Evaluating the requirements for the storage network backbone

Computer Technology Review, June, 2004 by Eric Blonda

Flexibility in a storage network switch allows the SAN architect the ability to host multiple network services (rate, protocol, etc.) in the most appropriate increment, under the same managed system. By limiting the number of ports on a single hardware interface, users can also contain port failures, limiting downtime to adjacent ports when swapping out the failed port. Investment protection comes not only from the ability to support additional interface technologies, but integrated services as well.

Visibility

In today's enterprise, SAN performance is typically expressed through a number of metrics that are collected directly from the SAN devices, helping administrators diagnose the overall effectiveness of the infrastructure in supporting business applications. SAN device performance metrics may come from the host, such as CPU utilization, from the storage array, such as disk-seek times or from the network, such as port utilization statistics. Traditionally, much more administrator time has been spent evaluating the performance of the host and array, where much more tangible metrics have been available.

The storage network, however, has yet to be considered either a suspect for degradation, or as an opportunity for overall improvement. This is largely due to the limited amount of visibility into storage network performance, and the false impression that a network has latent bandwidth when it is actually congested and causing application-level performance degradation. As the storage network backbone takes shape, understanding network performance, particularly when and where congestion is occurring, is one of the most effective ways to manage the overall performance of the SAN.

Understanding Congestion

When the purpose of a network is to allow for communication among devices that share common resources, every network presents the opportunity for congestion. This scenario is only magnified in storage networks, where there are typically many hosts, or initiators, communicating with relatively few storage ports, or targets. Many are under the false expectation that a network built with switches that are "non-blocking" (with the ability to service all ports at line rate) can never be at fault when overall SAN performance falters and application performance degrades. However, no amount of raw switch performance can overcome the fundamental bottleneck that arises from two or more initiators requiring access to the same target.

In response to congestion, the Fibre Channel protocol uses a very carefully architected system of fabric device communications and creditbased transmissions in order to ensure network stability and traffic integrity. When presented with the potential for congestion, such as more traffic from inbound ports than a destination port can handle, Fibre Channel switches do not drop frames--unlike Ethernet/IP networks. Instead, "backpressure" is used to squelch all inbound traffic bound for the same destination to a level where the target can accommodate all requests. Translated to the device level, typical switches equally share storage mapped to a storage port across all servers requiring access. Even conventional director-class switches, with Virtual Outbound Queue (VOQ)-based architectures, manage all traffic bound for the same out-bound port from the same queue.