Performance Monitoring of Enterprise Networks


Did you know that a video conferencing application needs a latency less than 300 ms to give a good quality video images in your IP Network? There are other real time applications that require the network to maintain an optimum performance. Let us explore what parameters of the network performance could be monitored, in this article.Network performance, needs to be assessed in real time to monitor parameters like jitter, delay and packet loss in the network, to assess its efficiency. This can be done by a Router, application running in server or network analyser tools.

In a nutshell: We have a client source that sends probe queries to a server that responds. The client source sends some packets to the server and the server sends it back with an acknowledgement. Both the type, and the content of the queries are configurable and the server generally is aware that this probe is used to measure RTT (Round Trip Time) and jitter delays. Different packet types can be used within the probe like ICMP echo, ICMP timestamp, http get, udp echo, etc. These probe packets are timestamped both at the source and the destination with the times that they are sent and received from both.

Timestamps:

Timestamps are used to measure the latency in packet communication in both client and server side. All nodes on the network should be preferably synchronized to a common time source that is accurate (Like Stratum 3 Level) using a protocol like Network Time Protocol, to ensure accuracy of measurements. These timestamps can be performed by using hardware or software or both (which is more accurate).

The time-stamping activity consists of measuring the following:

T1 – The time at which the packets leave the source (client).
T2 – The time at which the probe is received at the server.
T3 – The time at which the probe leaves the server.
T4 – Time at which the probe is received by the source (client).

Series of such timestamps are used to measure several metrics : For example RTT (Round Trip Time) is measured by computing the difference between T4 and T1 and then subtracting the difference of T3 and T2 from the former value.

Probe Types:

There can be several probe types which can be used to monitor the performance of the enterprise networks and some of them are: udp-ping-timestamp, udp-ping, icmp ping, tcp-ping, http-get, icmp-ping-timestamp, etc. A target is generally probed using its IPv4 address.

Tests:

Typically, a test is a homogeneous set of probes configured for all the devices that needs to be probed. Generally, multiple  types of probes are not a part of a single test. A test is carried out with multiple such homogeneous probes over which the performance metrics are calculated. So, during a test, such probes are generated and responses collected at a rate defined by the probe interval. An acceptable probe interval is set, and the probe is considered to be lost, if the probe doesn’t arrive back to the client source within that interval.

Performance metrics are measured with multiple tests, each consisting of different probe types sent between a source ip and destination ip. The interval between the tests, probe types and data types used in the probe are user configurable. The number of such probes to be used in each test could also be specified by the user and each probe can be assigned to a traffic forwarding class by specifying the type of DSCP marking to be used in the IP packet header.

Threshold limits are set for different delays and latencies and an alarm can be triggered each time this threshold is exceeded (SNMP traps need to be enabled for triggering alarms).

A lot of parameters can be configured in a test like: test name, target address, source address, probe type, probe interval, test interval, threshold, traps, probe count etc.

Key performance indicators:

¤ Round Trip Time: Time taken for the probe to reach the server from the client source and back.
¤ Egress Time: Time taken for the probe to reach from the client source to the server.
¤ Ingress Time: Time taken for the probe to reach from the server to the client time.
¤ Jitter: The delay experienced in the Ingress and Egress.
¤ Lost probes/ Successive Lost probes: The number of probes that don’t return/reach within the threshold timings. This is also monitored for successive probes.
¤ Latency: This value is calculated from the Round Trip Time values.

excITingIP.com

You could stay up to date on the various computer networking technologies by subscribing to this blog with your email address in the sidebar box that is mentioned as ‘Get email updates when new articles are published’