dynamox
7 Thorium

VMAX3 and Host Connectivity

Jump to solution

Since VMAX3 ports do not share CPU how does everyone create their port groups, also how do you distribute ports between multiple engines and SAN fabrics.  In this example it seems Director 1 on both engines is connected to Fabric A and ports on Director 2 all connect to Fabric B. Looks much simpler what we used to do on VMAX 20K but is this per best practices ?  Shouldn't we put some ports from Director 1 and Director 2 on Fabric A and same on Fabric B ?  Thoughts ?

Thanks

1-21-2016 12-39-27 AM.bmp

Labels (1)
0 Kudos
1 Solution

Accepted Solutions
Allen Ward
5 Tungsten

Re: VMAX3 and Host Connectivity

Jump to solution

Yes, that's exactly how I would (and do) do it. It gives you spread across fabrics, engines and directors to cover off both redundancy and traffic load-balancing. And it's easy to remember!

View solution in original post

0 Kudos
6 Replies
KW160
3 Argentum

Re: VMAX3 and Host Connectivity

Jump to solution

There are two main ways I am seeing PGs and fabric distribution designed:

Method 1 (even/odd ports):

Fabric A: 1D:4, 1D:6, 1D:8, 1D:10, 1D:24, 1D:26, 1D:28, 1D:30, 2D:4, 2D:6, 2D:8, 2D:10, 2D:24, 2D:26, 2D:28, 2D:30

Fabric B: 1D:5, 1D:7, 1D:9, 1D:11, 1D:25, 1D:27, 1D:29, 1D:31, 2D:5, 2D:7, 2D:9, 2D:11, 2D:25, 2D:27, 2D:29, 2D:31

Example PG: 1D:4, 1D:25, 2D:5, 2D:24

Main advantage: In case of a full fabric failure, the hosts will still have access to the full FE CPU pools on all directors.

Method 2 (even/odd directors):

Fabric A: 1D:4, 1D:5, 1D:6, 1D:7, 1D:8, 1D:9, 1D:10, 1D:11, 1D:24, 1D:25, 1D:26, 1D:27, 1D:28, 1D:29, 1D:30, 1D:31

Fabric B: 2D:4, 2D:5, 2D:6, 2D:7, 2D:8, 2D:9, 2D:10, 2D:11, 2D:24, 2D:25, 2D:26, 2D:27, 2D:28, 2D:29, 2D:30, 2D:31

Example PG: 1D:4, 1D:24, 2D:4, 2D:24

Main advantage: You cannot accidentally zone both host HBAs to only a single director.

In my opinion both methods are acceptable. I'm seeing a mix of both.

Allen Ward
5 Tungsten

Re: VMAX3 and Host Connectivity

Jump to solution

So, I know we already talked about this but somehow my "follow" settings got messed up and I never got notified of the question here. I'll chime in anyway for visibility to other users.

We have a 4 engine 400K deployed and we looked at a few different ways of managing the FA connectivity. We settled on connecting alternate directors in each engine to different fabrics. Since we have D0 & D1 redundant fabrics I decided to put all the odd numbers directors through the D1 fabric and the even numbered directors through the D0 fabric (I guess technically "0" is neither odd nor even, but in my mind it is more even than odd).

All our Port Groups are 4 ports with 2 from each fabric. So we pick an FA port and group that port number across directors 01, 02, 03, & 04. The next Port Group uses the same port number across directors 05, 06, 07, & 08. The next port group continues with the same pattern.

dynamox
7 Thorium

Re: VMAX3 and Host Connectivity

Jump to solution

thank you guys, trying to keep it as simple as possible (thank you for the feedback Allen) . For a two engine VMAX3 does this look "consistent/redundant/efficient/favorite adjective here"  . These 4 ports would become my first port group. Comments ?

2016-01-25_13-14-26.png

Thank you

KW160
3 Argentum

Re: VMAX3 and Host Connectivity

Jump to solution

Yes, that fits the method #2 I outlined above.

Allen Ward
5 Tungsten

Re: VMAX3 and Host Connectivity

Jump to solution

Yes, that's exactly how I would (and do) do it. It gives you spread across fabrics, engines and directors to cover off both redundancy and traffic load-balancing. And it's easy to remember!

View solution in original post

0 Kudos
dynamox
7 Thorium

Re: VMAX3 and Host Connectivity

Jump to solution

Thank you guys

0 Kudos