Re: Needing some mutliprotocol design advice!

highlandclinic1 wrote:

 

OK guys, I have created a drawing based on how I understand your posts.  If you would be so kind, please look it over and see if you can find any errors, and also see further questions below...

 

 

OK here are a few extra questions if I could get you to weigh in:

1. Given my setup, would it be      possible to move the Vmotion kernel port to any part of the SAN, or      is that even advisable?

if you have only 6 pnics, then your setup is ok and if you upgrade to esxi 5 you can use multinic vmotion and this will very much improve performance. it is not good to mix the vmotion with nfs/iscsi bandwidth.

 

1. I've read that I should always      configure Failover Detection with Beacon Probing and always Notify      the Switches, true?

if you have more than 3 pnics in a vswitch then only the beacon probing work correctly. more over is you have Top of rack swithch in between esx and core switch then you can use, but more pnics you need in the vswitch in order to work. and if for ip hash is should not be used, vlan flapping issue will occur

 

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1012819

http://blogs.vmware.com/vsphere/2008/12/using-beaconing-to-detect-link-failures-or-beaconing-demystified.html

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1005577

 

 

1. In this setup, can you see any      area where VLAN tagging will be necessaryin the Vswitches?  Keep in      mind that I'm planning to route both NFS and iSCSI traffic over a LAG      in the physical switch to the NetApp.  Does that matter for the      hosts point of view?

it is better to tag iscsi/nfs from a security point of view, when using LAG check the below links for using ip hash,

 

http://communities.vmware.com/message/2173828#2173828

 

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1007371

http://frankdenneman.nl/networking/ip-hash-versus-lbt/

 

 

 

 

1. How would MSLA effect this design.       It seems to me that this design assumes that I don't have physical      switches capable of MSLA. True?  If I had MSLA, would I route based      on something other than originating port, eg IP hash?

with your design diagram, you cannot use IP hash beacause you dont have MSLA, any way if you have multiple target in the storage any way you will get good throughput, as mentioned by Duncan use correct MTU,

 

if you have MSLA you can use IP hash but with ip hash... the link failover detection you need use correctly.

 

 

in  short, almost you made correct diagram, use correct MTU, use port Id teaming policy, configure one to one mapping for nfs/iscsi... your diagram is ready...thats it

 

Thanks guys, I feel like I'm getting pretty close to something usable here...