QoS Parameters to Network Performance Metrics Mapping for

SLA Monitoring
Hyo-Jin Lee, Myung-Sup Kim and James W. Hong*, Gil-Haeng Lee**
*Distributed Processing & Network Management Lab.
Dept. of Computer Science and Engineering, POSTECH, Pohang Korea
**Network Technology Research Lab., ETRI, Daejeon Korea
({really97, mount, jwkhong}@postech.ac.kr , [email protected])

Abstract
Service Level Agreement (SLA) is a formal negotiated agreement between a service provider
and a customer. The service level management (SLM) is the integrated management of all
functionalities in the SLA life cycle. When a customer orders a service from a service provider, an
SLA is negotiated and then a contract is made. In the SLA contract, QoS parameters that specify the
quality level of service that the service provider will guarantee are included. The service provider
must perform SLA monitoring to verify whether the offered service is meeting the QoS parameters
specified in the SLA. SLA monitoring involves monitoring the performance status of the offered
service and provide relevant information to the service level management system. In order for the
service level management system to verify whether the specified QoS parameters are being met, the
system must gather performance data from the underlying network performance monitoring system
and map such data to the QoS parameters. In this paper, we propose a formal mapping mechanism
between QoS parameters in SLA and the network performance metrics. Although we focus on the
network access service (e.g., leased-line service, xDSL service, VPN service) in this paper, we
believe that our mapping mechanism can be easily used in SLA monitoring of other services (such
as application, server hosting, contents). We also propose a general SLA monitoring system
architecture that can be used to monitor service levels for various services offered by network,
Internet and application service providers. Finally we present how our SLA monitoring system
architecture can be used for SLA monitoring of IP backbone network service.

Keywords: SLA, Service Level Management, SLA monitoring, QoS Parameters, Network
Performance Metrics, Measurement Mapping, Evaluation Mapping.

1. Introduction liberalization and rapid evolution of the

SLA is a formal negotiated agreement between telecommunication market is one of the major
a service provider (SP) and a customer. Usually in reasons for the increased significance of SLA. By
measurable terms, SLA is defined as the quality of the supported service of SLA, the service provider
services the service provider will provide [1]. The can differentiate itself from its competitors and
prioritize service improvement opportunities. From based services. TM Forum mentioned the
the viewpoint of the consumer, he may desire to importance of SLA monitoring, but did not present
access a service of his own inclination and to the details of SLA monitoring method or its
validate the quality of the service provided. This is architecture [2].
another reason for the increased importance of In this paper, we categorize SLM into seven
SLA. The SLA life cycle consists of the following functions. We define the concept of SLA
five phases: product/service development, monitoring in the network management layer. The
negotiation and sales, implementation, execution, most important factor in SLA monitoring is the
and assessment [2]. The service level management mapping between QoS parameters and network
(SLM) is the integrated management of these five performance metrics (NPMs). We also define the
phases in the SLA life cycle. Given the importance concept of this mapping and the requirements of an
of SLM, the concept and methods are described in SLA monitoring system. From this concept we
many other papers [13, 18, 19, 20]. present a generic SLA monitoring system
SLA provisioning and SLA monitoring are architecture. To validate our theory we applied our
critical to realize the SLA supported service in the design to an IP-backbone network service.
network management layer. SLA provisioning The organization of this paper is as follows.
configures the network and system infrastructure Related work is described in Section 2 and QoS
for quality insured service. In the network service parameters to Network performance metrics
area, many new traffic engineering methods such mapping is described in Section 3. Next, design
as MPLS [3] and Diffserv [4] have been proposed, issues of SLA monitoring system are provided in
and much valuable research has been performed [3, Section 4. We propose SLA monitoring system
21] for QoS-based traffic treatment. The role of architecture in Section 5 and apply it to an IP
SLA monitoring is to monitor the service status for backbone network service in Section 6. Finally,
each customer according to the agreed QoS concluding remarks are given and possible future
parameters in SLA and to provide a basis for the work is mentioned in Section 7.
billing and reporting system. Although much
research on network monitoring has been
performed and the result of network monitoring is 2. Related Work
very essential to SLA monitoring, it lacks the
2.1 SLM and SLA Monitoring
considerations of how to apply the result of
In this section, we define the seven functions of
network monitoring to SLA monitoring. In addition,
SLM and make clear the concept of SLA monitoring,
many service providers use their own proprietary
which is one of the seven functions of SLM. Various
SLA monitoring methods. Thus, further
service providers and customers need SLAs in the
consideration should be given to a standardized
current telecommunication market place [2]. Service
SLA monitoring method. Also, the general
providers are classified into three types: Network Service
guideline for SLA monitoring is necessary for
Provider (NSP), Internet Service Provider (ISP), and
service providers who newly start to launch SLA-
Application Service Provider (ASP). Customers are degraded. Next, it notifies the SLA provisioning to
divided into three types: individuals, organizations and restore the service. SLA reporting provides the
enterprises. A service provided by a service provider is performance information to customers periodically or
associated with other services. In this case, the service on-demand. Finally, SLA assessment demands payments
provider may be a customer of another service provider. to customers and accommodates customers with a
So, they are in contract with each other. Therefore, the penalty when the violation occures.
efficient and systematic SLM is needed to support SLAs
which are diverse and complex.
QoS Parameters
SLA Contract Problem Notification

