MODELLING AND SIMULATION OF AN ALARM BASED NETWORK MANAGEMENT SYSTEM FOR EFFECTIVE SLA MONITORING AND MANAGEMENT Karthik Vilapakkam Nagarajan and Gene Awyzio Intelligent Systems Research Group School of IT and Computer Science University of Wollongong Wollongong, NSW-2500, Australia E-mail:
[email protected],
[email protected] Web: www.uow.edu.au/~kv07/Vnk.html
Peter Vial School of Electrical, Computer and Telecommunications Engineering University of Wollongong Wollongong, NSW-2500, Australia E-mail:
[email protected] Web: www.elec.uow.edu.au/people/staff/p.vial
Abstract Modelling and Simulation is increasingly becoming popular in the design process of dynamic network systems. In network management, simulation analysis helps to validate the network models as well as enhance network service operations. In this paper we present simulation as an important process in understanding and analysing SLA management and monitoring. In general, simulation studies are not conducted by various service providers to the fullest possible extent and this leads to poor customer satisfaction. We describe the phase where simulation needs to be performed and what aspects of SLA management has to be simulated. A simulation model for SLM is constructed. We identify alarms as an important tool in effective network resource management. We use a simple case study that describes the need for simulation in effective SLA documentation and SLA monitoring. We analyse different scenarios within SLA levels using ARENA simulation software and demonstrate how it helps in identifying end user services and satisfying customer expectations. The model helps to better understand critical areas in SLA such as alarm monitoring, metrics to be monitored and reduce resource overuse. Through this approach we can perform successful simulation studies, identify the right simulation software for simulating every aspect of SLA management. The model is validated using sensitivity analysis, queueing principles and visualization using animation and hence our experimental results are both valid and reliable. A brief discussion of visual basic modules available in ARENA and its role in conducting network simulation studies has also been presented.
Key Words: SLA Monitoring, ARENA Simulation, Network Management, Thresholds, Simulation study.
1.Introduction Service Level Agreements (SLA) are becoming very important in today’s world as customers are dependent on ISP’s for their businesses and in-house use. SLA is becoming important to satisfy customers that they are being provided good service according to the defined contracts. A properly documented SLA will not only help the ISP to promote their business but will help the ISP to thrive in competitive market place. Not much emphasis has been laid on conducting simulation studies on identifying customer requirements and services to be offered to meet those requirements. Simulation studies will not only help to identify the above mentioned facts but will also help ISP beforehand to check if network meets the desired services to be offered, what metrics need to be monitored during SLA monitoring. This helps in effective SLA creation and also effective SLA monitoring. Many
new ISP who wanted to draft their first SLA find it very difficult as they are not very much sure of service level management goals and performance metrics to be included [13]-[15] One of the main importance of using simulations for SLA monitoring is that we can identify different types of customer with different expectations and priorities by doing different “what-if” analyses. By conducting simulation studies ISP’s can verify their SLA agreements and check if it meets customer expectations and whether the specified service could be provided. The major challenge in monitoring SLA is to establish an effective SLA and to understand the performance indicators (network delay, response time, network availability, service availability) associated with the network [20][24]. This will help to set realistic service expectations that will be guaranteed to users. The rest of paper is organized as follows. Section 2 details problem studied, Section 3 details on using ARENA for simulating alarm based network management system, Section 4 describes importance of simulation, Section 5 describes importance of simulation studies, Section 6 describes simulation models, Section 7 details on verification of simulation results and finally we draw conclusion in Section 8.
2.Problem studied in this paper In this paper we identify simulation as an important process in documenting SLA agreements. Until now simulations are not widely used by various service providers to the fullest possible limit and this leads to SLA that fails to keep up with customer expectations. We describe the exact phase where simulation needs to be performed and what aspects of SLA management have to be simulated. We support our approach through a simple case study that clearly describes the need for simulation in effective SLA documentation and SLA monitoring. We analyse difference scenarios within SLA levels using ARENA simulation software and demonstrate how it helps in identifying end user services and satisfying customer expectations [22].
