Building blocks of an exchange

$OJRULWKPVLQ/RJLF

Š
Š
KWWS$OJR/RJLFFRP
Published by Mondo Visione

Hamish McArthur
Business Development Director
Mondo Visione Ltd.
Central Point, 45 Beech Street
London EC2Y 8AD

mondovisione.com
T +44 207 953 9800
M +44 7906 147 527
F +44 207 953 9798

Graham Taylor | Gavin Brightman

Graphic Design

cargocollective.com/mrgt
[email protected]

Mondo Visione Building blocks of an exchange 3

4 Building blocks of an exchange Mondo Visione
Contents

7 Introduction

Introduction:
8 
Technical building blocks of an exchange

10 The matching engine

12 The ultimate test

Index solutions for China’s

14 
domestic equity market

16 Data in motion

18 Building a CCP

20 Powering exchange integrity

Algo-Logic Systems’ FPGA solutions

22 
accelerate exchange infrastructure

25 Company profile: Devexperts

Mondo Visione Building blocks of an exchange 5

 Introduction

The business of indices for

exchanges is a core strength
and a key driver of growth.
Herbie Skeete, managing director, Mondo Visione

introduction
Financial markets have always been volume marketplace, and for the venues themselves
influenced by changes in technology, but the current to gain insight into how data is being consumed.
rate of change is unprecedented. These changes create The volume of data is set to increase further as the
new opportunities for investors and intermediaries, expansion of internet connectivity to an ever greater
and present new challenges for regulators. number of devices - the ‘Internet of Things’ - creates
Technology is a great market enabler, a requirement for feedback control loops, increasing
especially for emerging markets, as it allows messaging output considerably.
them to leapfrog the legacy structures found While operational data provides increased control,
in developed markets. However, technological data is also the raw fuel for indices. The business of
innovation benefits investors everywhere in terms of indices for exchanges is a core strength and a key
greater access and cost savings. Improved access to driver of growth. Technology allows index providers
markets further democratises the investment process. to bring indices on board quickly to meet market
For hundreds of years, physical paper documents demand or to respond to competitive threats and the
and human beings dominated securities operations. challenges posed by increased regulatory oversight of
Technology has done away with much of the physical indices and benchmarks.
paper and now dominates markets everywhere. To operate efficiently in the fast moving
However, technology and communications have interconnected marketplace of today, market
transformed market structure driven by a tsunami participants are asking for technological solutions
of data which needs to be disseminated and analysed. which are flexible, innovative and cost effective
Sophisticated IT allows regulators and market allowing them to evolve within the global markets eco-
operators to make sense of this high-speed and high- system as regulations and technology develop. MV

Mondo Visione Building blocks of an exchange 7

Introduction

Technical building blocks of an exchange

Exchanges are venues that bring together buyers and mean that for trading platforms to be fit for purpose
sellers to effect a trade. Central to an exchange is the they need to be scalable, with high reliability and
trading system on which trades are executed. capacity. Above all trading systems need to cope
All exchanges have rules to ensure the integrity, with the volumes and transaction speeds of today’s
efficiency and fairness of the marketplace and have markets, and properly implemented and designed if
a department responsible for managing all aspects we are to avoid more ‘flash crashes’ and market
of regulation relating to the marketplace. This will glitches.
normally have sophisticated surveillance tools
to maintain an orderly market through real-time
Testing
monitoring of trading patterns that could indicate
breaches of the exchange or regulator trading rules. Markets still worry about technology glitches after
In today’s world, trading in many instruments the May 2010 ‘flash crash’. Regulators and market
is largely automated, providing greater scale of operators in reaction to concerns about the strength
action. Reaction speed is determined by latency, not of critical market infrastructure are demanding that
wit. Error and manipulation can also be scaled up. measures be put in place to reduce the occurrence of
Systems are tested to destruction and are housed in systems issues and improve resilience when systems
specialised environments engineered for low latency problems do occur. Resilience is provided through
and exacting reliability standards. system architecture but also through thorough testing
The trade cycle is made up of the following activities: of every element in the system through every path.
pre-trade decision, order routing, order execution, Thorough testing demands not only excellent technical
matching, clearing and settlement. As crucial as the knowledge but also domain knowledge of the markets.
pre-trade and trading itself are the post-trade processes
of clearing and settlement. After a buyer and seller
Indices
effect a trade in a financial instrument, only when
instructions from the counterparties are matched Indices measure the performance of a market or
up, and checks made that the securities and cash are sector, providing the underlying basis for many
held by buyer and seller is the settlement process is exchange-traded products, allowing the market that
completed. provides index data to profit from the data supplied
For the pre- trade decision to take place one needs to an asset manager offering an exchange-traded
to know the prices prevailing in the market for the fund. Exchanges either design or manage indices
financial instruments being traded. Market data, themselves or else they collaborate with an index
the bid and offer quotes, last trade and volume group to manage the indices and to maximise the
information distributed by exchanges in real-time reach of the indices to global investors. The integrity
or delayed, either directly or through specialist of benchmarks and indices is vital to the global
information providers. Market data is one of the key financial system.
sources of revenue for an exchange and is also a
means of attracting liquidity to an exchange.
Post-trade
Specialised network providers bring together
trading partners from around the world. Such Post-trade services are crucial for the functioning of
networks have ended the tyranny of distance so that an efficient financial system effecting the transfer
exchanges can now be local, regional or truly global. of ownership and cash after a trade has been made.
In this market overview primer we, through our Technology has effected fundamental changes in the
partners, will illustrate the processes outlined above. infrastructure of the financial services industry driven
by regulation and the use of emergent technologies
such as the application of blockchain technology.
Trading engine
Central to a trading venue is the trading platform,
Surveillance
typically a central limit order book on which
counterparties’ orders are matched. Market Trading venues have a responsibility to maintain
fragmentation and increased market complexity an orderly market through real-time monitoring

8 Building blocks of an exchange Mondo Visione

 Introduction

of orders, trades and quotes to identify market frustration that many businesses experience. Freeing
manipulation, insider trading, layering and other that information, and making it usable alongside other
breaches of the exchange trading rules or regulations. measures, can present real opportunities. Exchange
Technological advances have significantly affected the operators and post-trade service providers face rapidly
global securities market accelerating the movement evolving business dynamics and competition. Seizing
of capital across international borders markets more opportunities is therefore immensely important and
integrated. These technological advances have made to do so market operators need to see a fuller picture
the job of maintaining fair and orderly markets more using the data they already hold.
difficult, but at the same time have given rise to
innovative solutions to policing markets.
Hardware
For the past 50 years computing power has doubled
Market data
every two years. This trend, which is dubbed Moore’s
The market-data industry is going through law, is transformative because it means computing
unprecedented change. Irrespective of whether you power is increasing exponentially. Markets have over
are a classic information vendor, a trading venue, a the last decade experienced a revolution driven by the
broker or a niche information provider, this change Moore’s Law rate of change provided by technological
will affect your business. innovation and through regulation.
New technologies, new business models and
regulation are some of the disruptive forces at work.
Data centres
Market data consumption patterns are also changing.
This impacts trading vendors and information In today’s environment where financial institutions
vendors alike. have to become more agile as well as to keep tight
There is an inexorable trend towards voracious control on costs, trading venues often choose to
consumption of data by machines rather than being outsource their data centres to specialist providers
viewed on screens. These machines require richer who can provide state-of-the-art but cost-effective
data from a greater number of sources and across technology. Today’s high-tech data centres bring
more asset classes. together trading venues and market participants in
The key market data dynamic is the bundling a low-latency environment, supporting the evolving
of analytics and data is creating a fundamentally needs of the global financial services industry.
complicated ecosystem which will allow numerous
small firms to flourish. This will mean that in the
Networking
medium term the large dominant players will need to
right size. How they do this will be one of the more
Financial markets are constantly evolving, and market
interesting business stories in the coming years.
participants are always looking at ways to efficiently
access new counterparties, geographies, liquidity
Apps venues and trading services. Market participants
need to quickly connect and engage within the
In today’s world of interconnected market-places,
financial ecosystem. Trading venues whether local,
data dominates. Digital data is part of every aspect
regional or global need to connect to buyers and
of our markets. Making sense of this new reality
sellers. As markets become ever more interconnected,
is challenging, we need flexible yet cost-effective
connectivity between a trading venue and its clients
solutions for multi-asset trading, portfolio analysis,
is crucial. MV
decision support, compliance, market data and all the
other components interconnecting global financial
markets.

