Sampling link utilization for typical storage transactions may not identify periods of peak activity over time. A database application, for example, may require less than 20MBps per server but must contend with additional traffic in the SAN during periodic tape backups. In addition to performance monitoring utilities in SAN switches or network analyzers, then, it is useful to sample storage traffic over time to present a more comprehensive view of total utilization.
Providing performance statistics over extended periods provides useful information for both current utilization and network capacity planning. Vendors such as Concord Communications have been providing capacity-planning capability for traditional IP networks and more recently have introduced platforms for monitoring storage networks. Concord's eHealth-SAN, for example, periodically polls SAN adapters, switches, and interfaces to solicit point-in-time port utilization statistics, as depicted in Figure 9-5. This can be done by querying the SNMP MIB provided by the hardware vendor or by using proprietary APIs. Sampled performance numbers must then be correlated to data paths through the SAN and periodically rolled up in a database. The output can be displayed as bar or pie chart graphs, showing periods of peak utilization, potential bottlenecks, and utilization trends over a period of days, weeks, or months.
Figure 9-5. Concord eHealth-SAN historical SAN performance monitoring. Reprinted with permission from Concord Communications
Performance aggregation is also useful for monitoring service level agreements (SLAs), particularly for service providers whose clients need verification that service levels are met. Historical reporting both validates current service level compliance and provides data that may justify increasing available bandwidth for specific applications.