One of the most interesting features of the Qure Analyzer is the ability to compare different workloads. The possible applications for this feature are manifold: One could for example run identical traces against two different servers and benchmark them against each other. Or, for example, benchmark two different versions of SQL Server on identical hardware to get a feeling if, and how much, performance gain you would get by moving to a different version.
There are many different scenarios where this feature comes in handy, but I’m not going to describe it here in full. This is something I leave for you to work out in your own environment. However, I am going to describe some of the underlying concepts in the Qure Analyzer. The tool differentiates “controlled” and “uncontrolled” workloads. Here is the definition of these terms taken from the Analyzer User Guide:
Controlled Workloads contain an identical set of SQL commands and are replayed under two different conditions. Usually the testers modify only one dimension of the environment between replays. They may be evaluating a code change, a server upgrade or a different piece of hardware for example. Using Controlled Workloads offer one means to achieve an apples-to-apples comparison.
Uncontrolled Workloads contain a similar but non-identical set of SQL commands. When you capture two different sets of traces in production for example, these workloads are usually uncontrolled, because in a production environment, transactions occur as business needs dictate, and so they are not predictable.
In any comparison of workloads you define one workload as the baseline to which the other one is compared and benchmarked. Think of this as some kind of “before” and “after” snapshot where in between something else significant has happened. This could be the above mentioned new hardware, or the new software version, or maybe some tuning efforts that have taken place in between.
The result of the comparison is displayed on its own tab and can be aggregated and filtered in the same as single workload details report. The big difference is the colouring of the bars in the grid. Green stands for a decrease in resource consumption, while red stands for an increase in resource consumption.
Describing the full details of a workload comparison feature is beyond the scope of this review. I’d rather refer you to the excellent explanations in the User Guide or in this short series of detailed articles, published by DBSophic as part of their quarterly newsletter:
- http://www.dbsophic.com/Tips/tip-comparing-workloads-qwa.html
- http://www.dbsophic.com/Tips/tip-comparing-db-workload-part-2.html
Conclusion
The Qure Analyzer is a great tool for anyone concerned with SQL Server performance tuning. It is easy to use and can give you both, a quick overview as well as deep analysis into an SQL Server environment.
The fact that it is free makes it even more interesting and an absolute candidate for every SQL Server professional’s toolset.
]]>