- Choose Directors over switches. Director-class switches are specifically designed to be more resilient -- typically 99.999% or greater. Although more expensive, the switches will increase the uptime of your fabrics. A SAN will only be as resilient as its weakest link so consider recommending using directors for both core and edge in a standard core/edge design.
- Create SAN dual fabrics. Despite running high availability switches there is always the risk of a failure. Running dual fabrics with two host bus adapters (HBA) per server and two connections to the storage will improve availability considerably should a critical switch fail. Dual fabrics increase cost at the expense of availability but also offer other opportunities, such as switch firmware migration without downtime (upgrading each fabric independently).
- Keep up to date on SAN firmware. New bugs are identified continually so it is essential to keep HBA, switch and storage firmware up to date and in line with vendor support matrices. SANs always grow and as new servers are added the risk of an issue is also increased. Keeping up to date on firmware is a must. VARs may choose to include firmware maintenance as part of a value add proposition.
- Reduce SAN size. As SANs grow it may be possible to identify discrete elements of the infrastructure which can justify their own SAN (for example, development servers or email/messaging). VARs can recommend opportunities to subdivide SANs into separate fabrics which can be treated separately from a management perspective. Certainly, removing development servers from a production SAN will remove risk and may improve productivity for development users who don't have to suffer production-level change control. Don't forget that SAN fabrics do not have to be isolated and routing products can be used to re-join SAN "islands."
This was first published in April 2007