Database Reference
In-Depth Information
memory available during peak)
Baseline performance comparison between physical and virtual systems showing
application metrics, such as database transactions per second and transaction
latency for top systems during validation benchmark testing
Baseline analysis of physical systems, including database transactions per second,
latency, and throughput metrics for the top-5 systems
Detailed findings (detailed system metrics for each system baselined)
Processor load by system, business hours average, and peak load
Disk transactions/s, throughput and latency, % disk busy, and disk queue length
Memory metrics by system, including configured RAM, available RAM Mbytes,
and % business hour average and peak consumption
Detailed Database Transaction metrics by database instance and schema,
including transaction SQL, average reply bytes, average latency, max latency, hits
per second, and total hits
System inventory by system name, including OS and version, # CPUs, CPU MHz,
Total MHz, Installed RAM, # SCSI controllers, # disks, # NICs, and NIC speed
All of this information will play an important part in your design and implementation of
your SQL Server databases running on VMware vSphere.
Performance Traps to Watch Out For
This section will cover some of the common performance traps that you need to watch
out for during your SQL Server and vSphere infrastructure baselining and benchmarking
exercises. Some of the topics covered might be more obvious than others, but all are
important when you want to ensure performance and availability of your SQL Server
databases when they are virtualized.
Shared Core Infrastructure Between Production and Non-
Production
If there are any shared core infrastructure components between production and non-
production that might become saturated or overloaded, this could cause your baselines
and benchmarks to be invalidated in the best case, or, worse, potentially cause
production outages during performance benchmarks. As part of your analysis and design
process, you should evaluate what shared core infrastructure components there are that
might impact your baselining and benchmarking, such as shared storage arrays, network
components, and firewalls.
You need to determine if any of the links in the chain where you will be monitoring and
testing will likely exceed their capacity. It could be something as simple as exceeding