Business intelligence
To have information, but to be unable to use it, is a

Mondo Visione Building blocks of an exchange 9

The matching engine
by Feroz Cader, vice president, Head of Trading Systems,
MillenniumIT, Technology Services LSEG

The matching engine is the core of any trading a view and control into the system for reference
venue. In its simplest form, it facilitates the electronic data, client connectivity, matching and market-data
matching of orders from buyers and sellers, in a fair and across multiple markets and asset classes should be
orderly manner, within a regulated framework. This a standard operability feature along with automated
simple function however, sits at the centre of an release deployment and rollback capability.
increasingly complex and competitive global financial
system. It brings about many interesting challenges;
Multi asset and multi-micro
whether you are setting up a brand new exchange in an
extremely competitive developed market, upgrading the market
legacy systems of an existing exchange in a developed A matching engine should be capable of trading
market, or introducing electronic trading for the first multiple asset classes in different market structures
time in a frontier market. Building the matching engine on the same platform. This design feature allows for
requires skill as much as it requires experience and harmonisation across platforms and other systems
capacity to deliver. It is a highly specialised niche from pre- to post-trade. Even if a trading venue
requiring significant engineering expertise in software operates multiple segregated instances for different
design, operability, latency management and asset classes, a matching engine’s capability should
connectivity, combined with a deep understanding of allow a common code path to be deployed across these
trading across multiple asset classes and risk systems. This can significantly reduce maintenance
management capabilities. costs compared to maintaining separate systems for
multiple asset classes.
This is equally true for frontier or emerging
Availability and quality at source
market exchanges where running separate matching
The matching engine is mission-critical software engines across multiple asset classes has proven to
that global financial markets depend on to provide be prohibitively expensive, especially when such
participants with the confidence to execute their markets are in the early stages of launching asset
strategies. It must remain glitch-free and provide as classes beyond cash equities, and as such, trade very
close to 100% uptime as possible in a competitive low volumes.
environment while at the same time being flexible
enough to respond to regulatory change in near
real-time. Managing that change is much akin to A matching engine should
replacing the wheels of a moving truck while keeping
the truck on the road. be capable of trading multiple
The ability to deliver such quality and availability
starts at software design. Quality-at-source and
asset classes on the same
fault-tolerant architecture are critically important platform.
to keep the cost-of-ownership low and to minimise
opportunity costs from a time-to-market perspective.
Increasingly, new regulations are demanding
greater transparency for asset classes that are traded
Operability
over the counter (e.g. FX, IRS etc.). Having the
Even a defect-free system is prone to operator error. capability to on-board new asset classes with different
A typical reactive solution is automation. However, trading models (micro-structures) on the same system
there is no such thing as 100% automation and the will bring cost savings to the organisations leveraging
operations cycle is not limited to the start and end of this opportunity.
trading. It begins at the point of release deployment into
production. Operability should be a key architectural
Latency
building block and not an afterthought. A single
permissions-controlled dashboard that provides The arms race to be the fastest may arguably be over. It

10 Building blocks of an exchange Mondo Visione

is no longer a significant differentiator given that most management should be a key consideration in the
systems are capable of microsecond levels of latency. current environment of risk regulation being imposed
However, to a tier-1 market, consistency in latency is on trading venues. There are two types of risks that
an important factor and in most cases still a challenge can be managed by matching engine software, namely
due to latency peaks or outliers. Reducing outliers will price/size errors (also known as fat finger errors) and
limit the uncertainty faced by trading algorithms and exposure-related risks.
in turn minimise the risk of algorithms misbehaving. A price- or size-based error could negatively impact
Improving latency can help trading venues do more a fair and orderly market. A trade with an erroneous
with less in terms of the hardware footprint and risk price or size could cause market-wide panic and
management, even if single digit microsecond level adversely affect trading algorithms and valuations. At a
latency is not important. minimum, matching engines should be able to manage
this risk through price bands and order size checks.
Connectivity
A key aim of emerging and frontier capital markets A trade with an erroneous
is increasing liquidity, but barriers to entry for global
investors tend to be high. It is important during a price or size could cause
market’s evolution, to ensure technology facilitates market-wide panic and
rather than prevents increased investment inflows.
Industry standard connectivity both at order entry adversely affect trading
and market data level is critical to ensuring the
matching engine allows low cost and easy access to
algorithms and valuations.
the trading venue when hurdles to foreign ownership
are relaxed. Exposure risk is the risk of a default. Regulation and
The fight for liquidity is not limited to emerging and the complexity of unwinding a trade, post-execution,
frontier markets. In mature markets, very few trading is shifting the need to perform risk management
venues operate in a single market environment. upfront, while latency improvements are making this
Price formation is decentralised and as such, market possible. A matching engine should have the ability
participants need the ability to connect to multiple to be ‘participant risk aware’ by integrating with a
exchanges to provide best execution. They can do risk management system to maintain real-time risk
that by maintaining individual connections to each thresholds per participant without compromising on
trading venue but this can prove costly given the rate its matching performance.
of change in each market. Trading venues should
instead look to provide a single standard interface
Experience is key
and shield its participants from the need to maintain
multiple order and market data interfaces. To do this, Competition among exchanges for market share
the gateway framework of a matching engine needs to and new regulations are driving change at trading
be multi-venue capable while maintaining consistent venues at exceptional speed. Deploying a matching
microsecond-level latency without compromising engine is labour intensive. It is vital that a matching
functionality of any connecting venue. Further, the engine provider has both scale and a proven product,
ability to normalise connectivity across venues will along with experience and process efficiency to
be a differentiator with developed as well as emerging support a trading venue’s evolving requirements. As
markets given an increasing trend to create regional more venues turn to specialist providers they can
cross-border trading partnerships between venues. benefit from sharing the cost of development across
multiple organisations at a fraction of the cost, while
accessing innovative ideas from multiple sources
Risk Management
and simplifying member connectivity by leveraging
A matching engine’s ability to provide risk standard protocols. MV

Mondo Visione Building blocks of an exchange 11

The ultimate test
Iosif Itkin, CEO, Exactpro, London Stock Exchange Group

The modern society expects quality and client on-boarding and certification. Clients connect
reliability from the latest exchange platforms. to the exchange platform and perform the necessary
Thorough verification of exchange technology is not steps, while the tool captures these activities and
possible without adequate software testing tools. produces pertinent client self-certification reports.
Understanding their evolution can help deliver rapidly One can create a good functional testing tool and a
evolving advanced exchange systems.
Test automation tools allow simulation of connected
automated systems, running tasks in an unattended Passive testing tools can
mode and generating load as in real-life market
environment. The process of selecting a tool to test help obtain valuable results
a trading or post-trade system usually starts with about the system and the
looking at generic testing solutions widely used in
other industries. No one has been fired for using test library coverage.
standard commercial solutions, but the problem with
most of them is that they are targeted at desktop and scalable load-testing tool in order to generate hundreds
web interfaces. However, integration with trading APIs of thousands of messages per second and stress one’s
usually requires developing connectivity and session platform to its limits.
libraries as well as a framework that would allow to However, there is another area in between — intermittent
uniformly define message contents and expected defects that do not manifest themselves every time the
results. Essentially, one will need to build one testing tests are performed, potential black swans.
tool on top of another. Such defects are at the confluence of the functional
and non-functional testing. To trigger them, one needs
to repeat the tests many times and under load, and
Exchange testing tools evolution
they can be easily missed in the enormously large
load testing logs. These defects might not have severe
technical consequences, though.

