Comparative Evaluation of Available Bandwidth Estimation Tools (Prtg and Oaunetmon) in a Campus Wide Area Network

Azeez Nureni Ayofe. et al / International Journal of Computer Science and Information Technology (IJCSIT) 2010, Vol1 (1), 19-30

COMPARATIVE EVALUATION OF AVAILABLE BANDWIDTH ESTIMATION TOOLS (PRTG AND OAUNETMON) IN A CAMPUS WIDE AREA NETWORK.

Azeez Nureni Ayofe

Department of Maths & Computer Science, O. Osunade College of Natural and Applied Sciences, Department of Computer Science, Fountain University, Osogbo, University of Ibadan, Nigeria. Osun State, Nigeria. E-Mail address: [email protected] E-mail address: [email protected] [email protected]

Azeez Raheem Ajetola Department of Maths & Computer Science,

College of Natural and Applied Sciences, Fountain University, Osogbo,

Osun State, Nigeria.

ABSTRACT: In recent years, there has been a strong interest in measuring the available server selection tasks, among many bandwidth of network paths. Several others, without requiring access to methods and techniques have been proposed and various measurement tools network elements or administrative have been developed and evaluated. resources. However, there have been few comparative Application users on high-speed studies with regard to the actual networks perceive the network as an performance of these tools. This project end-to-end connection between presents a study of available bandwidth resources of interest to them. measurement techniques for both PRTG Discovering the least congested end-to- and oaunetmon and undertakes a end path to distributed resources is comparative analysis in terms of accuracy, important for optimizing network intrusiveness and response time of active utilization. Therefore, users and probing tools. Finally, measurement errors application designers need tools and and uncertainty of the tools are analyzed methodologies to monitor network and overall conclusions were made. conditions and to rationalize their performance expectations. Keywords: Network paths, Comparative study, Available bandwidth, Intrusiveness, Several network characteristics related Response Time, Probing Tools to performance are measured in bits per second: capacity, available bandwidth, 1.0 INTRODUCTION bulk transfer capacity, and achievable Active bandwidth estimation tools can not TCP throughput. Although, these only provide network operators with useful metrics appear similar, they are not, information on network characteristics and and knowing one of them does not performance, but also can enable end users generally imply that one can say (and user applications) to perform anything about others. independent network auditing, load balancing, and 19

Azeez Nureni Ayofe. et al / International Journal of Computer Science and Information Technology (IJCSIT) 2010, Vol1 (1), 19-30 Bandwidth estimation tools face estimation time, and increasingly difficult measurement dependency of accuracy on challenges as link speeds increase and probe traffic volume. network functionality grows more complex.  Consistency of estimates Consider the issue of time precision: on  Efficiency: required network load faster links, interval between packets for a given accuracy, and decrease, rendering packet probe required network load for a given measurements more sensitive to timing estimation time. errors. The nominal 1µs resolution of UNIX  Estimation time: required time timestamps is acceptable when measuring for a given accuracy. 120 µs gaps between 1500B packets on 100  Dependency of accuracy, Mb/s links but insufficient to quantify efficiency and estimation time on packet inter arrival time (IAT) variations of overall network load as well as 12 µs gaps on GigE links. overall machine (both sender and receiver) load. The research community is developing a set of metrics and techniques for active PRTG and OAUNETMON monitor bandwidth measurement. Many of them are network and bandwidth usage as well understood and can provide accurate well as various other network estimates under certain conditions. parameters like memory and CPU usages, providing system Also, some partial measurement and administrators with live readings evaluation studies of bandwidth estimation and periodical usage trends to tools have been published but my attention optimize the efficiency, layout, and has been drawn to the lack of publicly setup of leased lines, routers, available comprehensive and comparative firewalls, servers, and other network experimental results. components.

