Benchmarking for mixed workloads

In a production environment, a combination of online and batch workloads occurs throughout the workday. A benchmark that combines both helps determine the necessary server resources to handle such a load. This is day-in-the-life benchmarking.

In this combination of online and batch testing, use your usual online benchmark workload. Run your batch testing as you would normally run a production batch workload, even if it during peak hours. Start with a minimal number of CMDB batch processing, then increase it by as much as your environment can handle without impairing performance. Ensure that you configure for normalization and reconciliation private queues to handle the CMDB load separately from the online threads. Using private queues and limiting the number of threads for each queue restricts how much CPU each job uses and prevents the CMDB from consuming too much CPU and adversely impacting end-user online transaction response times. When combined with an online load, a CMDB load consists of updating CIs and loading new ones, which are generally run as continuous normalization and reconciliation jobs.

When you have reached a peak, anything more requires a separate server to handle a strictly CMDB load.