Business Level Management Performance Data

Input
Customer Interface Management

Output
Service Level Management

SLA Monitoring
SLA creation SLA assessment
Customer

SLA negotiation SLA reporting Network Monitoring

System
SLA provisioning SLA maintenance
Figure 2. SLA Monitoring
SLA monitoring

Among the SLM functions described above, SLA

Network Level Management
provisioning and SLA monitoring are the most important
Figure 1. Service Level Management Functions
in the network management layer. This paper focuses on

SLM is the integrated method to manage various SLA monitoring, which is important to assure the degree

SLAs from creation to assessment. We categorize the of QoS parameters and to use the monitoring results in

SLM into seven functions: SLA creation, negotiation, reporting and assessment. In other words, this paper

provisioning, monitoring, maintenance, reporting and presents a new concept of SLA monitoring, including

assessment, as illustrated in Figure 1. SLA creation three processes of Telecom Operations Map (TOM) [6],

creates an SLA templete for specified services. SLA which are network data management process at the

negotiation is the process of selecting applicable QoS network management layer, service problem

parameters in SLA and negotiating the penalty in case of management, and service quality management at the

SLA violation. SLA provisioning means that service service management layer. There are two types of input

providers configure the network element or topology to in SLA monitoring, as illustrated in Figure 2. One is

provide the service. After provisioning, service providers various QoS parameters according to services and the

must verify the degree of SLA assurance, which they other is SLA contract for each customer. After SLA

contracted with customers. To perform surveillance on monitoring uses network monitoring system, the output

QoS parameter degradation or violation, SLA monitoring of SLA monitoring is also dual. One is problem

is needed. When a violation of a QoS parameter is notification and the other is performance data.

detected, SLA maintenance analyzes the reason why

degradation has occurred and which QoS parameter has 2.2 QoS Parameters and Network Performance
Metrics in Network Service. used capacity or available capacity. Service providers
In this section, we explain QoS parameters and usually assure the maximum bandwidth to customers and
NPMs from the viewpoint of network service. There are it is stated clearly in SLA. The four QoS parameters
many kinds of network services such as leased line mentioned above are technology-specific, so they can be
service, IP-VPN service, xDSL services, Frame relay measured straightforward by NPMs. MTBF and MTRS
service, etc. Also, QoS parameters are the target of SLA are time-based, so they cannot be directly mapped by
monitoring and NPMs are needed to measure the NPMs.
network performance and guarantee QoS parameters.

