Optimizing SAN performance

SAN planning and performance management tools are critical when implementing a high-performance storage network.


Storage area networks (SANs) can significantly improve the performance of networked data movement. These performance gains are achieved through a combination of network architecture and the implementation of high-performance technologies such as Fibre Channel. A SAN centralizes storage and eliminates the need to deploy LAN traffic to move data between storage devices. Storage networks also allow multiple servers to access the storage simultaneously, providing exceptional potential for scalability and throughput.

Software vendors have responded to the opportunities for performance gains in SANs by delivering solutions to leverage the unique aspects of centralized storage. For example, a variety of software applications enable "LAN-free" backups, where little or no LAN traffic is generated during data archival operations.

Despite the significant opportunities for enhanced performance through SANs, there are a number of ways in which performance can be degraded. If a SAN is deployed with few data paths to a storage device relative to the traffic that the device receives and/or generates, performance may suffer. A data path that includes devices that are mismatched-in terms of throughput capability-may degrade performance to the level of the lowest-performing device. Even if the data paths and devices are implemented to optimize the bandwidth of each path and the device capabilities within the paths, the constant archival of new data and deployment of new applications and devices may change the usage patterns of the data paths. This can result in some paths that have excess bandwidth while others do not have enough. For this reason, SAN performance must be planned, monitored, and tuned.

Planning for SAN performance

SAN performance starts with the selection of the right equipment. Switches, routers, hubs, and bridges should provide as near-equivalent throughput as possible within each data path. Inserting a 1Gbps switch between two 2Gbps switches can lower the data throughput to 1Gbps.

If there are lesser-capacity components that must be implemented in the SAN, these can be relegated to less-critical data paths or used as redundant fail-over links across higher-bandwidth data paths.

If the storage administrator's budget does not allow for the purchase of the highest-capacity equipment, then a trade-off analysis can be performed to identify the right mix of bandwidth and availability risk versus device performance. It may be better, in some cases, to have additional bandwidth and redundancy available through multiple devices rather than one high-performance device. This might be true for certain backup applications, where additional links to ensure the backup occurs could be more important than performing individual block transfers quickly. Website data centers might also prefer greater bandwidth to individual device performance to a certain degree. This trade-off is particular to a given data center's situation and preferences.

SAN management software provides tools for quickly identifying and correcting performance issues. Visual indicators help administrators isolate trouble areas to specific components or segments of the SAN.
Click here to enlarge image

SAN management applications that include a planning tool can help with these assessments. A planning tool can provide a quick means to create topology (see diagram) alternatives so that administrators can better visualize and identify the trade-offs between candidate installations. Planning tools may also provide theoretical information and advice on how to set up an optimal SAN for a particular environment.

Monitoring SAN performance

Monitoring storage network performance can start as soon as the first piece of hardware is installed. Installing a SAN management application first can aid the installation process and ensure all the devices are operating correctly. If the management application includes a SAN planning tool, it may show discrepancies between the plan and the actual implementation. It can also help by monitoring the performance as devices are added to ensure the implementation performs as expected. The administrator may also want to try some theoretical trade-offs identified in the planning stage.

Once the SAN is installed, its performance must continue to be monitored. New data, new storage devices, ailing devices, and revised data zones (data segregated for specific hosts) can all change the behavior of the SAN such that it must be re-configured to obtain optimal performance. Some performance monitoring applications include visual indicators to help administrators quickly isolate trouble spots. Most SAN management applications allow users to set thresholds for performance so the administrator will be notified if performance is degraded below a certain level, absolving the administrator from having to constantly watch the SAN.

Tuning SAN performance

There are several types of monitoring in which administrators may be interested. For example, individual device performance, single link (direct path between two devices) performance, and full data-path performance can all affect access time to data. During normal usage, the user may want to set the performance thresholds and then forget about the SAN performance until a threshold is breached and the administrator is alerted. Once alerted to a problem, the administrator can then troubleshoot and tune the performance of the SAN.

Initial tuning efforts may focus on the area of degraded performance. If the trouble spot is on a section of a data path (link) that includes additional links between the storage subsystem and the server, the administrator may look at the performance of each device on either side of the ailing link to see if one of the devices has degraded performance. Most management applications allow users to look at individual device performance, often measured in megabytes per second.

If the devices are being used at full bandwidth, the data paths may need to be re-routed, or additional bandwidth capacity may need to be added.

If bandwidth usage is well within limits-and the upstream device (source) is performing adequately but the downstream device (destination) is not-then a problem may exist with that device or the connection between the devices. If a device, connector, or cable appears to be the cause of the performance issue, then correcting the problem may be as simple as replacing the ailing unit.

Administrators may also want to regularly enact proactive performance tuning. Some performance-monitoring applications provide the capability to test latency by sending data "bullets" from an initiating host to a target device, allowing the administrator to fine-tune SAN performance.

Optimizing SAN performance requires careful planning of the storage network implementation. A SAN management application that includes a planning tool can take some of the guesswork out of this task and greatly reduce the time involved in implementing and tuning the SAN. Once implemented, performance-monitoring tools can alert administrators to issues and help identify potential trouble spots. Using integrated planning and performance, SAN management software provides administrators with the ability to quickly and proactively evolve the SAN.

Robert Wright is president and COO at SANavigator Inc. (www.SANavigator.com) in San Jose, CA.

This article was originally published on April 01, 2001