To evaluate performance, track the percentage of utilization of active resources and throughput.
You should baseline your I/O profile when all systems are operating in a normal state. The values shown in the Normal Value column are planning estimates.
|
Metric Name Analyzer |
Name in detail view |
Normal Value (in milliseconds) |
Performance risk value (in milliseconds) |
| Read response time | Average disk reads/second | 1 to 10 | Greater than 10 |
| Write response time | Average disk reads/second | 1 to 3 | Greater than 3 |
Storage-area network considerations
Storage-area network (SAN) components are also part of the end-to-end data path analysis. Areas of concerns that need to be addressed could include inter-switch links (ISLs) and latency issues due to throughput. The switch configuration might not be in line with the desired infrastructure design, so we need to maintain the correct firmware and make sure we do not have any issues with SFP.
For example, a 48-port board has 2 ASICs sharing ports:- If communication is between ports on the same ASIC port, then you see a 700-nanosecond latency.
- If communication is between ports of different ASICs, or between different boards, traffic goes through the back panel, through core communication boards, and then at the end of the transfer goes through three ASICs which could take 2 microseconds.
Regarding throughput, each port can transfer the maximum capability of the ASIC (i.e., 16 Gbps with Gen5, 32 Gbps with newer switch architectures. Most switches can sustain the number of IOPS necessary to offer maximum throughput with a normal frame size of 2048 bytes. This might not be true with smaller frames.