Category list

Performance and Fault Management: A Practical Guide to Effectively Managing Cisco Network Devices (Cisco Press Core Series) - page 66

Summary

This chapter presented several statistics that can be important for network management: mean, mode, median, range, standard deviation, and variance. Although the amount of work needed to analyze a single MIB variable may seem intimidating, please note that it is not necessary to analyze every variable that you have in your network baseline. Although the application of the correct statistic will assist you in the analysis of data and variables , it is more important to use the knowledge of your business operations to define critical applications and how their successful performance is measured. After you have accomplished this, you can use the statistical methods in this chapter to translate "how successful performance is measured" into what needs to be measured.

References

Books

Buchanan, Robert. The Art of Testing Network Systems . New York, NY: Wiley, 1996.

Evans, Merran, Nicholas Hastings, and Brian Peacock. Statistical Distributions . New York, NY: Wiley, 1993.

Ferrari, D. Computer Systems Performance Evaluation . Upper Saddle River, NJ: Prentice Hall, 1978.

Gilchrist, W. Statistical Forecasting . New York, NY: Wiley, 1976.

Jain, Raj. The Art of Computer Systems Performance Analysis . New York, NY: Wiley, 1991.

Kececioglu, Dimitri. Reliability Engineering Handbook . Upper Saddle River, NJ, Prentice-Hall, 1991.

Naugle, Matthew. Network Protocols . New York, NY: McGraw-Hill, 1994.

Robertazzi, Thomas G. Computer Networks and Systems: Queueing Theory and Performance Evaluation , 2nd ed. Springer, 1994.

Spohn, Darren L. Data Network Design . New York, NY: McGraw-Hill, 1993.

Stallings, William. SNMP, SNMPv2, and RMON , 2 ^nd ed. Reading, MA: Addison-Wesley, 1996.

Terplan, Kornel. Benchmarking for Effective Network Management . New York, NY: McGraw-Hill, 1995.

Periodical

Foster, K. " Math on the Internet. " IEEE Spectrum , Volume 36, Number 4, April, 1999.

Standards

RFC 1757, " Remote Network Monitoring Management Information Base, " Steven Waldbusser, February 1995.

RFC 1857, " A Model for Common Operational Statistics, " M. Lambert, October 1995.

RFC 1902, " Structure of Management Information for version 2 of the Simple Network Managemenet Protocol, " Case J., McCloghrie K., Rose M., and Waldbusser S., April 1993.

RFC 2063, " Traffic Flow Measurement: Architecture, " Brownlee, N., Mills, C., and G. Ruth, January 1997.

RFC 2064, " Traffic Flow Measurement: Meter MIB, " Brownlee, N., January 1997.

Chapter 8. Understanding Network Management Protocols

Network managers need to have a base level of knowledge about the protocols used to do network management. This chapter will help you acquire this knowledge or review the protocols if you already are familiar with them. The chapter is designed to give you an overview of the most-used protocols and to discuss Cisco-specific information concerning them.

The protocols that are covered in this chapter are as follows :

Ping
Traceroute
Terminal emulators
SNMP
RMON MIB
Syslog
Cisco Discovery Protocol
Name service

Ping

Ping is commonly used to check connectivity between devices. It's the most common protocol used for availability polling. It can also used for troubleshooting more complex problems in the network.

Ping uses Internet Control Message Protocol (ICMP) Echo and Echo Reply packets to determine whether one IP device can talk to another. Most implementations of ping allow you to vary the size of the packet. Table 8-1 shows packet size statistics for Cisco IOS and Catalyst devices.

Table 8-1. Ping Packet Sizes for Cisco Devices

Device	Minimum Packet Size	Maximum Packet Size	Default Packet Size
Cisco IOS	36 bytes	18024 bytes	100 bytes
Cisco Catalyst	56 bytes	472 bytes	64 bytes

Most host-based implementations of ping send one packet per second. Cisco IOS allows you to send packets as fast as the CPU and the network device can generate them and the network can take them. Of course, such stress testing should be done with discretion on production networks.

There are several issues that you should be aware of when using ping. Pings are generated and answered by using the CPU on Cisco devices. Although pings do not usually take a huge amount of CPU time, you should be cautious about using heavily loaded devices to generate or answer pings. On Cisco IOS devices, the task that processes pings runs at low priority. Therefore, it is possible for a busy device to fail to respond to a ping. If a device you are trying to ping is not in the ARP table, the first ping will often time out on Cisco devices. Because of this, most network-management applications using ping will send several pings before announcing a connectivity loss event. For the same reason, you should be cautious about using ping to determine latency or throughput on a network because you may be measuring the ICMP performance of the end points more than the actual network performance.