Most management solutions for the VMware environment use the commodity performance and availability metrics provided by VMware APIs to drive their analysis, automation and reporting functions. However, in all environments, but most especially dynamically changing virtual ones, measurements of resource utilization such as CPU, IO, network packets, memory etc. don’t directly relate to the actual performance (throughput and response time) of the applications and services running in those environments. No matter how “intelligent” the rules put in place to manage using such metrics are fundamentally flawed. Such performance metrics that are too high (or low),or even statistically analyzed or combined to attempt to understand normal behavior still have no bearing on how much useful work is actually getting done, or if it is getting done efficiently (or in time!). Therefore automation based on such metrics and analysis is intrinsically sub-optimal (at best), and erroneous (at worst).
Only analysis that looks at the actual amount of work being done (transactional throughput) and the time spent getting it done (response time, or latency) can accurately and cost-effectively optimize performance and capacity risk/reward over time. Just because your average utilization is low doesn’t necessarily mean you have excess resources that can be consolidated, repurposed, or eliminated. And, it doesn’t mean that Application Performance is necessarily acceptable. Or, just because your average utilization is high doesn’t necessarily mean you are cost-optimized or failing to meet service performance goals. Your response time may still be unacceptable because of waiting on a particular, critical resource. You need to understand where, when and why applications are waiting on key resources. Waiting is called queuing, and queuing is always a red flag. Any management or optimization that doesn’t understand the components of response time: service time and queuing time simply misses the mark!
Commodity Metrics = Commodity Results.
Commodity Results = always buying more of the commodity! (which is what the commodity vendors want!)
Optimized Results = No Queuing AND no idle resources
TeamQuest Analytics constantly evaluate application/service Latency; fully understanding the past, present and future to continuously and cost-effectively ensure service delivery – optimizing the performance and capacity of your infrastructure and applications and services.