Connectivity
Availability
Availability Functionality

Delivery One way loss

Loss
NPMs RT loss
Latency
QoS Parameters One way delay
Delay
Bandwidth RTT delay
Delay variance
MTBF
Capacity
Utilization
MTRS Bandwidth
Figure 3. QoS Parameters for Network Service Throughput

Figure 4. Network Performance Metrics

QoS parameter is the instance to represent the
quality of service to customers. It should be easy for Network performance metric (NPM) means the basic
customers to understand the degree of assuring the metric of performance measurement in the network
service. QoS parameters can be different according to the management layer. In Figure 4, we categorize the NPMs
type of services. Figure 3 illustrates generic QoS into four types: Availability, Loss, Delay and Utilization.
parameters required in network service: Availability, The meaning of each NPM is as follows. Availability
Delivery, Latency, Bandwidth, MTBF (Mean Time means connectivity and functionality in the network
Between Failure) and MTRS (Mean Time to Restore management layer. Connectivity is the physical
Service). The definitions of these parameters are as connectivity of network elements and functionality
follows. Availability is the percentage of the feasibility of means whether the associated network devices work well
service in every particular service request. TMF701 [1] or not. Loss is the fraction of packets lost in transit from
defines that Availability of service is the key parameter sender to target during a specific time interval, expressed
that customers are interested in. Delivery is the converse in percentages. Loss consists of two metrics: one-way
of packet loss, which means that a percentage of each loss and round-trip loss. Delay is the time taken for a
service is delivered without packet loss. To some service packet to make the average round trip or one-way from
providers, Delivery means packet delay. It also depends the sender to the distant target and back. Delay consists
on the decision of service providers. Latency is the time of three kinds of factors: one-way delay, round trip delay
taken for a packet to travel from a service access point and delay variance. Utilization is the throughput for the
(SAP) to a distant target and back. It usually includes the link expressed as a percentage of the access rate.
transport time and queuing delay. Bandwidth means the
2.3 Network Monitoring The passive monitoring method obtains the current
Network monitoring is the process of measuring the status of the network by capturing the packet. Passive
value of NPMs. NPMs are measured by various network monitoring can monitor the network status without
monitoring technologies. Traditionally, there are three additional traffic. However, limited NPMs, such as
methods of network monitoring, which are active utilization and throughput, can be measured easily
monitoring [5, 10], passive monitoring [10, 11] and the compared to the active monitoring method. Some related
method of using SNMP agents [12, 33, 34]. Table 1 work in this are CoralReef [30], WAND [31],
shows the mechanism and related projects of three WebTrafMon [11] and NLANR PMA [32].
network monitoring methods. By using SNMP agents, we can measure the status of
The active monitoring method obtains the current the network device. For example, RMON [12] monitors
status of the network by setting up the test machine at the traffic information with SNMP agents. The method of
point which one wishes to measure, and then sending using SNMP agents is simple and scalable. But only the
extra traffic from one machine to another during a throughput and functionality of NPMs are measured by
specific time. Various NPMs can be measured by simple this method. Some related work on this are internet2 [33]
and easy tools, such as ping and traceroute. And system and MAWI [34].
load is very low because the amount of generated and Generally, Loss, Delay and Connectivity are the
analyzed traffic is small compared to passive monitoring NPMs that can be mainly measured by active monitoring.
method. But test packets can be lost at low priority, so it Utilization and Throughput are the NPMs which can be
is difficult to measure the exact network status. And test measured by passive monitoring. Functionality and
traffic may impose a burden on the current network. Throughput can be measured using SNMP agents.
Some related work on this are RIPE NCC Test Traffic Although we may obtain various NPMs using these
Measurement [5, 24], NIMI [25], Surveyor [26], network monitoring methods, it is difficult to apply these
NLANR AMP [27], PingER [28] and Skitter [29]. values to QoS parameters directly. The contract between
service providers and customers is performed using QoS
Table 1. Network Monitoring Mechanism and parameters and the network quality is measured using
Related Projects
NPMs. In other words, the NPMs for each QoS
Monitoring Mechanism Related projects
Method
parameter must be first decided before measuring the
Active ♦ Generate test traffic ♦ NIMI QoS parameters. A QoS parameter can be mapped to one
Monitoring periodically or on-demand ♦ Surveyor
♦ Measure performance of ♦ NLANR AMP NPM or many. This mapping depends on the type of
test packet or response ♦ PingER
service and can be very complicated. And the quality
♦ Skitter
Passive ♦ Capture the traffic by ♦ CoralReef
information of service should be presented in customer
Monitoring mirroring or splitting ♦ WAND friendly form, QoS parameters, not NPMs. Therefore, it
♦ Analyze the captured ♦ WebTrafMon
packets ♦ NLANR PMA is necessary to translate the measured NPMs to QoS
SNMP ♦ Using existing SNMP ♦ Internet2
parameters in SLA monitoring. So we define a new
agents ♦ MAWI
concept of a mapping between QoS parameters to NPMs
in Section 3.
 measured by Utilization of NPM. Utilization means the
3. QoS Parameters to NPMs Mapping throughout of link as a percentage. Contrary to these
technology specific parameters, MTBF and MTRS are
3.1 Mapping between QoS parameters and NPMs in
the QoS parameters which cannot be obtained from NPM
Network Service
directly. So service providers may satisfy these QoS
In this section, we describe a mapping between QoS
parameters by computing the time when the violations of
parameters and NPMs. We divide this mapping into two
NPMs occurred.
steps: the measurement mapping and the evaluation
Second, we explain the evaluation mapping.
mapping. Before we generalize our concept of QoS
Evaluation mapping is mapped to evaluation function to
parameters to NPMs mapping, we mention a specific
verify QoS parameter from measured NPM values. In the
example of mappings in network service, as described in
bottom side of Figure 5, we show an example of
Figure 5.
evaluation mapping. Availability of QoS parameter is
Measurement Mapping considered with connectivity and functionality in the
QoS Parameters NPMs network service. So we should check the connectivity
Connectivity
Availability Availability Functionality and functionality of the network service. When a
Delivery One way loss
Loss RT loss malfunction of a network device or disconnection occurs,
Latency
One way delay
Bandwidth Delay RTT Delay we measure the unavailable time. Then by applying the
Delay variance

