This is the most important metrics from SAN and storage perspective in my experience. Also, some of this metrics can help you with solid understanding of problems and organize any type of reports.
Storage perfomance metrics.
Here we start from hardware audit — gathering information about:
- Type and model of storage in vendor product line (block, file, object, unified) — this information will help you with common knowledge of customer infrastruscture. The first look on this metrics — and you can say about EOL, EOSL, OES, high-end, midrange, something about internal components (CPU, FC) etc. Here we can save information about current hardware failures;
- Number of installed hard drives — at first sight it’s useless, but it`s not. In my practice customer may have storage with total number of drives near maximum of possible, so it may be very good point to think about new storage instead trying to get more IOPS, espessially if existing storage has 6-12 mounth to EOS.;
- Number of cache memory — only one point concerning possible upgrade (more size, more qty);
- Number and purpose of installed licenses (for usable space, any features like replications, snapshots and so on);
- Number of front-end and back-end ports — and all about its parameters (speed, type, upgrade possibility).
Depending on hardware audit we can come to some conclusions (exclude not interested storage, for example) and form some of logical metrics:
- Input/output operations per second — IOPS;
- Latency or response time;
- Size of LUNs (as a result you can calculate queue depth);
- CPU utilization;
- Cache utilization (cache usage rate, cache write pending, cache read/write hit);
- Throughput in Mbps (using this formula — https://editorsean.com/articles/convert-mbps-iops/)
- As a result of metrics above you can calculate block size and workload profile (50% read/50% write) but only for monitoring period. Warning!! Be carefull with storage which have a mixed type of workload.
SAN perfomance metrics.
Next step is gathering information about SAN hardware — directors, top of rack switches, blade chassis switches:
- Like in storage hardware audit part — type and model of hardware throught the «filter» of EOS, EOSL, EOL etc.;
- Total number of modules (in case of directors and backbones), ports, power supplies, blowing direction etc.;
- Number of trunk, ISL, IFL links;
- Number of connected devices (hosts, storages, tapes).
Depending on hardware audit we can come to some conclusions and form some of logical metrics:
- Port utilization (in general, from separate device point of view)
- Fan-in/Fan-out ratio;
- Buffer-to-buffer credits on ISL links;
- Oversubscription ratio;
- Correlation with best practices regarding zoning, topology and vendors recommendations;