As we performed test over prtg and This project provides a comparative oaunetmon, comparative analysis study of two Available Bandwidth procedures were being developed. Among (ABw) techniques (PRTG and these, a number of generic criteria were OAUNETMON) taking into account highlighted. These criteria have been the statistical conditions of the ABw, classified into two groups: a first set of several improvement of criteria specified as simple numerical and OAUNETMON in terms of accuracy scalar metrics that provide a partial and efficiency, and a performance comparison of tools performance, a second evaluation of OAUNETMON under a set of criteria that provide comparative DiffServ environment. evaluation of tools performance. The same study was carried out for The first set includes metrics that can be PRTG and the results obtained were useful to quickly choose or discard some compared with OAUNETMON and techniques and tools according to simple recommendations were made only to constraints such as required time or know which among the two to allowed probe bandwidth, some of these choose when a bandwidth is to be metrics are the following: measured in order to obtain  Total probe traffic generated; accurate, reliable and up-to-date  Maximum attainable accuracy; results.  Total estimation time;  Traffic IN; 2.0 RESEARCH JUSTIFICATION  Traffic OUT. The following are the contributions of The second set of criteria for comparative this research work in particular: analysis between different techniques  A well understood bandwidth includes but is not limited to the following: analysis could be used to build  Accuracy: maximum accuracy for a multicast routing trees more fixed time, maximum accuracy for a efficiently and dynamically. limited amount of probe traffic, Ideally, multicast routing trees dependency of accuracy on would be built so that packets 20

Azeez Nureni Ayofe. et al / International Journal of Computer Science and Information Technology (IJCSIT) 2010, Vol1 (1), 19-30 travel along a tree that minimizes applications as well as networks duplicate packets and latency while operation. maximizing bandwidth; - To give aggregate, detailed overview about the traffic  To provide proactive problem patterns on the network of each detection, efficient troubleshooting tool and to recommend the one and rapid problem solution that is more acceptable in terms concerning bandwidth especially of intrusiveness, reliable and when there is the need to improve consistent. on the low bandwidth of a network; - To evaluate available bandwidth measurement tools (prtg and  To determine which among the tools oaunetmon) and undertake a is good for bandwidth measurement comparative analysis in terms of with a mean of getting detail results accuracy, intrusiveness and in terms of accuracy, reliability, response time of active probing intrusiveness and good response tools. time; - Know the challenges facing a low bandwidth network and to  To quickly and easily set up and run probably proffer suggestions on a monitoring station for networks. how best to manage such to give With just a few mouse clicks a a better performance. person can log the amount of data flowing through routers and leased 4.0 RESEARCH METHODOLOGY line, monitors CPU utilization, To compare the performance of the analyze the traffic by type, or check measurement tools studied; disk space utilization; OAUNETMON and PRTG, a set of  To monitor the bandwidth usage of experiments were carried out on an leased lines, routers, and firewalls isolated network that comprises three via SNMP, packet sniffing, or Cisco 1700 series routers that is SNMP NetFlow; enabled connected through  To be able to obtain a graphical FastEthernet links. The network built overview as well as detail statistics for tests, comprise sender and receiver of network usage; hosts, belonging to different LANs on  To have opportunity of storing and which several cross-traffic sources are analyzing traffic in a switched active. network such as Fast Ethernet and Gigabit Ethernet. The used hosts are equipped with Debian Linux and have the same 3.0 RESEARCH AIM AND OBJECTIVES hardware configuration. A couple of The aim of this project work is to carry hosts are used as the sender and the out a research work on the receiver part of the measurement tool, measurement of comparative evaluation while the other hosts are used as the of available bandwidth estimation tools source and the destination for the in a campus Wide Area Network (Prtg cross-traffic. The choice of an isolated and Oaunetmon). network is motivated by the need to totally control the network under test. The objectives of this research work The methodology adopted so far include include the following: the following: - To ease the deployment of platforms that takes the advantage of the huge  An extensive review of relevant amount of available experimental and related literatures on resources to provide a solid previous work done in projects experimental basis for research on comparing bandwidth active bandwidth estimation. measuring tools; - Assess to what extend current  Running of the two bandwidth bandwidth estimation tools can be measuring tools on the same used as basis for providing a network with equal number of bandwidth estimation service for system; 21

