Whenever a new application is rolled out on the network, IT must be able to predict with confidence how it will perform for the business. Moreover, other mission-critical applications must not be negatively impacted. Solutions that provide network performance analysis capabilities should be able to provide immediate feedback on how new applications behave on the network and can predict the impact to existing, business-critical applications.
Here are three warning signs that a new application on the network is not performing well on the network and / or is negatively affecting other business-critical applications.
1) Performance in production doesn’t match pre-deployment testing/analysis
Pre-deployment testing can take many forms, from a relatively simple set of tests followed by predictions of the network impact, to complete test networks that simulate the entire usage pattern of the application. While the latter approach can be complicated and time consuming, the former technique is simple to perform with a few basic tools. By installing the application and capturing some representation transactions with packet-based network analysis software, one can perform “what if” analysis that allows one to “scale” these representative transactions to the expected number of users of the application once in production. The resulting analysis provides a clear assessment as to the readiness of the network for the new application. Additionally, it will give the data needed to compare application performance once the application is in production. Any discrepancies are a red flag and should be addressed.
2) Network baselines change dramatically after deployment
Without network baseline data organizations are operating blind, whether it pertains to the specific issue of deploying a new application, or just general day-to-day network performance management. Baseline data should include long-term (weekly cycles work well) sets of network utilization data from key network interfaces; node, protocol and application utilization over time; and an assessment of the typical “noise floor” of the network – typical packet loss numbers, number of retransmissions, latency measurements for key applications, etc. After deploying a new application (and as part of the pre-deployment testing) a new set of baseline data should be generated and compared against data from before the rollout. Unexpected differences should be addressed.
3) A change in network “health”
Network health ties into the aforementioned “network noise floor”, but changes in network health can usually be seen much more quickly than comparing long-term baseline data. Enterprise-quality network analysis software will constantly monitor the vital signs of the network by performing background expert analysis, and it will send alerts when vital signs begin to drift. For example, if an organization deploys a new application and its network analysis software begins sending alerts that there are too many TCP retransmissions, this is a strong indication that this new application has tipped a balance. (even without even knowing the network’s current threshold for TCP retransmissions).
No matter what solution an organization selects for analyzing how a new application behaves on a network, it must meet the following criteria …
1) Provides detailed information through all seven layers of the OSI model
2) Enables collection of key network data over meaningful time cycles (like a week) to develop high quality network baseline data
3) Quickly identifies if the network is being stressed and where
4) Offers “What If” capabilities for analyzing the impact of new developments and implementations?