MTBF Capacity unavailable time to the above formula, we can verify the
MTRS Utilization Bandwidth
Throughput Availability of QoS parameter.
Evaluation Mapping Measurement mapping and evaluation mapping are
Availability of device unfunctional time + disconnected time
QoS Parameter = 100 − ×100 (%) essential for SLA monitoring. These mappings are highly
total monitored time

dependent on the kind of services. In this paper, we make

Figure 5. Mapping Example in Network Service
a guideline only in network service. When the available

First, we explain measurement mapping in network NPMs for a QoS parameter are decided, the value of

service. Measurement mapping is the process of NPMs is measured by using some network monitoring

determining some NPMs to a QoS parameter. methods. Afterwards, we can apply the measured value

Availability of QoS parameter is mapped to Availability of NPMs to the evaluation function from evaluation

of NPM. Availability of QoS parameter means the mapping. After obtaining the value, we can assess a QoS

feasibility of service in every particular service request. parameter in SLA.

Availability of NPM refers to connectivity among

network elements and functionality of the network 3.2 Generic Mapping between QoS parameters and

elements. Functionality means whether network elements NPMs

work well or not. Delivery of QoS parameter is mapped We present the generic mapping between QoS

to Loss or Delay of NPM, but this depends on the parameters and NPMs, as illustrated in Figure 6. We

decision of service providers. Latency of QoS parameter formalize the mapping using the theory of set and

means Delay of NPM: one-way delay, round trip delay functions. Through this formalization, we make clear the

and delay variance. Bandwidth of QoS parameter is concept of QoS parameters to NPMs mapping, which is a
mandatory process in the construction of SLA verify a QoS parameter from measured NPM values.
monitoring system. Evaluation function, which is element of set E, can be
N
First, we define the three sets : Q, N and E. Set Q different for each element of Set 2 . However, it is
represents a set of QoS parameters, set N represents a set difficult to define the evaluation function according to
of NPMs and set E is the set of evaluation functions. QoS parameters, and it also depends on the type of
N
Set 2 is the power set of N. The relationship between service.
N
set Q and set 2 is the measurement mapping: m(x), In this paper, m(x) and n(x) are defined as the QoS
N
and the relationship between set 2 and set E is the parameters to NPMs Mapping. Therefore, to evaluate
evaluation mapping: n(x). QoS parameters in SLA monitoring, it should be decided