Azeez Nureni Ayofe. et al / International Journal of Computer Science and Information Technology (IJCSIT) 2010, Vol1 (1), 19-30  Implementation of these measuring network such as memory and CPU tools (OAUNETMON and PRTG) on utilization. the same operating system;  Conducting two series of With PRTG Traffic Grapher the user reproducible laboratory test in a receives comprehensive bandwidth and fully controlled environment using network usage data that helps to two different sources of cross-traffic; optimize the efficiency of the network.  Experimenting on high-speed paths Understanding bandwidth and resource in real networks where I had a consumption is the key to better complete knowledge of link network management: Avoid bandwidth capacities and had access to SNMP and server performance bottlenecks; counters for independent cross- Find out what applications or what traffic verification; servers use up your bandwidth; Deliver  Obtaining the final results for both better quality of service to your users by the oaunetmon and prtg after a being proactive; Reduce costs by buying consistent and reliable period to bandwidth and hardware according to avoid doubt in the statistical value actual load. and graphical interpretation values obtained; The Freeware Edition of PRTG Traffic  Comparing the accuracy and other Grapher is completely free for personal operational characteristics of the and commercial use and may be tools, and analyze factors impacting downloaded for free. PRTG Traffic their performance; Grapher is designed to run on a  Then, based on the data collected Windows machine in your network for from the above studies, 24 hours every day, and constantly recommendations and suggestion on records the network usage parameters. improving the performance of the The recorded data is stored in an bandwidth networks will also be internal database for later reference. made. Using an easy to use Windows interface you can configure the monitored Our research takes one step further the sensors as well as create usage reports. testing and comparing publicly (prtg and For remote access PRTG Traffic Grapher oaunetmon) for available bandwidth comes with a built-in web server to estimation. First, I conducted two series of provide access to graphs and tables. reproducible laboratory tests in a fully controlled environment using two different All common methods for network usage sources of cross-traffic. Second, I data acquisition are supported: SNMP: experimented on high-speed paths in real Simple Network Management Protocol is networks where there is a complete the basic method of gathering knowledge of link capacities and had access bandwidth and network usage data. It to SNMP counters for independent cross can be used to monitor bandwidth traffic verification. Finally, I compare the usage of routers and switches port-by- accuracy and other operational port as well as device readings like characteristics of the tools, and analyze memory, CPU load etc. Packet Sniffing: factors impacting their performance. With its built-in Packet Sniffer PRTG can inspect all network data packets passing the network card to calculate the bandwidth usage. NetFlow: The 5.0 Overview of PRTG - Paessler NetFlow protocol is supported by most Router Traffic Grapher Cisco routers to measure bandwidth PRTG Traffic Grapher is an easy to use usage. Windows application for monitoring and classifying bandwidth usage. It provides 5.1 How to Choose a Method of Data system administrators with live readings Acquisition and long-term usage trends for their There are four methods of Data network devices. PRTG is mainly used for Acquisition when using PRTG for bandwidth usage monitoring, but may also measuring bandwidth. They are: be used to monitor many other aspects of a  SNMP Monitoring; 22

Azeez Nureni Ayofe. et al / International Journal of Computer Science and Information Technology (IJCSIT) 2010, Vol1 (1), 19-30  Packet Sniffing; The developed non intrusive traffic  Netflow Monitoring; monitoring and analysis system  Latency Monitoring. (OAUNETMON) is different from similar tools because it is non intrusive. It uses 5.2 PRTG Network Monitor System open source software and has the Architecture capability to show traffic information on PRTG Network Monitor consists of three Global, host and protocol basis. It can main parts: keep historical information about PRTG Core Server: The central part of a network usage. It can also give a PRTG installation is the “Core Server” that graphical overview as well as a detailed includes data storage, web server, report statistics. It has the ability to monitor engine and notification system. and analyse traffic in a switched network such as Fast Ethernet and Gigabit Ethernet. Apart from monitoring  PRTG Probe: A “Probe” performs the and analysing network traffic, actual monitoring. It receives its OAUNETMON can also carry out proxy configuration from the Core Server, log analysis using Webalizer. runs the monitoring processes and

delivers monitoring results back to The Obafemi Awolowo University the Core Server. A Core Server can Network Monitoring (OAUNETMON) manage an unlimited number of Experiment was conducted with the aim Probes in order to achieve multi of monitoring live ntwork without location monitoring. adversely imparting on the performance  PRTG Web Interface: The user and also to identify and monitor traffic interface is completely web-based. patterns (both speed and volume) on the The users access the configuration basis of host (IP address), protocol and and monitoring results using a time of the day. standard web browser.