Ideal passive testing tool

There are a few specialised testing tools available that

already support FIX, FAST, ITCH, binary exchange
protocols and have expandable architectures. The user
specifies what to send into the market and what the
expected result from every gateway should be. This
process is called active testing. If the tool fits the purpose
and the QA team is capable enough, thousands of What characteristics should a passive tool have to
test cases to validate most of the exchange platform’s capture defects at this point? It should be scalable, stable,
functionality under test can be prepared and maintained. have minimal impact on the system under test; it should
The execution of huge regression libraries produces collect all events from all components in the system; it
massive amounts of data. Additional insights can be should efficiently capture all of the data for analysis;
obtained from analysing these large data sets. The it should notify the user when something goes wrong;
next evolutionary step in test tools is passive testing it should support the ability of the user to configure
where one uses non-intrusive test tools that capture scenarios that trigger such a notification; finally, when
all messages generated in the system, normalise them, something indeed goes wrong, the user should be able to
and store them for processing. Passive testing tools can restore the transactions around the event.
help obtain valuable results about the system and the An ideal passive testing tool for exchanges functions
test library coverage. The Exactpro team has also found exactly like a production monitoring / market
that passive test tools are invaluable in supporting surveillance solution.

12 Building blocks of an exchange Mondo Visione

Production monitoring system Production algo-trading engine

A good active testing tool should be scalable, stable, Is that it? Can the test tools of the next generation be
connected to many test environments, and easily just algo-trading and market-surveillance platforms?
configurable. It should be able to validate data Not really. There are two distinct reasons for that.
consistency across various feeds, store the audit trail, First, imagine trying to fit several algo engines and
and be capable of executing complex scenarios. an additional monitoring solution into your budget.
It will substantially increase the cost of testing. The
other limitation in using ordinary real-life systems
Ideal active testing tool
is that their intention is to do what is right. However,
the purpose of testing is to find what is wrong. Testing
tools of the next generation will be very much like
production systems, but deformed and mutated to
decrease the hardware footprint and to enable all
possible permutations.
In a more distant future, there might be test tools
based on machine-learning techniques. However,
they can be enabled only if the test tools of the next
generation are scalable and mutable. As singularity is
not here yet, human intelligence is still required to create
and operate test tools and large regression libraries. The
best part of any testing solution is the human mind
behind it. Exactpro builds software to test our clients’
software and has developed tools for trading, clearing
The above is clearly an algo-trading system. Instead
and surveillance systems testing. We constantly work
of running Excel in and getting Excel out, connect
on evolving our test tools, so if one ever needs a testing
something that behaves like a market participant with
tool of the next generation and/or a comprehensive
an algo engine.
quality assurance process, we can help. MV

As singularity is not here yet, human intelligence

is still required to create and operate test tools and
large regression libraries. The best part of any
testing solution is the human mind behind it.

Mondo Visione Building blocks of an exchange 13

 Index solutions for China’s
domestic equity market
by Eddie Pong, director,
Research and Analytics, FTSE

International investors’ access to the Institutional Investor (RQFII) programme in 2011, the
domestic ‘A-share’ segment of China’s equity market simplification of the R/QFII application process, the
was heavily restricted until the introduction of the rapid growth in the approval of R/QFII licences, the
Qualified Foreign Institutional Investor (QFII) extension of the R/QFII scheme to other asset classes
programme in 2003. In the first few years after the and the resulting expansion of the aggregate R/QFII
launch of the programme, China’s regulators granted quota are all testaments to this policy shift.
QFII quotas at a slow and steady pace. As of March 2015, 414 QFII and RQFII licenses
However, in the past few years, China has accelerated have been granted, a 25 per cent increase compared
the liberalisation of its domestic capital markets. to last year, and the total quota approved increased by
The launch of the Renminbi Qualified Foreign 46 per cent, reaching US$ 125.3 billion.

Historical R/QFII Aggregate Quota by Investor Type

140 450

400
120

350
100
300
Quota (US$ bn)

No. of Licenses
80 250

60 200

150
40
100

20
50

0 0
N 06

N 09
N 08
N 04
ov 3

N 05

N 07
M 06

M 09
M 08
M 04

M 05
M 03

M 07

N 10

N 14
N 13
N 12
M 10

14
M 13
M 12
N 11
0

M 11
ay

ay

ay

ay

ay

ay

ay

ay

ay

ay

ay

ay
ov

ov

ov

ov

ov

ov

ov

ov

ov

ov

ov
M

Asset Management Companies Securities Companies Sovereign Wealth Fund

Commercial Banks Insurance Companies Pension Fund
Other Investors No. of Licenses (RHS)

Source: FTSE Russell, CSRC, SAFE, data as at 31 March 2015

14 Building blocks of an exchange Mondo Visione

 A further step in the liberalisation of access to adjusted market value, assuming no quota restriction;
China’s A-share market came with the introduction of • Customised indexes based on an investor’s own
the Shanghai-Hong Kong Stock Connect programme QFII/RQFII allocation.
in November 2014. The programme permits The index series follows the existing FTSE GEIS
international investors with Hong Kong brokerage methodology, where China is reviewed as a single
accounts to access A-shares and mainland investors entity.
to access Hong Kong-listed shares, subject to separate In the diagram we show how the FTSE Global China
quotas (initially US$ 47.9 billion for ‘northbound’ A Inclusion Index Series is constructed by adding
trading and US$ 39.9 billion for ‘southbound’ the China A-share market available to international
trading). The programme is open to both institutional investors to the FTSE All-World and FTSE Global
and retail investors. Small Cap indexes.
Allocation = QFII + RQFII Quota
Existing GEIS (Quarterly Update)
A-share benchmark solutions for
international investors Large Large

