PowerStore: General Performance Terminology

• CPU Utilization: The percentage of CPU Utilization - This can be either CPU-wide or reported for each core, or sets of cores assigned to host specific functions, such as Data Path, Management, SDNAS, and so on
• IOPS: Number of read and write requests that the system serviced within a specified period of time. The IOPS metrics include the average total IOPS, average read IOPS, and average write IOPS.
• Bandwidth: Our GUIs and tools truncate to this from Bandwidth Utilization, which is what these data are reporting.  The rate at which data that is being moved across/by a component of the system, or the entire system itself. The bandwidth metrics include; average total bandwidth, average read bandwidth, and average write bandwidth, and may contain average read and write bandwidth for mirror operation. Bandwidth Utilization is calculated by multiplying the IOPS by the IO size.
• Latency: How fast the system, or component, responds to read and write requests on average during a sampling interval. The latency metrics include; average total latency, average read latency, average write latency, and may contain average mirror latency to reads and writes.

• IO Size: Average size of read and write IO operations within a specified period of time. The IO size metrics include the average total size, average read size, and average write size.
• Queue Depth: Number of pending IO requests.
• Random access I/O: I/O Which is addressed to regions on the Volume which are not for sequential Logical Block Addresses (LBAs).
• Sequential access I/O: I/O which is performed in LBA sequence.
• Single Threaded Operation: This indicates that the program performing the I/Os is sending the I/O request as a single sequence of IOs and can result is inefficient use of the i/o resources. Thus, once a single threaded I/O stream has filled up a Queue in the data path it must wait for an IO to complete before the next IO is the sequence can be queued for processing.
• Multi-Threaded Operation: This indicates that the program performing the I/Os is sending the I/O requests as multiple sequences of I/Os, While multi threading requires coordination of the I/Os if makes more efficient use of the available resources of the system. By using multiple threads, and dividing the work across multiple threads, the overall throughput for the process is much greater.

Port network performance metrics:

• Packets: The number of packets that are received and transmitted over the port (Ethernet Ports)
• Bytes: The number of bytes that are received and transmitted over the port (Ethernet Ports)
• Loss:  Loss of signal and sync count per second (Fibre Channel Ports)
• Invalid count: Invalid crc and transmit word count per second (Fibre Channel Ports)

Replication:

• Replication Remaining Data: The amount of data (MB) left to be replicated to the remote PowerStore
• Replication Bandwidth: The replication rate (MB/s)
• Replication Transfer Time: The amount of time (seconds) required for copying the data
• Session Bandwidth: The Sync session rate (MB/s)

Offload Performance Metrics:

• IOPs: I/O commands per second
• Bandwidth: Bytes per second
• Latency: Time in seconds to run a command