For this experiment, open source Core and probe are Windows services which software was used for the reason of ease are run by the Windows system without the modification and cost. On a high-speed requirement for a user to be logged in. network, the system may not be able to handle all packets in time. If the The monitoring system runs on the Linux processing speed or the system itself operating system. Monitoring was done operates at speed insufficient to capture using open source IPTraf software. A all packets, the analysis result is wrapper script was used to start the unreliable. IPTrafprogram with suitable command line arguments to sort the output of IPTraf The project was to set up a single based on IP address and protocol and it monitoring station that will monitor all creates suitable inputs for PRTG and the traffic. Webaliser. The figure below shows the architecture of OAUNETMON as it was used to After starting up IPTraf and initializing monitor the traffic on the University of some variables, the wrapper script is Ibadan Network. responsible for processing the summary information from IPTraf into suitable log formats both for PRTG and Webaliser. The software aims at doing the processing very fast and as such it is not complicated. In order to maintain system stability and minimize memory leaks, the software restarts every 24 hours. Other script used are run_prtg, prtg_reader, run_webaliser and webaliser_caller.

6.0 OAUNETMON OVERVIEW

Azeez Nureni Ayofe. et al / International Journal of Computer Science and Information Technology (IJCSIT) 2010, Vol1 (1), 19-30

Figure 1: Design Architecture of Prtg

PRTG _Reader Webalizer_Caller

PRTG Webalizer

Webalizer PRTG Log File Log File

Web Server Web Pages

Web Browser (User)

Network of University of Ibadan

Azeez Nureni Ayofe. et al / International Journal of Computer Science and Information Technology (IJCSIT) 2010, Vol1 (1), 19-30

INTERNET

CISCO 1600 ROUTER SNMP ENABLED

1 GIGABIT SWITCH 10/100/1000MBPS 1 GIGABIT INTERFACE MONITORING STATION PENTIUM IV CPU

100 Mbps 100 Mbps Firewall WEB SERVER 100 Mbps Proxy 1

INTERNET WEB NON WEB/FTP SERVER WEB /FTP

University of Ibadan (UI) Network

Oaunetmon Traffic Monitoring Setup at University of Ibadan

Figure 2: Design Architecture of oaunetmon

Azeez Nureni Ayofe. et al / International Journal of Computer Science and Information Technology (IJCSIT) 2010, Vol1 (1), 19-30 iv. Speed required in 7.0 RESULTS AND DISCUSSION generating the output; v. Code size and 7.1 INTRODUCTION vi. Bandwidth requirement The results obtained for both the for the network the tools PRTG and OAUNETMON are analyzed. presented in the tables below. For 7.2.1 The Network each of the tools there is a Coverage corresponding value for the The network coverage value tells following: whether the monitoring was on i. bandwidth Traffic IN (kbyte all the time and covered the and kbit/second); complete time period. ii. Bandwidth Traffic OUT (kbyte and kbit/second); From the table above, the result iii. Sum of Bandwidth Traffic IN shows that OAUNETMON has and Bandwidth Traffic OUT nearly 100% network coverage (kbyte and kbit/second); unlike PRTG with poor network iv. Coverage (%); coverage characteristics. This v. Total volume for parameters feature undoubtedly shows from i-iv above; efficiency and reliability in vi. The Average for parameters OAUNETMON which is a from i-iv above. characteristic to rely on for any Each of the above parameters are properly bandwidth measuring tool in observed and studied into detail for each of order to obtain accurate result. the tools and conclusions were drawn from their characteristics. 7.2.2 Daily Result It is important to note that all the hardware Characteristics requirements for both tools were the same The daily results obtained shows in order to obtain dependable and that PRTG did not give any consistent results on the same network. result for few days. In OAUNETMON, this did not arise. For OAUNETMON the results were obtained For each of the day, the results at every five (5) minutes interval for a were fully and the required data complete day. But for PRTG, the results were recorded. Daily Result obtained indicate a daily values for each of Characteristics indicates that the parameters stated above. OAUNETMON is still far better in relying on the precision of the Hence, the Totals for 10-June-08 from 9:30 result unlike PRTG which PM to 11-June-08 till 9:30 PM indicate a omitted some results in most of daily average as it was obtained for PRTG its daily readings. which is on daily basis. The values obtained for each of the tools were used to determine 7.2.3 Speed Required for the bandwidth requirements for the Implementing the network. Tools.

