CPU
CPU statistics provide information about the processing load on the HCP system. HCP reports CPU statistics in these graphs:
CPU Usage
This graph shows both the percent of CPU capacity used by the operating system kernel (OS in the graph legend) and the percent of CPU capacity used by HCP processes (HCP in the graph legend).
CPU IO Wait
This graph shows the percent of CPU capacity spent waiting to access logical volumes that are in use by other processes.
These two statistics together equal the total processing load on the system.
If CPU usage is consistently high across all nodes and system performance is degraded, the namespace application workload may be too heavy for the system to handle efficiently. In this case, you may need to add nodes to the system or upgrade the existing nodes to nodes with greater CPU capacity.
If CPU usage is high on a recurring basis, check the system log to see whether the high CPU usage correlates with recurring events such as services running. If the high usage correlates with services running, you may want to change the service schedule.
High CPU usage on only a small number of nodes may mean that applications are repeatedly using the same IP addresses to access the system. In this case, you may want to suggest to tenant administrators that their applications use DNS or some other mechanism to help balance the workload across all the nodes in the system.
Consistently high CPU IO wait with low CPU usage may mean that HCP cannot access the system storage fast enough to keep up with application demand. In this case, you may need to add storage to the system so that attempts to access storage are spread across a larger number of logical volumes.
In an HCP SAIN system with spindown storage, high IO wait on nodes with logical volumes that can be spun down may indicate that these volumes are being spun up frequently. Check the LUN Utilization graph for the load on spindown volumes. If the load is high, you may want to redefine service plans to have a longer wait time before objects are moved to spindown storage. To see which logical volumes are spindown volumes, check the Hardware page in the System Management Console.
A brief period of high CPU IO wait that corresponds to increased workload does not necessarily indicate a problem.