m(x) n(x) which NPMs should be mapped and which evaluation

Q 2N E
..
e1 functions should be applied to the selected NPMs.
q1 {n1,n2}
e2
q2 {n2, n3}
.. e3
q3 {n4}
.. e4
q4 4. Design Issues of SLA Monitoring System
{n5,n6} e5
QoS Parameter .. ..
NPM Evaluation
Function
N In this section, we describe the design requirements
NPM
(N1,N2,N3,N4)
of an SLA monitoring system for network services of
H I J K NSP and ISP. The SLA monitoring system receives
Monitoring Method Agent Type
Monitoring Point Monitoring Period
(A,P,S) (aa, na, ta, sa)
service contents, such as QoS parameters and customer
Figure 6. Generic Mapping Method SLAs as its inputs from the business process system, and
uses network and system monitoring technologies to
m(x) represents a decision which NPMs are used to gather corresponding network performance values. Next,
N
measure each QoS parameter. In m(x), we use set 2 , it evaluates the quality of a given service and gives the
not the set N, because the mapping between set Q and set results to the upper layer systems, such as billing and
N may not be one-to-one mapping. Moreover, reporting system. The consideration of the SLA
monitoring method (H), point (I), period (J) and agent monitoring system is based on the previously mentioned
type (K) should be considered in m(x) for the effective QoS parameters to NPMs mapping, and follows the basic
measurement mapping, as shown in the lower side of structure described in the SLA Handbook [2] from TM
Figure 6. The monitoring method includes active Forum. The following gives the general requirements and
monitoring, passive monitoring and the method using some design issues for the SLA monitoring system.
SNMP agents. Monitoring point is usually the service First, the SLA monitoring system should be designed
access point where customers get provided service. And with no dependencies on the SLA provisioning
monitoring period is the total time for NPM monitoring. technologies. In the network management layer there can
There are four agent types: application agent (aa), system be various traffic engineering methods [3, 21] to
agent (sa), network agent (na) and traffic agent (ta) [13]. configure the QoS supported network service, but the
Among them, the agents used in network service are quality of the SLA based service should be measured
network agents (na) or traffic agents (ta). based only on the QoS parameters and corresponding
n(x) represents a decision of evaluation function to NPMs. Second, the SLA monitoring system should be
designed in a layered and distributed architecture. The Mapper module, the Customer SLA module and the
customers of many network services such as IP-VPN, Monitoring module. The role of the Parameter Mapper
xDSL services are geographically widespread, so there module that is illustrated in Figure 7 is to manage the
can be many dispersed service access points in which an QoS parameters to NPMs mapping. For each QoS
SLA agent should gather network performance values. parameter a decision about NPMs, monitoring location,
So the distributed and layered architecture is suitable. period, method and agent type is made. Based on this
Third, the monitoring packet and communication data decision, the SLA agents are deployed at the proper site
should not overburden the underlying network. The in the service network. Another important consideration
functional modules and the communication data should in the Parameter Mapper is that the evaluation function
be well defined, as well as the monitoring method. for each QoS parameter is created and stored in the Map
Fourth, flexibility and extensibility should also be DB. This evaluation function is used in the Performance
considered. The SLA monitoring system evaluates the Analyzer to validate the quality of a given service for
quality of a given service for each customer according to each customer. Besides the evaluation function, the other
its contracted SLA. Over a period of time, the number of mapping information is stored in the Map DB.
customers and the content of SLA can vary as well. To
Related SLM Components
cope with this situation, the consideration of Customer
Business
SLA
Maintenance Reporting Billing
Service
Planning
System System System System
and Development
extendibility and flexibility is crucial.
SLA Monitoring Reporting
Next, we are going to describe the design issues of System Manager
Problem
Manager Parameter
an SLA monitoring system. The essential step is to Data Mapper
Store
define the QoS parameters to NPMs mapping. Here, one
Customer Performance Map
should also determine the monitoring location, method, SLA Analyzer DB

period and agent type. There are three types of Parameter

Customer SLA PM PM PM Mapper module
module Collector Collector
monitoring methods (active, passive and SNMP-based) Collector