7.2 RESULTS OBTAINED When both PRTG and The discussion on the results OAUNETMON were both obtained from these experimental implemented on the network, it works was analyzed under the was clearly observed and noticed following criteria: that OAUNETMON runs faster; i. The Network Coverage two times the time required by characteristics of each of the PRTG to be implemented on the tools; same network. ii. Daily results characteristics; iii. Speed required for 7.2.4 Speed required in implementing each of the generating the tools; output 26

Azeez Nureni Ayofe. et al / International Journal of Computer Science and Information Technology (IJCSIT) 2010, Vol1 (1), 19-30 Of course, when the speed used by a ii. Total Volume for particular tool is faster than the Bandwidth Traffic other, what this mean invariably is OUT (Kbyte) = that the time when the output will 933, 737.354 be generated shall definitely be Therefore, the amount of bandwidth different. OAUNETMON takes the required in this case is calculated as opportunity of having a better follow: running time to generate its result in a better time than PRTG. It was Bandwidth = (Total Volume for noticed that 3 minutes after Bandwidth Traffic IN (Kbyte)/8) Mbps OAUNETMON would have started = (2,202,013.789/8) generating its results that PRGT will Mbps just begin displaying its result on = 275251.723625 Mbps the same network. is the Bandwidth required using Oaunetmon. 7.2.5 Code Size PRTG has a better advantage over b. FOR PRTG OAUNETMON in this regard. The From 23-May-08 to 01-Mar-08, size of code required for the following results were obtained from implementation on the network for prtg: PRTG is 7,302kb whereas (i) Total volume for OAUNETMON has 8,323kb. Bandwidth Traffic IN 7.2.6 Bandwidth requirement (Kbytes) = There are two approaches to getting 484,585,022.15 bandwidth in this case: 0 i. Comparing the average Total (ii) Total volume for Volume Bandwidth obtained Bandwidth Traffic OUT (Kbytes) = for both the prtg and 413,246,247.382 oaunetmon; To calculate the equivalent bandwidth ii. Comparing the Random requirement as we have in oaunetmon Selection of Bandwidth for prtg, the following procedures are to obtained for prtg and a daily be followed: bandwidth obtained for oaunetmon. Bandwidth/per day = Total volume for bandwidth Traffic IN 7.3 Comparing the average Total (Kbyte)/Total Volume Bandwidth number of days

=484,585,022.150/74 a. For OAUNETMON: This is an important =6548446.24527 parameter for determining the efficiency, Hence, the bandwidth required by prtg dependability, accuracy and precision of = (6548446.24527/8) Mbps any bandwidth measuring tool. = 818555.78 Mbps From 10-June-08 From the calculation shown above, it is 9:30 pm to 11-June-08 9:25pm- clear that the bandwidth required when 9:30pm as shown on the statistical oaunetmon was used is better compare table obtained for oaunetmon and to bandwidth required by ptrg. which was obtained at 5 minutes interval the following values were The bandwidth required by the obtained: oaunetmon is 275251.723625 Mbps i. Total Volume for while the bandwidth required by the Bandwidth Traffic IN prtg is 818555.78 Mbps which is almost (Kbyte) = 2, four times bandwidth required by the 202,013.789 oaunetmon.

Azeez Nureni Ayofe. et al / International Journal of Computer Science and Information Technology (IJCSIT) 2010, Vol1 (1), 19-30 7.4 Comparing the Random Bandwidth April,2008 and 29th April,2008 the values Bandwidth Traffic OUT for PRTG are 4,585,835.226Kbyte and For OAUNETMON, from the result 2,887,869.427Kbyte respectively. In this shown above, it is noted that the case, the Bandwidth requirement for Total Volume for the Bandwidth Traffic IN is PRTG is poor when compare to 2,202,013.789 Kbyte and Total Volume for oaunetmon. the Bandwidth Traffic OUT is 933,737.354Kbyte. This approach shows that neither OAUNETMON requires less/higher But for the PRTG, since there is more than bandwidth nor PRTG requires such. a day value for Total Volume of Bandwidth There is no definite conclusion with this Traffic IN and Bandwidth Traffic OUT, the approach. analysis is therefore base on randomly selection of values. Finally, with analysis of the two approaches to bandwidth requirement For example, on 23rd May,2008 the of the two bandwidth measuring tools, it Bandwidth Traffic IN is 2,106,288.017Kbyte is noted that oaunetmon is better in and the Bandwidth Traffic OUT is bandwidth requirement than prtg. 1,185,471.328Kbyte for prtg and this is better than oaunetmon. But on 30th

