In design and implementation, I think that application I/O pause or hang is very important. Even though a redundancy link is available between a SAN device and a server or storage, other SAN devices have failed. I tested two SAN switches connected to each other through several ISLs to verify how much time is required for failover when I pull out one ISL between two switches. The result was 5-20 seconds. In some application, the result would be a critical issue. Could you tell me why time is required and is there is any solution to reduce those times through the changing of setup-value in a SAN switch?

    Requires Free Membership to View

    Login

The newer switches provide a mechanism for ISL "trunking," which should eliminate the problem you are seeing. You may be able to just upgrade your current switch firmware to get this functionality.

The failover time for host paths is sometimes dependent on the HBA driver settings. Driver settings can override the path failover settings for timeout information. Most drivers default to somewhere between 60 and 180 seconds. You can change that (depending on vendor) down to less than 5 seconds in many cases and still get reliable failover. (Again, this depends on the vendor. Check your documentation to be sure.)

It also may be useful to limit how many HBAs are configured in a particular switch zone. In each zone set, you can create an access zone between each host HBA and the storage it is supposed to see. You can do this for each HBA in the fabric for the zone set. Doing so will limit error traffic to only that specific host HBA and the storage port it uses to access its LUNs.

Chris

Editor's note: Do you agree with this expert's response? If you have more to share, post it in our Storage Networking discussion forum.

This was first published in February 2002

Join the conversationComment

Share
Comments

    Results

    Contribute to the conversation

    All fields are required. Comments will appear at the bottom of the article.
    Back to top