Despite this trend towards the liberalisation of Mid Mid

China’s domestic capital markets, A-shares are so

far ineligible for inclusion in FTSE’s Global Equity Small Small

Index Series (GEIS). This is because China A-shares

Developed Adv. Sec. Developed Adv. Sec. China A
market does not yet meet the requirements for market Emerging Emerging Emerging Emerging

accessibility, free capital repatriation and clearing and FTSE Emerging Index FTSE Emerging Markets China A Inclusion Index
settlement cycle that are governed by FTSE’s Country FTSE Emerging All Cap Index FTSE Emerging Markets All Cap China A Inclusion Index
FTSE All-World Index + FTSE All-World China A Inclusion Index
Classification Committee. FTSE Global Small Cap Index + FTSE Global Small Cap China A Inclusion Index
Nevertheless, many international investors have
now received substantial quota allocations to enable
FTSE’s China Expertise
them to access the A-share market and, increasingly,
require index solutions to reflect their levels of access. FTSE has a long-standing presence in China and in
As a result, FTSE has launched the FTSE Global 2015 celebrates the fifteenth anniversary of its first
China A Inclusion Index Series . This index series has Chinese equity indices, launched in 2000 under a
two primary objectives: to provide a comprehensive partnership with Xinhua Finance Limited.
range of benchmarking solutions for these market In 2010, FTSE acquired full ownership of these
participants and to prepare for the potential inclusion indices, which were renamed the FTSE China Index
of A-shares in FTSE’s standard indexes. Series. FTSE’s China indices have become the
The index series also aims to serve as a benchmarks of choice for institutions worldwide and
benchmarking tool for international investors underlie a wide range of index-tracking mutual funds,
wishing to access China A-share market, to reflect the ETFs and derivatives.
prevailing levels of market access for international As the leading index provider for China, FTSE
investors and to be transparent and easy for users to maintains offices in Hong Kong, Beijing and Shanghai,
understand. supporting clients in the region with a full suite of
The FTSE China A Inclusion Indexes provide market research, distribution, support, product issuer and
participants with following benchmark options: asset owner services.
• Global benchmarks with China A-shares included As at end-December 2014, ETFs with US$24.4bn
and weighted by the aggregate approved quota in assets worldwide used FTSE China indices as their
(including QFII/RQFII); reference performance benchmarks, including ETFs
• Global benchmarks with China A-shares included on the flagship FTSE China 50 and FTSE China A50
and weighted by free float and foreign ownership- indexes. MV

Mondo Visione Building blocks of an exchange 15

Data in motion
Creating an intelligent data management infrastructure
to maximise strategic insight and efficiency

For market operators, the ability to derive Executives need to base strategic decisions on a
real-time market intelligence from various forms of complete picture, but data alone does not provide
data is an essential ingredient for success. Internally, it that. Relevant pieces are scattered across the data in
helps them understand what is working and what is various shapes and forms, making them hard to find.
not, and it allows them to be proactive in making When data is structured and ordered, it becomes
improvements. Externally, it helps them understand documented, accessible information. But even
their customers, which leads to deeper relationships then, information does not automatically lead to
and better services. knowledge.
Imagine being able to provide senior management Events and information that are stored in disparate
with up-to-the-minute reports following a new systems can be turned into knowledge by applying
product launch. Just think of the scope for improving analytical models and exploratory visualisation to
satisfaction by sending alerts when customers’ order- correlate them and find hidden patterns in the data.
to-trade ratios are nearing a point where they will Visualisation helps market operators quickly identify
breach thresholds. anomalies and outliers in large datasets. When
In reality, market operators often struggle with the visualisation is overlaid with business knowledge and
real-time analysis that underlies market intelligence. expertise, market operators can gather new insights
Today even smaller venues must cope with enormous
volumes of historical and streaming data. Since data is
recognised as a strategic asset, they are collecting and The goal is to take raw
storing more than ever before. Reference data, market
data, trading data, account data and counterparty structured and unstructured
data all reside in their core systems and various data and turn it into
silos, but it is in several formats. It must be extracted,
cleansed, normalised and made accessible to an array information, knowledge and
of business applications in a cost-efficient manner. It
is important to capture the data in its most granular
wisdom, and ultimately
form to avoid loss of quality. monetise it.
The goal is to take raw structured and unstructured
data and turn it into information, knowledge and
wisdom, and ultimately monetise it – and that is not that enable strategic decision making and accelerate
an easy feat. business growth.
This can be done without sacrificing quality or
incurring excessive costs. An open data management
framework can be implemented at the foundation of
the enterprise architecture. As such, it can support
Wisdom all core systems that enable the trade life cycle, from
the order through to trade, clearing and settlement.
Essentially, the data management framework serves
Knowledge as an ‘information bus’ that collects all the data
from various operational systems and external
sources of relevance, and transfers it to a centralised
Information infrastructure where it can be managed. Since it is
‘open’, the market operator is free to use its preferred
method of data storage and add applications as
Data necessary.
Think of this as a paradigm shift from the

16 Building blocks of an exchange Mondo Visione

 Connecting data and Consider the following examples:
allowing users to see it in their • Data from a trading engine, the billing system and
an external market data feed could be aggregated
own context can be incredibly and cross-referenced to analyse performance. Market
valuable. If data is correctly operators could identify a shift in demand for specific
products and determine the cause. They can identify
articulated it can reveal which products generate the most revenue and
attract the most liquidity. Armed with this market
patterns that were previously intelligence, they can make strategic decisions about
hidden and make otherwise product design and fees, and determine the optimal
way to allocate resources.
unknown relationships visible. • Historical data can be cross-referenced with order
and trade data to determine the cause of increased
latency at a specific time. With this insight, market
traditional data warehouse concept where data is at
operators can have an informed dialog with
rest: it is simply loaded, transformed and extracted,
participants to help achieve an optimised experience
and decisions are made based on stale data. In the
and increase customer satisfaction.
data management framework, however, data is in
To compete in today’s data driven, globally
motion. Live structured and unstructured data from
interconnected environment, information is key
the markets, internal business systems, social media,
to gaining an advantage and staying agile and
mobile applications, the news and other sources are
relevant. Moreover, global regulators are requiring
fed into the framework and combined with historical/
various data sets to be made available upon request
static data. Then it can be sliced, diced and refined
and often within tight timeframes. While accurate,
using complex event processing. It can be sequenced
accessible data is an undeniably crucial component,
and replayed. Market operators can run real-time
obtaining that data in a timely manner – especially
streaming queries and combine historical queries
across multiple formats and sources with infinitely
with streaming data. Calculations can be made on the
increasing volumes – continues to be a challenge
fly for predictive analysis.
for many market operators. Startup marketplaces
The framework provides a unified API for
can avoid falling into this trap by putting a data
integration, ensures relational integrity and resilience,
management framework on the IT roadmap from the
and stores high velocity data without delays.
beginning, while incumbents can solve an existing
Importantly, the data can be tapped and collected
problem by adding it to their enterprise architecture.
without affecting other systems. When the data is
Market operators can start by implementing a data
visualised on dashboards, executives can make more
management solution and collecting all data in its
effective decisions.
most granular form. Before long, they will realise
Connecting data and allowing users to see it in
the hidden value in the data and begin applying data
their own context can be incredibly valuable. If
management techniques in various new ways to gain
data is correctly articulated it can reveal patterns
business insights they never envisioned before. Finally,
that were previously hidden and make otherwise
they can leverage data to develop innovative offerings
unknown relationships visible. There are myriad
and expand customer value while potentially creating
ways to use market intelligence to improve
new revenue for the marketplace. MV
performance and customer satisfaction, as well
as comply with regulations. Theoretically, market
operators just need to identify what needs to
be analysed and which data are needed to do it.

