HP LeftHand P4000 SAN Solutions - Manual (Page 223)

Table 72 Performance Monitor statistics (continued)
Storage system Volume or
snapshot Cluster Definition Statistic
X - - Percent of bidirectional network capacity used
on this network interface on this storage system
for the sample interval. Network
Utilization
- X X Number of outstanding read requests. Queue Depth
Reads
X X X Number of outstanding read and write requests. Queue Depth
Total
- X X Number of outstanding write requests. Queue Depth
Writes
X - - Average time, in milliseconds, for the RAID
controller to service read and write requests. Storage System
Total Latency
X X X Average read bytes per second for the sample
interval. Throughput
Reads
X X X Average read and write bytes per second for the
sample interval. Throughput Total
X X X Average write bytes per second for the sample
interval. Throughput
Writes
Performance monitoring and analysis concepts
The following list describes general concepts related to performance monitoring and analysis.
Workloads
A workload defines a specific characterization of disk activity. These characteristics include access
type, access size, access pattern, and queue depth. Application and system workloads can be
analyzed, then described using these characteristics. Given these workload characterizations, test
tools like iometer can be used to simulate these workloads.
Access type
Disk accesses are either read or write operations. In the absence of disk or controller caching,
reads and writes operate at the same speed.
Access size
The size of a read or write operation. As this size increases, throughput usually increases because
a disk access consists of a seek and a data transfer. With more data to transfer, the relative cost
of the seek decreases. Some applications allow tuning the size of read and write buffers, but there
are practical limits to this.
Access pattern
Disk accesses can be sequential or random. In general, sequential accesses are faster than random
accesses, because every random access usually requires a disk seek.
Queue depth
Queue depth is a measure of concurrency. A queue depth of one (q=1) is called serial. In serial
accesses, disk operations are issued one after another with only one outstanding request at any
given time. In general, throughput increases with queue depth. Usually, only database applications
allow the tuning of queue depth.
Configuring and using the Performance Monitor and Adaptive Optimization 223