SLA SLA SLA

and four types of agents (application agent (aa), system Agent Agent Agent

Monitoring module
agent (sa), traffic agent (ta) and network agent (na)) [13].
Figure 7. SLA Monitoring System Architecture
The evaluation function should be determined for each
QoS parameter from network performance values
Customer SLA module in the left side of Figure 7
gathered by SLA agents. One should also determine the
stores the information of each customer contract. This
stored data format and the period to be kept in storage.
module should be designed with good flexibility because
The communication data format among each functional
the amount of stored information can vary easily
module is designed to avoid high network traffic, and is
according to the popularity of the service. The LDAP
another critical factor to be considered.
[22] is one possible implementaion solution for the
Customer SLA module.
5. SLA Monitoring System Architecture
The center of Figure 7 shows the Monitoring
In this section, we represent an SLA monitoring
module which is designed in a layered architecture.
system architecture as illustrated in Figure 7. This
There can be multiple SLA agents for one service,
architecture has three main components: the Parameter
because there can be multiple QoS parameters for one
service and multiple NPMs can be mapped to one QoS Therefore, we decided on the QoS parameters for the
parameter. Also, in some cases, multiple SLA agents IP-backbone network service with the following three:
should be deployed at a regionally dispersed place. The Availability, Latency and Loss. The content of SLA
Performance Metric (PM) Collector gathers the negotiation may be as follows: The backbone availability
performance data measured at each SLA agent and sends over 99.99% should be assured. The average round trip
these to the performance analyzer. The quality evaluation time (RTT) should be less than 50 ms. The delivery ratio
for each QoS parameter and each customer SLA occurs should be more than 98.0%. From our definition of QoS
in the performance analyzer. The performance analyzer parameters to NPMs mapping, there should be a decision
evaluates the quality using the evaluation function stored about m(x) and n(x).
in the Map DB and customer contract information in the
Customer SLA DB. The analyzed information is stored Table 2. Measurement Mapping for IP-Backbone
Network Service
at the Data Store module for a certain time period. The
QoS NPM method point period type
stored data format and stored period is an important point Parameter
Availability Functionality active edge 5 min na
to consider. If a problem such as SLA violation is router
Connectivity active edge 5 min ta
detected, the Performance Analyzer sends an alarm router
Latency RTT active edge 5 min ta
message to the Problem Manager, which detects this router
failure or problem and sends an alarm message to an Delivery RT loss active edge 5 min ta
router
administrator or SLA maintenance system immediately.
The Reporting Manager is to create a report for the
First, we describe about m(x) which is illustrated in
billing system and the reporting system from the Data
Table 2. The QoS parameter Availability is mapped to
Store.
two NPMs: functionality and connectivity, which are
monitored at each edge router and every five minutes
6. SLA Monitoring for IP-backbone Network using an active monitoring method. The network agent
Service
(na) is used to acquire availability NPM value and the
traffic agent (ta) is used for connectivity NPM value. In
To validate our proposed architecture, we applied it
the same manner, Latency is mapped to RTT and
to an IP-backbone network service and designed an SLA
Delivery is mapped to round trip (RT) loss. By this m(x)
monitoring system. The service provider of IP-backbone
mapping the monitoring method, point, period and agent
network service belongs to NSP among our previous
type are decided. For the IP-backbone network to
categorization such as KT, NTT and AT&T. The
monitor the given NPMs, we should set up SLA agents at
customer of this service can be any user who is
each edge router which generates test packet every five
connected to this IP-backbone network. These can be
minutes and sends the results to the upper layer PM
companies, organizations, other ISPs using leased line
Collectors.
service, and individuals who are using xDSL service.
The QoS parameters in these service should be decided
with the terms to represent whether the backbone
network is in a healthy state or not.
 Table 3. Evaluation Mapping for IP-Backbone amount of data is smaller than that generated by passive
Network Service traffic monitoring. In addition, the Parameter Mapper,
QoS Evaluation Function Map DB and Customer SLA DB which receive data
Parameter
Availability 100 − device unfunctional time + disconnected time ×100 (%) from extraneous systems are located at the same machine.
total monitoredtime
Latency ∑ RTT (m sec)
total number of RTT test packet 7. Conclusion and Future Work
Delivery number of lost packet
100 − × 100 (%) In this paper, we proposed a new concept of SLA
total number of test packet
monitoring: a QoS parameters to NPMs mapping and a
Next, the n(x) mapping is used to determine the generic architecture for the SLA monitoring system. The
evaluation functions to measure the quality of each QoS QoS parameter to NPM mapping combines measurement
parameter, as described in Table 3. As mentioned mapping and evaluation mapping. From this mapping
previously, these evaluation functions are stored in the one can decide which type of SLA agents are needed to
Map DB and are used by the Performance Analyzer. evaluate the quality of the provided service in an easy
and systematic way. We also provided a mapping