Mondo Visione Building blocks of an exchange 17

Building a CCP
The financial crisis revealed that the finance transactions.
industry’s approach to risk management was flawed. Through portfolio margining (combining positions
Since then, new regulations have been created to push in different products to offset margin), CCPs can help
more activity onto regulated marketplaces and central members reduce their overall margin requirement.
clearinghouses (CCPs), which are deemed safer than Depending on the degree to which collateral
bilateral trading and clearing arrangements. optimisation services are effective, there may be
CCPs manage, mitigate and mutualise risk by significant reductions in the amount of overall
serving as the counterparty on all trades (the buyer to collateral required.
all sellers and the seller to all buyers). The members Central clearing enables participants to trade and
put up capital in a mutualised guarantee fund. They clear with everyone without having to worry about
also post initial margin when they take a position, direct counterparty risk. CCPs have a lower risk profile
and variation margin when the market moves against than traditional banks because they always have
them. If a firm cannot meet a margin call, its positions balanced positions, except in a default situation. They
are liquidated. When necessary, CCPs can tap into are in the business to manage risk, not to speculate.
their shared capital and their own capital. Moreover, they must comply with regulations that
By clearing through a CCP, firms can benefit
from multilateral netting, where several parties’
transactions are summed instead of being settled CCPs have a lower risk
individually. Multilateral netting not only reduces
credit, operational, settlement and liquidity risk, but it profile than traditional banks
also reduces firms’ balance sheets. It is a cost-efficient because they always have
mechanism for managing initial and variation margin
because firms have one net position and one collateral balanced positions, except in
pool to cover their margin calls. Central clearing also
leads to a reduction of regulatory risk capital charges.
a default situation.
CCPs can drive operational efficiency by enabling
members to optimise collateral management using require certain levels of capital and risk management
real-time information, straight through processing processes. Markets are safer as a result, which attracts
and the auto allocation of funds on a client level. They more global participants, boosts liquidity and drives
can help members identify surplus collateral, free up competition on exchanges.
valuable liquidity and maximise value from securities Four key components are needed to build a CCP.

18 Building blocks of an exchange Mondo Visione

The first is proven and tested technology. CCP processes, and it collects the right amount of margin
technology needs to be flexible because both startup and default fund contributions. Only products that
and established organisations need room to grow and can be properly risk managed and liquidated in a
add new products as they move toward a multi-asset, default scenario should be put into the CCP.
multi-currency and real-time environment. Fourth is the CCP capital structure. CCPs have
After a trade is executed, the trade title is transferred a waterfall mechanism whereby various levels of
to the CCP. To this end, they need a system to handle safeguards protect against the default by one or
novation (CCP matching) and netting. They need a more members. The waterfall should include both
real-time risk management tool so they can constantly participant and CCP capital and be designed to create
evaluate their risk. In the event of default, CCPs must the right risk incentives to ensure a functional default
be able to close out a portfolio at the market prices management process.
assumed in their model. Members post margin with the CCP, and the CCP
To do portfolio-margining, CCPs must be able has a default or guarantee fund. If the guarantee
to run a series of rigorous calculations to identify fund levels are higher, then the margin levels tend to
which products are correlated, demonstrate the be lower. CCPs model and analyse the size, nature,
strength of the correlation, and determine whether diversity and volatility of their members’ portfolios
the correlation is stable over time. They also need to
monitor performance and back test to ensure that the
correlation will hold up in current market conditions. Only products that can be
CCPs need a system to hold various collateral
assets including bonds, equities and cash. It must be properly risk managed and
possible to segregate collateral in customers’ accounts
so their assets are protected should a clearing member
liquidated in a default
default. By using real-time prices and positions, CCPs scenario should be put into
minimise the margin requirements they impose on
their members and customers.
the CCP.
Data management is critical. Customers want to
trade and clear more products faster than ever before,
so they strike the right balance between margin and
which introduces a new level of complexity, especially
guarantee fund levels.
in margining. CCPs need graphical user interfaces and
The risk management framework of a CCP
gateways to aggregate data flowing in from various
comprises several layers, all of which are fundamental
exchanges and liquidity pools. New, updated and
to building a strong organisation. A few best practices
historical data enables them to perform sophisticated
should be kept in mind.
risk calculations that determine margin levels, and
When building a CCP, it is critical to develop
influence the degree of leverage and the amount of
prudent member admission criteria and ensure it
capital to be committed to positions.
is implemented. Margins should be calculated to
Additionally, CCPs need various systems to handle
cover for the market risk inherit by the CCP. CCPs
treasury functions, reconciliation and corporate
should set strict guidelines and haircuts for the types
actions.
of collateral they are willing to accept. Finally, they
Second is regulatory status that delivers capital
should stress test participants’ portfolios, strike a
efficiency and complies with global standards for
balance between margin and guarantee fund levels,
CCPs while meeting local or regional demands. CCPs
as well as determine how the pieces of the waterfall fit
are systemically important financial institutions, so
together and when margin levels should change. MV
they must be certified to ensure they meet regulatory
standards. In most legislations they also must be set
up as separate legal entities and operate as standalone
companies with their own board and capital.
Third is competence and know-how. Although
central clearing is less risky, it is not risk free. It is critical
that the CCP is operated correctly with appropriate

Mondo Visione Building blocks of an exchange 19

Powering exchange integrity
By Robert Lang, vice president, head of Product
Management, Risk and Surveillance Solutions, Nasdaq
and Tony Sio, head of Sales and Account Management,
SMARTS Market Surveillance, Nasdaq

Fair markets are free from manipulative and between venues. A participant may try to manipulate
deceptive behaviour, with clear rules that are enforced the price of a particular stock in order to make a profit
fully and without bias. Orderly markets are those on an options contract on that stock. Alternatively
which are reliable, without unreasonable price the manipulator may be profiting through over-the-
fluctuations or errors. Without these characteristics counter contracts or other linked products.
both investors and potential issuers will stay away, In addition to possibly missing market abuse,
starving the exchange from both directions. An compliance staff may be wasting time and resources
effective surveillance program plays a key role in investigating scenarios based on single venue, single
every exchange’s future. asset class data. Behaviour that appears unusual in
Recent changes in technology, regulation and the context of one venue may be normal when viewed
global economics have revolutionised the securities against the backdrop of multiple markets or related
markets. Dynamic exchanges have adapted quickly instruments. Hence, the ability to view consolidated
and launched new products and services to meet trading data from multiple sources in a single view
new customer demands. But to remain successful, has become essential, as has the ability to review alerts
exchanges must also ensure that their surveillance in the proper context, and compare them to what is
program keeps pace with change and that they have happening in electronic communications, news and
the same abilities to view the market as the traders other areas.
they are monitoring. The minimum requirements
include a full consolidated audit trail and real-
Order-based analysis over
time, automated and cross-product/cross-market
monitoring. execution
Historically, surveillance has focused on market
integrity and market orderliness separately. Rule
Cross-market, cross-product
breaches as well as unlawful and dishonest conduct
surveillance relate to market integrity; errors, system stability and
Globalisation and fragmentation allow market market activity relate to market orderliness. The ability
participants to seek opportunities in multiple assets to reconstruct every aspect of the market at any point
and geographies, and take advantage of differences in time, and at the very lowest level of granularity,
in structure, pricing, latency, hours and order types – is the foundation for monitoring both areas. Trade
exchanges need to be able to detect these advantages. and top-of-book data is not sufficient to accurately
The effectiveness of a single market’s surveillance understand modern surveillance issues asmany types
efforts is reduced if the system cannot incorporate of manipulation cannot be detected at the trade level.
trading data for the securities across all the venues Regulators are devoting more attention to order-level
on which those securities trade and their related manipulation. With order-level analysis you can get
instruments. For example: closer to determining true intent.
• Layering (Spoofing): Analysts may be alerted
to layering of the order book on one market, but
Real-time control
without cross-market functionality, they may not
be able to identify that the same account is trading Exchanges cannot wait until the end of the day or the
on the opposite side of the market from the layered week to analyse trading activity.
orders on another venue. While historical analysis is still extremely
The markets are so inter-connected that surveillance important in determining abuse, analysts need to
cannot be done in a siloed manner. The economic be alerted instantly when a market event occurs. A
relationship between the traded instrument and failure to take action could unleash extreme volatility,
related instruments must be considered to identify spread contagion to other markets, cause huge losses
gaming and manipulative behaviour that may occur and erode investor confidence. A prime example is

