Subscribe to this feed

IT performance can appear complicated which can make the diagnosis more difficult than it needs to be. An important performance indicator for the End-user is response time, which can be translated to end-to-end network performance metrics: packet loss, one-way-delay and jitter. These are important network characteristics for the network administrator to document the end-to-end quality of the network.

In this blog I want to start by taking a look at packet loss, which is a common network problem that affects application performance. So why is packet loss such an application performance killer? This mainly relies to the nature how TCP handles packet loss. If within a certain period of time (known as timeouts) the receiving party does not acknowledge a data packet, the sender sends it again. Timeouts caused by packet loss can add up to large delays as timeout values double with each retry.

We find out that a lot of IT admins think that packet loss of 1 or 2% is a small number and should have little impact on the application performance. Unfortunately, already a very small amount of packet loss will have a big impact on the TCP throughput, which in turn will affect the application performance. We recommend that packet loss rates higher than 0,01% should be investigated.



Server Retransmission - who ist the originator?


Client Retransmission - who ist the originator?

Identifying packet loss is with GeNiEnd2End Network little effort. You install either the free software endpoint on already existing PCs/Servers or position at demarcation points the cost-effective GeNiJacks. GeNiEnd2End Network will report besides packet loss also one-way-latency, jitter and other important network quality parameters.

GeNiEnd2End Network is the perfect addition for already existing network-based APM solutions like SuperAgent from CA, Application Xpert from Opnet or TruView from Fluke Networks. It streamlines the diagnosing of network performance problems considerable. With GeNiEnd2End Network it is a piece of cake to prove that "it's not the network".

So let me review in more detail the reasons why an active endpoint-based solution like GeNiEnd2End Network is the best mate of a network-based APM approach:
Within a network-based APM solution packet loss is reported with the metrics client and server retransmission. However, the originator (network, server of client) and the direction (backward/forward) of the packet loss can't be determined clearly. Have a look at the bounce charts above that reveal this.

However, with GeNiEnd2End Network you can decide quickly whether packet loss is caused by the network or not. Detecting packet loss is the first step for the network team in diagnosing application performance problems caused by the network. GeNiEnd2End Network streamlines the troubleshooting workflow and enables the network team to become productive in isolating the problem domain.

Interested to attend a product web session? Then let us know.

GeNiEnd2End Network - the real Troubleshooting tool for the network engineer!