The table below gives a better interpretation of the above parameters used in analyzing the differences between the prtg and oaunetmon. NETWORK SPEED OF RATE OF CODE DAILY RESULT BANDWIDTH COVERAGE RUNNING GENERATING SIZE FEATURES REQUIREMENT (KB) (%) SECONDS OUTPUT

PRTG 91 57 Slower 7,302 Omission of Higher results were noticed Oaunetmon 99 26 Faster 8,323 Full results Lower

Table 4.1: Showing the characteristics features of prtg and oaunetmon as demonstrated on the network.

8.0 Summary both the network operators and Some partial measurement and user applications to make use of evaluation studies of bandwidth a better tool in order to get a estimation tools have been better network auditing load published, nonetheless, my balancing and a good server attention has been drawn to lack selection as well as the of publicly available realization of other aims within comprehensive and comparative which this research was carried experimental results. Not only out. that, there is lack of common and consistent practices and 9.0 CONCLUSION procedures to test available I have presented the systematic active bandwidth measurement comparative analysis of active tools. bandwidth estimation tools (prtg and oaunetmon) performed over With this new research carried a broad range of tools, scenario out on comparative analysis of (both emulated and real) as well both prtg and oaunetmon, it is as experimental conditions, therefore a great advantage for 28

Azeez Nureni Ayofe. et al / International Journal of Computer Science and Information Technology (IJCSIT) 2010, Vol1 (1), 19-30 which results are publicly tools, what stands to be a challenging available. issue now is how to develop a tool that will compute available bandwidth at Techniques tools ton estimate different time of the day. capacity, available bandwidth and bulk transfer capacity have This, undoubtedly, will give overtime been studied jointly. statistics of usage pattern (peak an off Furthermore, I have tried to peak) and to try to investigate why the analyze and to explain errors network is not utilizing the maximum and uncertainties observed on capacity of all times and what could be the tools I have studied. consuming the unused bandwidth.

The results obtained shown that Carrying out this project will go a long oaunetmon is the least intrusive, way in explaining into detail the concept reliable, consistence and of bandwidth utilization and dependable and in some cases management. It will assist the network can be very slow through with manager understand full how the better accuracy and precision available bandwidth can optimized. when compare to prtg REFERENCES. Finally, oaunetmon offers the 1. Hong J. Kwon S. and Kim. best performance with regard to the criteria studied: (1999) WebTrafMon: Web- i. The Network Coverage BasedInternet Intranet network characteristics of each of the tools; traffic monitoring and analysis ii. Daily results system. Elsevier Computer characteristics; iii. Speed required for Communications. Vol. 22 issue: implementing each of the 14. pp 1333-1342. tools; iv. Speed required in 2. O.O. Abiona, C.E Onime, A.I generating the output; Oluwaranti, E.R Adagunodo, L.O v. Code size and vi. Bandwidth requirement Kehinde, S.M Radicella: for the network and the Development of a Non-Intrusive tools analyzed. The study presented in this project Network Traffic Monitoring and was on the small number of criteria Analysis System. African that seemed to be the most important. However, this study must be Journal of Science and completed by considering other Technology (AJST) Science and parameters and by evaluating the measurement tools on real networks Engineering Series Vol. 7, No, 2, settings. pp.

As a result from the tests carried out, 3. Ahmed Ait Ali, Fabien Michaut, a number of failure conditions for the Francis Lepage: End-to-End analyzed tools have been identified, as well as the dependency of the Available Bandwidth estimates accuracy on factors such as Measurement Tools: A system load and network properties. Comparative Evaluation of

Performances. 10.0 FUTURE WORK I have seen a great challenge to this 4. Kevin Lai, Mary Baker: project; apart from making a Measuring Bandwidth comparative study to these bandwidth 29

Azeez Nureni Ayofe. et al / International Journal of Computer Science and Information Technology (IJCSIT) 2010, Vol1 (1), 19-30 5. Jacob Strauss, Dina Katabi, Frans Kaashoek: A Measurement Study of Available Bandwidth Estimation Tools. MIT Computer Science and Artificial Intelligence Laboratory.