What you will learn: To understand SAN performance as well as what is normal SAN behavior, you must have a baseline to compare against.
To understand SAN performance, and to understand what is normal (or abnormal) SAN behavior, you must have a picture or baseline to compare against. How much detail on SAN performance you need will also determine which SAN performance tools you use.
You should establish a baseline of what is normal or typical performance for your storage environment as well as the applications running in it. Knowing the typical IOP rates and throughput rates for the storage devices, as well as the common error rates, average queue depths and response times, will allow you to make quick comparisons if the SAN develops performance issues.
Monitoring SAN performance involves primarily three metrics: IOPS (I/O operations/second), bandwidth and latency.
Fibre channel adapters are used to connect storage devices to a SAN. A fibre adapter on a server (either virtual or physical) can include PCI, PCIx or PCIe busses that are attached to Fibre Channel, Ethernet, InfiniBand or SAS ports. You can use these adapters to look at IOPS, bytes sent and received, errors and queues.
if you are using MAN and WAN links to support remote backup and data replication or archiving applications, you wil need to factor IP-based bandwidth services (along with SONET/SDH and WDM items) into the SAN performance equation. If this is the case, you'll need to do this with an eye not only on bandwidth, but on latency and retransmission errors as well.
Other metrics for measuring SAN performance include:
- bandwidth (Gbps or Mbps) on a storage system, switch, LUN, file system or per port;
- transactions, frames, requests, words or packets; and
- I/O size and type (read or write), used to determine or report on random or sequential activity.
Measuring IOPS and bandwidth can tell you how much work or activity is taking place on the SAN. Measuring latency tells you how effectively the SAN is doing its work, as well as whether the SAN is meeting its service objectives. By using switches and server adapters to view error rates, you can pinpoint the source of SAN performance problems. Error rates can include loss of signal or synchronization, excessive database loss, retransmissions, link failure or invalid CRCs.
SAN performance monitoring tools
You can obtain SAN performance metrics from tools available from your SAN manufacturer, third-party storage management vendors or even operating system-based utilities. Tools for collecting, reporting, analyzing or simulating SAN performance for servers, operating systems, adapters, switches and storage systems are available from top-tier companies such as Brocade, Cisco, EMC, Emulex, Qlogic and Symantec, but also smaller companies such as Agilent Technologies; Anue Systems Inc.; Ethereal Inc.; Finisar Corp.; HyperIO LLC; LeCroy Corp.; Network General; NetQos Inc.; Network Instruments LLC; Onaro Inc.; and Shunra Software.
Green storage environments
Power and cooling or energy consumption are also becoming factors in the SAN performance equation. Many SAN administrators are now looking at IOPS/kWh, bandwidth/kWh and capacity/kWh to avoid over-consolidation and subsequent performance bottlenecks. Other energy metrics for a green storage environment include transactions per watt; Mbps/watt; SPC, TPC or Exchange users per watt; and backed-up data per watt.
About the author: Greg Schulz is founder and senior analyst with the IT infrastructure analyst and consulting firm StorageIO Group. Greg is also the author of Resilient Storage Networks (Elsevier) and a contributor to Storage magazine and other TechTarget venues.