Managing Latency in LR Results Reporting
In my organization, we have regional load injectors based in the network locations of our facilities (Malaysia, Germany, & US). Malaysia performance even for a single user can be as high as 2000% worse response times than what we are seeing in US. I understand that the average latency in Malaysia is between 250-400ms whereas our location is 30-60ms in US.
This is my first experience with a location with latencies this high and am having a difficult time with the question from leadership "how much impact does latency have on the numbers presented by LoadRunner". For example, if a given response time is 5s in US and 20s in Malaysia, they want to know how much of that delta may have been caused by Malaysian latency rates. Has anyone dealt with this problem before? Are there any industry standards to correct for and properly analyze results to take latency into account? Say in the example above, if we project Malaysia with the same latency as in US - could this potentially reduce response times to closer to US standards?
Any insight into this thorny issue would be appreciated!
Re: Managing Latency in LR Results Reporting
This is more of a test process issue versus a tool issue. With so many uncontrolled network elements in your test it is essential for you to have a reference network controlk factor. When you are testing a new application, use a reference "control" generator on the same network segment as your application under test with a single user of each type present. Use the statistics from your "control group" to determine the absolute impacts of network complexity for the rest of your load generator population.
Essentially what you are doing now is testing two unknowns togetherm, application and network, and then trying to decouple the influence of one part from another. The network monitor won't help you here, you need the control generator.
Replace ineffective offshore contracts, LoadRunnerByTheHour
. Starting @ $19.95/hr USD.
Put us to the test, skilled expertise is less expensive than you might imagine.
Twitter: @LoadRunnerBTH @PerfBytes