20 Building blocks of an exchange Mondo Visione

the Knight Capital case, in which the company lost Beyond the surveillance system
about US$170,000 per second over the course of 45
Today’s surveillance systems can provide value
minutes. Real-time capabilities are critical to early
beyond traditional surveillance. The surveillance
detection of unusual trading patterns that could be
system provides the most accurate low-level store
potentially catastrophic breaches of trading rules and
of market information held by an exchange, and
practices.
advanced surveillance systems have flexible
visualisations and data mining tools on top of this
Technology information, which can be leveraged by other teams.
Market operations teams can leverage the
Modern surveillance technology uses data mining
investment to become more proactive in boosting
and pattern recognition techniques to monitor the
market control, like being able to reconstruct trading
markets. First, it combs through large amounts of
system events or to detect sudden surges/drops in
historical data and builds a profile of normal, day-
order activity.
to-day behaviour. Then it captures trading activity
Front-office teams can leverage surveillance
in every product listed on the exchange in real time,
systems to see exactly where orders have been placed,
and compares it with the norm to detect certain types
when and at what price, greatly improving the service
of events. Anomalies can be uncovered quickly, and
provided by the exchange to its members.
exchanges can immediately make a decision to launch
In conclusion, exchanges need an effective
a preliminary investigation or even halt the market.
surveillance program in place to ensure market
integrity, maintain investor confidence and grow
Data integration: Storing and handling data is a
the business. Technology is only one component.
struggle that has implications for surveillance. Data
Strong rules and regulations, effective exchange
volumes have grown from about 10 million messages
processes and continual education and staff training
per day in some markets to more than 250 million
are essential. As the market develops and matures,
as more orders are being entered and cancelled - a
the surveillance program also must evolve to
few seconds either side of an alert, they are probably
accommodate various players, different trading
covering about 20,000 events in a single instrument.
strategies and ever increasing volumes of data. MV
High speed data feeds containing huge volumes
from multi-asset sources need to be consolidated
into a single comprehensive view to create a clear
picture of normal market behaviour against which
to proactively measure deviations.
Proven in the world’s fastest and largest markets,
SMARTS market surveillance and compliance
Big Data Buzz: In surveillance, big data is not the
solutions from Nasdaq are the industry benchmark
next big challenge to conquer; it is a very current
for the global exchange, regulator and broker
affair. Trade data is important, but to accurately
community. SMARTS Surveillance technology is
understand market behaviour, systems need to
installed in more than 50+ national exchanges and
analyse other types of external information including
regulators and 100+ market participants across 65
news, economic data such as interest rate changes,
markets.
and even weather forecasts. Social feeds should also
be examined - while difficult for exchanges to gather
Contact Details: S
 [email protected]
this information, market participants are already
Business.nasdaq.com/TECH
greatly investing and exchanges should look at how
they can incorporate ‘big data’ into their surveillance
programs.

Mondo Visione Building blocks of an exchange 21

$OJRULWKPVLQ/RJLF

Š
KWWS$OJR/RJLFFRP

Algo-Logic Systems’ FPGA solutions

accelerate exchange infrastructure

A tremendous new technology shift is about networking systems and builds multiple FPGA
to impact the entire global exchange industry components for ultra-low-latency trading firms. It
infrastructure. Heterogeneous computing platforms provides a broad range of highly scalable and reliable
with field programmable gate array (FPGA) cores that run on off-the-shelf FPGA hardware
technology are implementing the next generation of platforms with high-speed Ethernet interfaces.
exchanges. FPGA technologies provide ultra-low- These platforms have been deployed worldwide in
latency, jitter-free operation, high throughput, and co-located trading datacentres.
consume very little power as compared to CPUs and
GPUs.
Trade-to-Tick System
FPGAs are widely used for real-time networking
and supercomputing application. FPGA technology Algo-Logic’s Trade-to-Tick (T2T) System receives orders
uses fine-grain parallel hardware to accelerate from market participants and disseminates market
exchanges so that financial transactions complete data to subscribers. Within the T2T system, the order
quickly and deterministically. Deterministic trading sequencer module assigns timestamps to orders and
infrastructure satisfies government regulations for forwards the orders to the matching engines. The T2T
fair markets and the faster hardware scales to handle System’s uniqueness lies in its ability to avoid delays
the traffic needed for global exchange connectivity and jitter caused by sequential software execution, bus
and forecasted trading volume growth trends. transfers, memory copies, and cache misses that are
With roots at Stanford University, Algo-Logic present in standard software-only servers.
Systems is the leading gateware developer for The Trade-to-Tick path starts with Order Management
Gateways that receive orders from market participants
over the network. Sequencers benefit from hardware
accelerated time-stamping of all the orders incoming
FPGA technologies provide from market participants. The Matching Engines use
ultra-low-latency, jitter-free hardware accelerated in-memory databases, called a
Key-Value Store (KVS), for maintaining open orders,
operation, high throughput, and aggregated order book building, sorting, as well
consume very little power as as matching of outstanding orders based on price,
timestamp, volume, and other priorities. Matched
compared to CPUs and GPUs. orders are distributed to clearing houses, order loggers,

Order Management Gateway Sequencer Matching Engine

Packet Inspection
& Filtering
Ultra-Low- TCP/UDP Distributed Aggregated Matching
Market Data Latency Offload
Protocol Parsing & Time
Key -Value Order Book of Open
Participants PHY+MAC Engine
Order Inspection Stamping
Store (KVS) Building Orders
Order Routing &
Load Balancing

Market Data Disseminator Distributor of Matched Orders

UDP
Ultra-Low- Ultra-Low- TCP/UDP Ultra-Low-
Market Data Latency
Multicast
Latency Offload Latency
Subscribers PHY+MAC
Offload
PHY+MAC Engine PHY+MAC
Engine

Figure 1 : Major components in an FPGA Accelerated Exchange

22 Building blocks of an exchange Mondo Visione

 $OJRULWKPVLQ/RJLF

Š
KWWS$OJR/RJLFFRP

The Trade-to-Tick System’s uniqueness lies

in its ability to avoid delays and jitter caused by
sequential software execution, bus transfers,
memory copies, and cache misses that are
present in standard software-only servers.
and to market data disseminators which multicast
events to market data subscribers.
The sub-components within Algo-Logic’s Trade-to-
Tick (T2T) System includes the Ultra-Low-Latency
(ULL) PHY+MAC, Transmission Control Protocol
Total logic-to-logic roundtrip latency
(TCP) Endpoint, Full Order Book with aggregated
in-memory matching engine, Protocol Parsers and Total fiber-to-fiber roundtrip latency
Low Latency Library.
Parallel Logic Packet Streaming Interface

Ultra-Low-Latency PHY+MAC
Stratix V

SERDES IO
Packet Source Transceiver
mac_10g_top
Reconfig
To achieve the lowest Ethernet latency, Physical (Avalon ST
Controller
Level (PHY) and Media Access Control (MAC) interface, 64 bits
mac_10g_tx
> 156.25 MHz)
functions are optimised for speed. Algo-Logic’s Stratix V Native SFP+Fiber
Ultra-Low-Latency (ULL) PHY+MAC receives and Monitoring logic
Statistics Counters PHY. PMA Module
(optional)
sends packets with a roundtrip latency measured Direct Mode Transcelver
in tens of nanoseconds. The ULL PHY+MAC design Packet Sink
mac_10g_rx
(Avalon ST
is compatible with multiple FPGA platforms that Stratix V
interface, 64 Bit
support high-speed Serializers and Deserializers > 156.25 MHz)
Transceiver
Reset
(SERDES). The MAC interfaces to FPGA logic via
Controller
standard 64-bit Avalon-ST or AXI4-Stream bus
standards. Figure 2 : Ultra Low Latency PHY/MAC

