The FC-SW-2 standard for storage area networks (SANs) uses an algorithm called Fabric Shortest Path First (FSPF) to select the most efficient path to follow when moving frames through a SAN fabric. While this works very well for small and medium multi-switch SAN fabrics, there are some issues with very large SANs.
FSPF maintains a link state database (sometimes called a topology database) containing all the link states in the SAN. It associates a cost with every link on every switch in the SAN fabric and chooses the lowest-cost path through the fabric from source to destination. Usually the cost is proportional to the number of hops and the total cost of a path is the sum of the costs of the individual links.
Although FSPF is efficient, it is not without its costs. To make FSPF work each switching node has to announce link-state changes to its neighbor switches when setting up a path. Each switch is also involved in recomputing the shortest route through the fabric. As McData points out in a white paper titled "Performance considerations for large-scale SANs": "Thus, if there are N switches in a fabric, N2 order of announcements and computations are performed. Optimization applicable to less densely meshed networks results in (Nlog2L) route tests, where N is the total number of switching nodes and L is the total number of links."
This adds up quickly. The same paper points out that a SAN fabric with 100 switching nodes can have between 5000 to 10,000 route announcement messages and may perform more than a million comparisons to compute the new routing tables. All this messaging and calculation can impact performance and manageability in large SANs.
One solution is to reduce the possible routes through the fabric by techniques such as a core-edge topology where most transactions involve a source and destination on a single switch. In general the name of the game is to reduce the use of inter-switch links (ISLs) by careful planning.
IBM has a note explaining FSPF at:
publib-b.boulder.ibm.com/Redbooks.nsf/RedbookAbstracts/tips0034.html
"Performance Considerations For Large-Scale SANS" is available from McData at:
www.mcdata.com/downloads/whitepapers/LargeSANS2002.pdf.
Rick Cook has been writing about mass storage since the days when the term meant an 80K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For the last twenty years he has been a freelance writer specializing in storage and other computer issues.