3. Simulation of Alarm Base Network Management System Using ARENA Simulation Software Arena simulation software is mainly used for creating animated models and representing any system virtually. It is built with the help of SIMAN simulation language. Arena supports all types of applications and has nearly 60 modules. It is very useful tool for simulation of call centers, telecommunication queueing systems and manufacturing, materials processes. Arena has five templates which allow for the support of a specific application. They are Block, Elements, Common, Support and Transfer. Based on the
layout of simulation, modules can be picked up and can be connected. All the information related to that particular module needs to be entered in to its dialog box. Animation is included with many modules and so we can do both the simulation and animation modeling processes. Model jump wizard in arena help us to quickly build the models and helps to specify the key parameters for models like entity, number of stations etc. SIMAN blocks are available to all users in the SIMAN template. Arena is built using this SIMAN simulation language and integration via DAO(Data Access Objects), ODBC (Open Database Connectivity) can also be done. ARENA has an input analyzer for determining appropriate input to a model. Output analyzer facility help us to view the simulation results after the simulation has run to completion. It also helps provide graphical display through histogram, plot and meters. Real-time controlling and monitoring is done through ARENA RT for the purpose of making the simulation model to interact with external clients[4][1][2][3][19]. Alarms perform a routine check to make sure all the operations taking place within network are going on smoothly, however if a check fails a process gathers information about type of check that has failed, time of the day and the level of importance. Three main levels exist within alarm monitoring feature based on importance and nature of failure - Critical, Major and Minor. Alarm surveillance is used to monitor the condition of the network usage thresholds set through SLA devices. One of the main areas of focus is the usage levels that helps us to identify a particular resource usage and determine its availability. Alarms can be generated for different SLA violations (using more bandwidth, heavy traffic, congestion in networks) [8] [11].
4. Importance of Simulation Modelling and Analysis in Network Management Representing adequately and correctly the system under study was given the primary importance while modelling. “How to effectively configure alarms” in network management requires good simulation study with proper simulation software. This will help us to experiment with different thresholds for different customer levels which is an important factor in SLA documentation. Inaccurate simulation study will not only cause service degradations to customers (service not being provided according to SLA), but will also cause poor SLA monitoring as correct threshold patterns are not identified and right performance metrics are not monitored. The main aim is to study the appropriateness and effectiveness of using simulation software to solve problems related to SLA management and monitoring [23].
5.Neglecting simulations studies leads to service failures One of the main reasons why Service providers fail to provide defined service according to SLA specifications is due to negligence of conducting simulation studies. Simulations are not done either due to lack of time, rush to be first in the market, pressure from customers and finally competition in the market place. This leads to service measures that are not identified properly. Some important factors that needs to be included when conducting SLA simulation studies are ?? Does the simulation study results help to find out whether the desired network can support wide range of customers who have different thresholds in terms of availability, response time and service guarantee. ?? Can the network handle e-business applications that might increase the network load due to transaction between the web server and database server. ?? What type of applications require monitoring check to check if the service provider is providing service as per SLA
??
Is the network set-up able to cope up with additional load when there is increase in number of users We suggest that simulations should be used in the first place to do different “what-if” analysis based on which service levels have to be established (should also consult vendors (or) suppliers) and then SLA agreement should be created for customers. “What-if’ analysis helps us to check the service performance limits and identify misbehaviour in network configurations. Thus corrective measures can be taken before services failures occurs. Other reasons include lack of resources for monitoring and reporting services [8]-[10].
Modelling and Simulation Simulation Study [20]-[22]
objectives:
SLA
The following objectives have been identified 1. Service provider will be committed to service being provided to customers 2. Ensuring user satisfaction by checking whether the network can meet end user requirements 3. Identifying the metrics to be monitored, which particular area where faults are likely to occur and generating alarms by setting thresholds and prioritising the alarms based on their importance. 4. Record the different scenarios examined using simulations and analysing the output and validating the model.
The type of scenarios examined in this SLA simulation study The following types of simulation study have been identified 1. What type of service each group of customer needs to be simulated 2. How important is the service to be used is to the customer (business (or) personal use) 3. Additional service to be provided to each group of customer based on the service level priority 4. SLA needs to cover all essential service metrics that will be monitored and detailed explanation of various policies used The successful running of ISP’s depends on variety of factors but to be more specific proper planning, good simulation studies, right monitoring tool and good pricing scheme [8]-[10].
6.Explanation of various simulation modules developed 6.1.Service Level Management Process Module In service level management process we investigated the service level policy implication on network users. Users arrive in Arrive module and then are sent to Server module for the intended service they require. Using Branch module we control the user entity flow through set of branches. User entity arriving in branch module is checked for specific branch condition and then sent to next module specified. We use probabilistic branching as this is suitable for us to understand policy control on users. Specifically, 0.65 probability criteria that most of users choose the policy service type A in the model as the SLA here is reasonably good (assumption of good network service, resource allocation). Rest of the users choose policy 1 as it has two main processing levels. We use Duplicate module implementing this process and users can select payment on monthly as well as weekly basis and then entities go to the service agreement level in server module. If the first two types of policies are not suitable then users have no alternative than to choose policy 2 that has various resource limitations such as Number in Queue less than 50, (NQ(select policy 2_R_Q)