Ultra-Low-Latency TCP Endpoint

RX TCP TCP TX
Algo-Logic’s TCP Endpoint maintains reliable 10G 10G
Ethernet Endpoint: Endpoint: Ethernet
connections for data flowing directly into FPGA TX Client Client RX
MAC MAC
hardware. It supports opening, maintaining, and
closing TCP Connections. It has a latency of just 76.8 RX
10G TCP FPGA TCP TX 10G
nanoseconds from the time that a packet arrives at the Ethernet Ethernet
Endpoint: Application Endpoint:
PHY/MAC to the time that the order processing logic MAC TX Client Logic Client RX MAC
can start to consume that data. By using multiple
full duplex ports within a single FPGA, each device 10G RX TX 10G
TCP TCP
can handle hundreds of Gigabits/second of line-rate Ethernet Ethernet
Endpoint: Endpoint:
traffic. The implementation is portable between MAC
TX Client Client RX
MAC
Altera and Xilinx FPGA devices and compatible with
all widely deployed FPGA platforms from board Multiple TCP Proxy Configurations
vendors that include Nallatech, Terasic, Solarflare,
Bittware, Alpha-Data, and PLDA. Figure 3 : Ultra Low Latency TCP Endpoints

Mondo Visione Building blocks of an exchange 23

$OJRULWKPVLQ/RJLF

Š
KWWS$OJR/RJLFFRP

Aggregated order-book with Low Latency Library

order matching algorithms The Low Latency Library provides a set of
The order book implements the sorted matching modular components that are integrated together
function in an FPGA hardware device. In large to implement a complete trading system with the
exchanges, where scalability and redundancy are shortest possible time to market. These components
important, a distributed order book is used to include Internet Protocol (IP) packet decoding, TCP
handle large numbers of symbols and open orders. session processing, UDP datagram processing, and
A fast, in-memory database is implemented using a register interface that allows the hardware to be
an FPGA-accelerated Key Value Store (KVS). Each easily controlled from a software API.
Exact Match Search Engine (EMSE2) inside the Because Algo-Logic provides such a wide library
KVS transacts 150 Million Transactions Per Second of pre-built ultra-low-latency infrastructure
(MTPS). The KVS allows orders to be added, modified, components that utilise user-friendly C/C++ APIs,
and deleted with sub-microsecond latency. minimal effort is required for integration into
existing systems. These components are tested,
validated, and deployed worldwide across multiple
Order Book: AAPL exchanges.
Last Trade Block

Conclusion
Cur: Vol:

Top of the Order-Book Block

100 408.30 408.47 121 To compete effectively, exchanges are changing
the way that they build their infrastructure. Rather
Level-II Price-Level Book
Bid Size Ask Size
than continue making further painful efforts to
optimise legacy software systems, exchanges are
408.30
Illustration to come
100 121 408.47
408.28
408.27
44
75
300 408.57
408.92 36
turning to FPGAs to achieve higher throughputs
408.25 100 408.96 100 with deterministic latencies. Global exchanges that
408.24 100 408.10 1700 recognise these advantages are the ones that will
408.04 100 409.18 400
407.76 100 410.09 54 maintain their edge. As Charles Darwin said, “It is
407.35 400 425.00 110 not the strongest of the species that survives, nor
407.09 100 490.00 100
406.00 54 0.00 0 the most intelligent that survives. It is the one that is
most adaptable to change.” MV
Figure 4 : Order Matching

Protocol Parsers As Charles Darwin said,

Algo-Logic’s protocol parsers extract data from the It is not the strongest of the
order flows. The parsers cover multiple markets for
different exchanges across the globe. Parsers built in species that survives, nor the
hardware include support for all major worldwide
protocols, including FIX, OUCH for NASDAQ, XPRS
most intelligent that survives.
for DirectEdge, BOE for BATS BZX, ArcaDirect for It is the one that is most
NYSE ARCA, Arrowhead for TSE, CME MDP3 for
SBE, and Native Trading Gateway for LSE.
adaptable to change.
To facilitate new markets, an automated process
generates a new parser, enabling rapid deployment
of hardware in new markets and on-going support
for changes in existing markets. The protocol parsers, Contact Details: A
 lgo-Logic Systems, Inc.
combined with business logic, extract and process data http://Algo-Logic.com
as soon as fields of interest become available in the data. Email: [email protected]

24 Building blocks of an exchange Mondo Visione

Company profile: Devexperts

Devexperts has been offering market The HTML 5-enabled system offers a comprehensive
operators a revolutionary approach to technology set of desk-top terminals, mobile terminals and HTML
build since 2000. Originally a small software 5 web terminals across smartphones, tablets and PCs.
development company with expertise in broker
technology, Devexperts changed that industry with
Data: Capturing the market
platforms such as TD Ameritrade’s Thinkorswim. Our
expertise derived from the experience of supporting Two issues dominate data provision and analysis for
brokerage operations 24x7 - hence developing the market operators today: the drive towards greater
know-how necessary to also support exchanges’ transparency for all market participants, and the
technology requirements and provide superior pressure to improve market supervision. They both
servicing to their clients and members. Devexperts is require a higher level of information granularity than
currently participating in some of the exchange world’s previously needed, with a greater capacity to query
most strategic projects, working with electronic and analyse captured data.
communication networks (ECNs), alternative trading To support market operators, Devexperts’ platforms
systems (ATSs) and exchanges to support secondary allow them to look at real-time market data, historical
market operations via our unique, streamlined and data and reference data across symbols or in an
cost-effective approach to delivering IT solutions. aggregated view, according to the trading venue’s
The technology we use has evolved over the last 15 need – hence addressing the requirements for market
years – as has our approach to project management. transparency and reporting.
Having developed the platforms of its early investors, To deliver greater market transparency, traceability
Devexperts became savvy and built up its own internal and audit, Devexperts provides a data solution
development department, taking the knowledge with a capacity on par with the Consolidated Audit
accumulated over time and using it to develop the Trail (developed by the Securities and Exchanges
firm’s own line of products. Commission, in the US) and with a five-year proven
Devexperts offers a set of components, which can be record. The solution currently captures data from all
put together like Lego blocks to build the foundation of the US exchanges in order to provide a traceable log
for a solution. With some additional customisation, of what happened at any given point in the market.
that solution can be tailored to clients’ specifications. Data capture, storage and retrieval is available
Those building blocks exist for each possible function right out of the box- but it can have its pre-defined
- from a matching engine to a messaging bus. The functionality extended through customisation. The
system provides connectivity to internal and external model Devexperts uses is data agnostic since any data
information sources so that market data can be plugged provided to the platform is normalised. As such, the
in, normalised and then used within interoperable system is able to handle information from multiple
components - fully integrated, yet flexible. sources in multiple formats and protocols. Market
Building an exchange is a complex project, which data could be well defined and highly structured, but it
makes the componentised approach to technology can also be more open, so that the set of data captured
invaluable. The building blocks can be combined in is expansible - hence allowing for future requirement
different configurations to build different solutions, changes. For example, it might be expanded to
hence bringing flexibility and speed to the process increase the granularity of data around a specific
of customising solutions to any given challenge – characteristic, or it might take in new details such as
yet based on proven technology. While exchanges the requirements for trader identification under the
will typically look for assistance with their back Markets in Financial Instruments Regulation (MiFIR).
office and services, Devexperts can provide a full The market snapshot that the system provides is
front-to-back service offering, with a full set of front- accessible for the exchange’s internal processes but
end components - including exchange terminals it can also be made accessible to clients in order to
supporting different asset classes and displaying all of dramatically increase transparency.
the information that the exchange is giving out. It uses Since both the marketplace and its members have
Java-based desktop technology giving it the capacity the same information available to them, there is an
to run across operating systems, Mac or Windows, ‘official history’ of what has truly happened in the
without any impact on experience or performance. market removing doubt around market events and

