Network Performance

1. Network Performance

2. Performance Management-What • What’s performance management? • understanding the behavior of a network and its elements in response to traffic demands • Measuring and reporting of network performance to ensure that performance is maintained at a acceptable level

3. Delay • Delay = Latency + propagation delay + serialization delay • Propagation delay: the time it takes to the physical signal to traverse the path; depends on distance. (add 6 ms for 1000km Fibre link) • The delay from Beijing to Guanzhou is about 34 ms (CERNET), the distance is about 3000Km. • Serialization delay is the time it takes to actually transmit the packet; caused by intermediate networking devices, includes queuing, processing and switching time (normally, less than 1ms for one networking devices, but not firewalls or heavily loaded routers) • Comfortable human-to-human audio is only possible for round-trip delays not greater than 100ms • Latency : Setup delay

4. Jitter • Jitter is the variation of the delay, a.k.a the 'latency variance,' can happen because: • variable queue length generates variable latencies • Load balancing with unequal latency • Harmless for many applications but real-time applications as voice and video • Applications will need jitter buffer to make it smoothly • Tolerable Jitter range for VOIP is: 20ms – 30ms

5. Packet Loss • Loss of one or more packets, can happen because ... • Link or hardware caused CRC error • Link is congested or queue is full (tail drop or even RED/WRED) • route change (temporary drop) or blackhole route (persistent drop) • Interface or router down • Misconfigured access-list • ... • 1% packet loss is terrible and unusable! • Tools: ping etc.

6. Throughput • Network throughput is the average rate of successful message delivery over a communication channel. • This data may be delivered over a physical or logical link, or pass through a certain network node. • The throughput is usually measured in bits per second (bit/s or bps), and sometimes in data packets per second or data packets per time slot.

7. Bandwidth Utilization • The channel efficiency, also known as bandwidth utilization efficiency, in percentage is the achieved throughput related to the net bitrate in bit/s of a digital communication channel. • For example, if the throughput is 70 Mbit/s in a 100 Mbit/s Ethernet connection, the channel efficiency is 70%. • In this example, effective 70Mbits of data are transmitted every second.

8. Network Availability • Network Availability is the metric used to determine uptime and downtime • Availability = (uptime)/(total time) = 1-(downtime)/(total time) • Network availability is the IP layer reachability • Better > 99.9%

9. Packets Per Second (PPS) • Important for performance: network performance is highly affected by PPS, such as delay or packet loss, because the serialization delay will increase because of the load of the intermediate routers • PPS is a very important metric to detect DOS/DDOS traffic

10. CPU and Memory Utilization • How much the CPU and Memory are used. • CPU utilization better less than 30% • For global routing routers, at least 512M memory is needed

11. Quality of Service • Quality of service is the ability to provide different priority to different applications, users, or data flows, or to guarantee a certain level of performance to a data flow. • For example, a required bit rate, delay, jitter, packet dropping probability and/or bit error rate may be guaranteed. • Quality of service guarantees are important if the network capacity is insufficient, especially for real-time streaming multimedia applications such as voice over IP, online games and IP-TV, since these often require fixed bit rate and are delay sensitive, and in networks where the capacity is a limited resource, for example in cellular data communication.

12. QoS • QoS: Quality Of Service • QoS is technology to manage network performance • QoS is a set of performance measurements • Delay, Jitter, packet loss, availability, bandwidth utilization etc. • IP QoS: QoS for IP service

13. SLA and QoS • SLA: Service Level Agreement • SLA is the agreement between service provider and customer, SLA defines the quality of the service the service provider delivered, such as delay, jitter, packet loss etc. • SLA is a very important part of the business contract, and also can be used to distinguish the service level of different ISPs Business Technology SLA QoS

14. SLA example: Level 3 Delay Packet Loss Availability Jitter Bandwidth

15. SLA example: Sprintlink

16. Measurement Technology • We’ve known what metrics used to describe network performance, but how to measure them? • Technologies and tools • ping, traceroute, telnet and CLI commands etc. • SNMP • Netflow (Cisco), Sflow (Juniper), NetStream (Huawei) • IP SLA (Cisco) • Etc.

17. ping • Normally used as a troubleshooting tool • Uses ICMP Echo messages to determine: • Whether a remote device is active (for trouble shooting) • round trip time delay (RTT), but not one-way delay • Packet loss • Sometime we need to specify the source and length of packet using extended ping in router or host • Why using large packet when ping? (to test the link quality and throughput.) • Large packet ping is prohibited in Windows, but Linux is ok

18. Sample Ping Freebsd>% ping 202.112.60.31 PING 202.112.60.31 (202.112.60.31) 56(84) bytes of data. 64 bytes from 202.112.60.31: icmp_seq=1 ttl=253 time=0.326 ms …… 64 bytes from 202.112.60.31: icmp_seq=6 ttl=253 time=0.288 ms 6 packets transmitted, 6 received, 0% packet loss, time 4996ms rtt min/avg/max/mdev = 0.239/0.284/0.326/0.025 ms router# ping Protocol [ip]: Target IP address: 202.112.60.31 Repeat count [5]: Datagram size [100]: 3000 Timeout in seconds [2]: Extended commands [n]: Sweep range of sizes [n]: Type escape sequence to abort. Sending 5, 3000-byte ICMP Echos to 202.112.60.31, timeout is 2 seconds: !!!!! Success rate is 100 percent (5/5), round-trip min/avg/max = 1/1/4 ms

19. traceroute • Can be used to measure the RTT delay, and also the delay between the routers along the path • Unix/linux traceroute uses UDP datagram with different TTL to discover the route a packet take to the destination, Microsoft Windows tracert uses ICMP protocol, If Windows tracert appears to show continuous timeouts, the router may be filtering ICMP traffic – try a Unix/Linux traceroute • After the Nachi worm, many ISPs filter ICMP traffic. So ping can not work, but traceroute is ok 19ms 2ms 15ms 2ms router2 router3 router1 H1

20. Sample Traceroute Router# traceroute 202.112.60.37 Type escape sequence to abort. Tracing the route to 202.112.60.37 1 202.112.53.169 0 msec 0 msec 0 msec 2 202.112.36.250 20 msec 20 msec 16 msec 3 202.112.36.254 28 msec 28 msec 24 msec 4 202.112.53.202 24 msec * 24 msec

21. Visual Route • Visualization of traceroute information • http://www.visualroute.com

22. telnet and CLI commands • Using telnet manually or scripts programmed with Expect to telnet the network device then issue the CLI commands is also a useful and basic monitoring method to get performance data • It’s necessary because some data can only be accessed through CLI commands, and not supported by SNMP etc. How about config file?