Parameter Mapper.
guideline for network services. In the network service,
Map DB.
Customer Business Customer SLA. the QoS parameters can be Availability, Delivery,
System,
Service Planning Latency, Bandwidth, MTRS and MTBF. We also
and Development Performance Analyzer.
System Data store. provided the points of consideration in the design of the
Problem Manager.
SLA Maintenance Reporting Manager.
System, SLA monitoring system and presented its generic
Reporting System, PM Collector.
Billing System
Backbone router.
architecture. We believe that this paper can serve as a
SLA agent.
- loss check guideline for service providers who intend to deploy an
- RTT check
backbone network
- connectivity check SLA-based network service and set up an SLA
- functionality check
data flow
monitoring system.
Figure 8. Layered and Distributed SLA monitoring We plan to complete our validation with the
System in IP Backbone Network Service
implementation of the SLA monitoring system we
designed for the IP-backbone network service. We intend
Figure 8 shows the SLA monitoring system we
to make an overall guideline of the QoS parameter to the
designed for the IP-backbone network service. The SLA
performance metric for various services such as hosting,
agents which are deployed at every edge router measure
application and content services. A refinement of the
network performance metrics using active method and
proposed SLA monitoring system architecture is also
send the measured NPM values to the PM Collectors.
needed to apply to various services.
There can be two or more PM Collectors because the
backbone edge routers are located regionally far away
from each other. It is necessary to deploy multiple PM
References
Collectors to reduce the monitoring traffic.
The Performance Analyzer, Data Store, Problem [1] TM Forum, "Performance Reporting Concepts and
Manager and Reporting Manager can coexist at the Definitions," TMF701 v2.0, Nov., 2001.
same machine because in this kind of monitoring system [2] TM Forum, "Service Level Agreement Management
 Handbook," GB917 v1.5, Jun., 2001. Delay Metric for IPPM," IETF RFC2679, Sep., 1999.
[3] T. Choi, S. Yoon, H. Chung, C. Kim, J. Park, B. Lee, [16] G. Almes, S. Kalidindi, M. Zekauskas, "A One-way
T. Chung, "Wise: Traffic Engineering Server for A Packet Loss Metric for IPPM," IETF RFC2680, Sep.,
Large-scale MPLS-based IP Network," Proc. of 1999.
NOMS 2002, Florence, Italy, Apr., 2002, pp. 251-264. [17] G. Almes, S. Kalidindi, M. Zekauskas, "A Round-
[4] S. Blake, et. al, "An Architecture for Differentiated trip Delay Metric for IPPM," IETF RFC2681, Sep.,
Services," IETF RFC2475, Dec., 1998. 1999.
[5] M. Alves, et. al, "New Measurement with the RIPE [18] K. Appleby, et. al. , "Oceano - SLA based
NCC Test Traffic Measurement Setup," Proc. of PAM management of a computing utility," Proc. of the 7th
2002 Workshop, Colorado, USA, 2002. IFIP/IEEE International Symposium on Integrated
[6] TM Forum, "Telecom Operation Map," GB910, v2.1, Network Management, Seattle, WA, USA, May 2001,
Mar., 2000. pp. 855 -868.
[7] CAIDA, "Network Measurement FAQ," Jan 17, 2002, [19] CISCO white paper, "Successful Implementation
http://www.caida.org/outreach/metricswg/faq.xml. Strategies for Service-level Managment," CISCO,
[8] IPPM, http://www.advanced.org/IPPM/. 2000.
[9] Thomas Lindh, “An Architecture for Embedded [20] P. Bhoj, et. al, "SLA management in federated
Monitoring of QoS Parameters in IP-Based Virtual environments," Proc. of IM’99, Boston, MA, USA,
Private Networks,” Proc. of PAM2001 Workshop, May 1999, pp. 293-308.
Amsterdam, Apr., 2001. [21] A. Feldman, et. al, "NetScope: Traffic Engineering
[10] T. Lindh, "A New Approach to Performance for IP Networks," IEEE Network, Vol. 14, No. 2,
Monitoring in IP Networks-Combining Active and Mar./Apr. 2000, pp. 11-19.
Passive Methods," Proc. of PAM 2002 Workshop, [22] W. Yeong, T. Howes, S. Kille, "Lightweight
Colorado, USA, 2002. Directory Access Protocol", IETF RFC1777, Mar.
[11] S. H. Hong, J. Y. Kim, B. R. Cho, J. W. Hong, 1995.
"Distributed Network Traffic Monitoring and Analysis [23] M. Murray, K. Claffy, “Measuring the
using Load Balancing Technology," Proc. of Immeasurable: Global Internet Measurement
APNOMS 2001, Sydney, Australia, Sep., 2001, pp. Infrastructure”, Proc. of PAM2001, Amsterdam, Apr.,
172-183. 2001.
[12] S. Waldbusser, "Remote Network Monitoring [24] RIPE NCC Test Traffic Measurements,
Management Information Base," IETF RFC1757, http://www.ripe.net/test-traffic/index.html.
Feb.,1995. [25] NIMI, http://www.ncne.nlanr.net/nimi/.
[13] L. Lewis., P Ray.,"On the migration from enterprise [26] Surveyor, http://www.advanced.org/csg-ippm/.
management to integrated service level management," [27] NLANR AMP, http://moat.nlanr.net/AMP.
IEEE Networks , Vol. 6, No.1, Jan., 2002, pp 8-14. [28] PingER, http://www-iepm.slac.stanford.edu/pinger/.
[14] J.Mahdavi, V. Paxson, "IPPM Metrics for [29] Skitter,
Measuring Connectivity," IETF RFC2678, Sep.,1999. http://www.caida.org/tools/measurement/skitter/.
[15] G. Almes, S. Kalidindi, M. Zekauskas, "A One-way [30] CoralReef,
 http://anala.caida.org/CoralReef/Demos/cerfnet/link. 1992 ~ 1995 Univ. of Western Ontario, 연구교수
[31] WAND, http://wand.cs.waikato.ac.nz/ 1995 ~ 현재 포항공과대학교 컴퓨터공학과 부교수
[32] NLANR PMA, http:// moat.nlanr.net/PMA. <관심분야> 분산처리, 네트워크 트래픽 모니터링,
[33] Internet2, http:// monon.uits.iupui.edu. 네트워크 및 분산 시스템 관리, CORBA, Internet
[34] MAWI, http:// tracer.csl.sony.co.jp/mawi. 관리