Mondo Visione Building blocks of an exchange 25

supporting smaller firms in their capacity to keep up information via FIX messages. Alternatively this can
to increasing complexity and volumes of data. be a subscription-based service via which clients can
Given the current concern around volatility events, receive a subset of symbols from the exchange’s data
this ability is highly valuable. There would be no feed. Depending on configuration, the system can
question about what happened in a ‘flash crash’ deliver bulk feeds or subscription-based live feeds.
similar to the event of May 2010; to find out what
happened the exchange’s audit team would simply
Historical data
dig in and see who conducted what activities. This
functionality could also be utilised for a market abuse Devexperts has a unique historical data technology.
inquiry or meeting any other regulatory requirement. The firm has been storing Level 1 and Level 2 tick data
Most of the historical and real-time aggregated since 1 January 2010 from every American exchange
data are accessible through a single set of services across all asset classes: equities, options, futures,
and application programming interface (API). From indexes and FX. The amount of information captures
an operational point of view this means there is no is extensive e.g. quotes in the US options market,
need to build different interfaces to accept different which due to the number of instruments listed and
types of data. The user can just switch a couple of the amount of quotes given across US exchanges
parameters to change from historical to real-time and is among the largest feeds in the world. This data is
between aggregated or reference data, making it a instantly accessible in real-time and it is also available
simple model to use. Data types that are typically used from a historical database – retrievable from the
by an exchange can be pushed into the Devexperts moment subsequent to capture all the way back to the
normalised structure, making life easier for those who start of the data capturing period in January 2010.
need to develop technology solutions to run on top of As an out-of-the-box feature, the Devexperts
the exchange platform. Be it client systems or internal technology makes data available from any period in
systems, processing data or storing data, Devexperts the reporting history in a matter of seconds. Typically
has that technology. this is sold as a service, providing clients with access
Our technology is built entirely in-house – as to the cloud platform, with data and a set of client-
such, there are no third-party production sites. For facing services, APIs and subscription services. One
historical data storage systems there are no third of the main features it offers is a market replay from
party products and therefore no third-party license the cloud, which is akin to a time machine for the
costs in Devexperts technology projects. It is all a market. Either via API or from the web it can replay
single package which the firm completely owns and specific moments in time for symbols covered by
offers complete support for. the market data subscription at the users’ preferred
speed. Whether yesterday or one year ago the time to
provide a particular replay takes approximately five
Market data
seconds or less, depending on the nature request.
To manage market data Devexperts offers a low- While the replay service streams directly from the
latency messaging platform that handles real-time cloud, a second scenario is possible using a Devexperts
data and utilises an internal system to distribute that query engine. The web service query engines lets
data. This system can also function as a standard the exchange extract the data for a specific symbol,
external protocol through the use of a separate API, experience, or event based on user-defined parameters,
- allowing customers to connect and get the data with upon which it extracts the file or streams it.
very little code development. The platform supports
Java, C++ and most web services. Devexperts
Supermodel
can deliver an exchange project with a full set of
protocols right out of the box with compatibility Rather than use a relational data model, which is used
with every major programming language, and every in traditional SQL databases and data warehouses,
major technology. It supports standard distribution Devexperts has developed its own engine for
mechanisms such as Unicast and Multicast; it can historical data. It is a database in the broader sense;
also be converted to use Financial Information a storage engine for unstructured data which is then
eXchange (FIX) protocol if a client prefers to deliver structured to provide quantitative analysis. This big

26 Building blocks of an exchange Mondo Visione

data engine is able to run queries against the data and
process it efficiently in large amounts.
To run high-speed analysis of data, the platform Security Security Client
Token System
uses an on-demand, big-data processing model. It Token
validity info
Data
integrates with the Spark Framework, which lets Pre-staging
the user run queries across clusters of processors,
provided on-demand, creating a parallel-processing
capability without any need for owned infrastructure. Security
gateway
At present, all the historical data that Devexperts service
Final Exchange
holds is compressed at a ratio of 1 to 10, leaving the indexed
Cloud storage,
replay and Staging connectivity
storage
firm managing about 200 terabytes of compressed analytics platform

data which represents five-and-a-half years of data

Extract-
from the American market. Uncompressed it would Transform-Load
be nearly two petabytes.
With this technology users are able to address and Pre-staging

query data, while providing constant query times

Hot stand-by
across all time periods. As a consequence, when the availability zone
user is extracting the data, or processing data with an
on-demand cluster or just replaying data, Devexperts
provides very fast response times. Those response The structure of storage is defined along with the
times are not impacted by the depth of history that data, so the system will apply any changes to the
a query accounts for, or by the type or number of data model across the entire platform, redefining
symbols that are being queried. or extending the data scheme and extending those
Since the platform covers all aspects of real-time changes into data storage. Although feed handlers for
and historical data, it fully underpins the provision specific exchanges are built in proprietary code the
of data services to exchange members as well as messaging engine is open source – and is currently in
supporting the internal processes of an exchange. its fifth generation. Access to the storage engine varies
This is done by providing access to data as and when according to the licence type.
needed, regardless of the formats that might be used. In order to tailor this model to suit clients across
geographies and instruments, Devexperts offers a
variety of approaches to the development model.
Balance of strength and flexibility
Teams of coders are accessible to clients in order to
Flexibility is built into the system, and wedded to the develop systems specifically for individual customer’s
reliability of the platform itself. As such, the exchange needs. Alternatively – and perhaps more suitably for
gains additional resilience in tough circumstances new exchanges, a lighter approach allows a shared
without needing to adopt the kind of rigidity that may team to be created between the client and Devexperts
limit the platform as regulatory demands change in in order to leverage our many ready-made components
the future. with the client’s demands. This process can reduce IT
expenses and enable faster times-to-market – while
Data input Exchange Exchange Exchange still delivering very specific features. The use of shared
Data Feed Data Feed Data Feed
teams is highly recommended in environments where
Data dxFeed Feed dxFeed Feed dxFeed Feed Data storer and ready-made components can enable the very fast
processing Handler Handler Handler converter service
deployment of custom-built solutions.
Platform dxFeed Data Distribution Platform
onDemand cloud- The market operator is hence able to decide whether
based historical
middleware Multiplexors, Delayers, Calculated Data
data storage and it needs a dedicated technical capability or a lighter
Engines, Data Aggregators, Charting Servers
access service support model with only occasional development of
product - or maybe a monthly development cycle.
APIs Java API C API C== API
Above all, Devexperts offers flexibility – we
Consumer End user Application Redistributor
encourage different approaches to different
requirements. MV

Mondo Visione Building blocks of an exchange 27

Building blocks of an exchange Mondo Visione XXV