Database Reference
In-Depth Information
getting an accurate benchmark much easier.
Benchmark Workload Model Based on System Nonfunctional Requirements
The first method can be used if you developed the system originally. In this case, you
may have a copy of the nonfunctional requirements that need to be met, and the original
testing team should have documented the workload model. Over time the workload
model and nonfunctional requirements will likely have changed, and hopefully your
capacity planning team has kept them up to date as part of their standard operating
procedures. If this is the case, you can use this model, in combination with a load
generator, to perform system benchmarks as part of your virtualization project. If the
model and nonfunctional requirements have not been kept up to date, you will need to
modify them to make them as valid and realistic as possible. This type of benchmark
isn't as accurate as a benchmark based on recorded production performance.
Benchmark Workload Model Based on Recorded Production Performance
The second method of developing your benchmark involves recording a baseline from
actual production transactions or system performance metrics and then replaying the
baseline against an equivalent virtualized non-production copy. The recording and
replaying may be achieved by using a specialized tool that has been developed for this
purpose. Alternatively, you may take the recorded system baseline and create a
workload model from it manually that can then be fed into a load generator such as
Apache Jmeter or HP Load Runner.
Baseline Your Performance Benchmark
You would normally baseline the configuration of your benchmark testing by running a
series of simulations and recording the results to ensure they are consistent. When the
test results have stabilized or normalized, you have a baseline of your performance
benchmark. This baseline can then be used as a reference for further benchmark testing
where the configuration has been modified. This is an important step to ensure the
validity of your results and so you have a control you can measure further tests against.
Using Your Baseline and Your Benchmark to Validate Performance
You would use your chosen benchmark to try and simulate the production load against
the non-production version of the system prior to migrating it. This is to ensure the
virtualized system can meet the performance characteristics that were recorded during
the baseline of the production system. Depending on the results of your performance
benchmark tests compared to your production baseline, you may need to make design or
configuration adjustments.
Once you're satisfied with your performance benchmark compared to the baseline, you
 
 
Search WWH ::




Custom Search