Performance analytics statistics reference

This topic describes the various performance statistics that are available for the Delphix Engine and how they can be used to analyze and improve performance.

The Delphix Engine is shipped with a default set of statistics that are collected on Delphix Engine virtual appliance, as listed below. The statistics are stored for up to 30 days for historical analysis.

Statistic	Description
CPU Utilization	Total CPU utilization for all CPUs. This statistic includes both kernel and user time.
Network Throughput	Measures throughput in bytes and packets, broken down by sent vs. received data and by the network interface. Each network interface shows four graphed lines: bytes sent, bytes received, packets sent, and packets received. To help easily correlate bytes and packets, the same color is used for both bytes and packet values.
Disk IO	Measures a number of IO operations, and the latencies and throughput of the underlying storage layer. The statistic is represented by the graphs - a column chart for IO operations, a heat map for latency distribution, and a line chart for throughput. IO operations are grouped by reads and write. A shaded rectangle on a latency heat map represents an IO operation (read or write) which falls within a particular time range (bucket). The shading of rectangles depends on the number of IO operations that fall within a particular bucket - the higher the count the darker the shading.
NFS	Measures a number of IO operations and the latencies and throughput of the NFS server layer in the Delphix Engine. Its graphical representation is similar to the Disk IO graph. It is useful to diagnose the performance of dSources and VDBs that use NFS mounts (Oracle, PostgreSQL).
iSCSI	Measures the number of IO operations, and the latencies and throughput, of the iSCSI server layer in the Delphix Engine. Its graphical representation is similar to the Disk IO graph. It is useful to diagnose the performance of Microsoft SQL Server dSources and VDBs.