이길행
1986 KAIST 전산학과 석사
1996 KAIST 전산학과 박사
이효진 1986 ~ 현재 한국전자통신연구원
2001 동국대학교, 컴퓨터공학과 네트워크연구소 서비스네트워크
학사 연구부 SLA 관리체계연구팀장/책임연구원
2001 ~ 현재 포항공과대학교, 컴 <관심분야> 분산처리, 로드 밸런싱, 네트워크 관
퓨터공학과 석사 과정 리, DBMS
<관심분야> SLA 모니터링, 네트워크 모니터링,
XML 기반의 네트워크 관리

김명섭
1998 포항공과대학교, 전자계산
학과 학사
2000 포항공과대학교, 컴퓨터공
학과 석사
2000 ~ 현재 포항공과대학교, 컴퓨터공학과 박사
과정
<관심분야> 분산처리, 네트워크 트래픽 모니터링,
네트워크 및 분산 시스템 관리

홍원기
1983 Univ. of Western Ontario, 전
산학 학사
1985 Univ. of Western Ontario, 전
산학 석사
1985 ~ 1986 Univ. of Western
Ontario, 전산학과 강사
1986 ~ 1991 Univ. of Waterloo, 전산학 박사
1991 ~ 1992 Univ. of Waterloo, Post-Doc fellow