04-05-2011 05:51 PM - edited 03-01-2019 09:52 AM
What is the difference and hence the impact of creating a global VSAN (VSANs being visible to both fabric-interconnect) as opposed to creating a VSAN specific to FIC-A and FIC-B?
I create Global and then assign one to vHBA-a on to FIC-A and the other to vHBA-b on to FIC-B.
Also, with FC switching (UCS 1.4) does this(defining globally) get worse?
Solved! Go to Solution.
04-05-2011 07:01 PM
SKumar,
This comes down to design. Most people like to keep their SAN fabrics completely separated. Unlike in switching where uplinks are criss-crossed every way imaginable for redundancy, in the FC world, it's all about separating your fabrics. For this you would normally connect one FI to a single upstream SAN switch (Fabric-A), and the other FI to their other (Fabric-B). You would create the VSANs on the Fabric Interconnect rather than global as you don't need both VSANs available on both FI's.
An alternate approach is to create them globally, but in this design you would need either a link from each FI to EACH SAN switch or utilize VSAN trunking (assuming both VSANs exist on both Upstream SAN switches). The complication here, is when defining your vHBAs you must assign them to a VSAN on either A or B. This is a hard pinning and offers no hot redundancy benefits. There's no failover with FC like there is for Ethernet (aka Fabric Failover). All FC redundancy is handled at the OS level by multipathing drivers.
The first design keeps your FC access much simpler and deterministic. You know that all vHBAs connected to Fabric Interconnect-A go to VSAN X, and all vHBAs connected to Fabric Interconnect-B go to VSAN Y. This is assuming you're only using two VSANs for your environment.
The only time you'll need FC Switching mode is if you plan on using direct attach FC/FCoE arrays with your Interconnects. There's no increase in complexity here in regards to VSAN design, same best practices apply.
Regards,
Robert
04-05-2011 07:01 PM
SKumar,
This comes down to design. Most people like to keep their SAN fabrics completely separated. Unlike in switching where uplinks are criss-crossed every way imaginable for redundancy, in the FC world, it's all about separating your fabrics. For this you would normally connect one FI to a single upstream SAN switch (Fabric-A), and the other FI to their other (Fabric-B). You would create the VSANs on the Fabric Interconnect rather than global as you don't need both VSANs available on both FI's.
An alternate approach is to create them globally, but in this design you would need either a link from each FI to EACH SAN switch or utilize VSAN trunking (assuming both VSANs exist on both Upstream SAN switches). The complication here, is when defining your vHBAs you must assign them to a VSAN on either A or B. This is a hard pinning and offers no hot redundancy benefits. There's no failover with FC like there is for Ethernet (aka Fabric Failover). All FC redundancy is handled at the OS level by multipathing drivers.
The first design keeps your FC access much simpler and deterministic. You know that all vHBAs connected to Fabric Interconnect-A go to VSAN X, and all vHBAs connected to Fabric Interconnect-B go to VSAN Y. This is assuming you're only using two VSANs for your environment.
The only time you'll need FC Switching mode is if you plan on using direct attach FC/FCoE arrays with your Interconnects. There's no increase in complexity here in regards to VSAN design, same best practices apply.
Regards,
Robert
04-06-2011 02:41 PM
Thanks Robert, appreciate your response.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide