VMware Cloud Community
m1kkel84
Contributor
Contributor

FC Switch zoning or 2 FC switches- performance gain?

Hello.

I have 4 esx servers - currently 1 hba in each server. 8 GB FC, and 8 GB FC switch aswell.

Lets say that i installed a total of 2 hba's in each server, that would give me load balancing and fault tolerancy. Lets say i did that, and used only one FC switch, but created 2 hard zones in the switch.

Would i get more performance if i used 2 switches instead of one zoned switch? (i know i would loose fault tolerancy, but i can always add another switch when i get some money freed up)

Thanks in advance!

Best regards MIkkel

Tags (4)
0 Kudos
5 Replies
a_p_
Leadership
Leadership

Regarding high availability, a second HBA as well as a second switch is definitely worth thinking about.

Regarding performance, it depends on your storage system and whether it supports the Round-Robin path policy. I assume the switch is powerful enough to handle multiple paths, so IMO there will be no real performance difference in using one or two switches. However, the question, whether you will need more than 8GBit/s or not is something you need to find out yourself by evaluating the workload.

Regarding zonig, I recommend setting up a single initiator zoning in any case (1 host HBA zoned with all storag HBAs).

André

0 Kudos
AndreTheGiant
Immortal
Immortal

Also with a single switch you can have multiple-paths and maybe more performance (but of course a single point of failure).

I say "maybe" because it depends by your switch (is the internal thoughput enough?) and by your LUN layout.

For example, with more LUNs you can have each LUN on different path and (in this way) have more performance.

Andre

Andrew | http://about.me/amauro | http://vinfrastructure.it/ | @Andrea_Mauro
0 Kudos
m1kkel84
Contributor
Contributor

My storage system is a HP P2000G3 - and yes, it supports round robin path policy.

The switch is a HP Storageworks 16/24 Full Fabric Enabled San Switch. It's a rebranded brocade switch ofcourse. So yes, i assume it is powerful enough?

Well i currently only have 1 single portet jba in each server - path policy set for fixed. 1 12disk raid 5, 1 12disk raid 10. SAS 10k 300GB Disks. I have 5 Luns's on the raid 5, and 3 Luns on the raid 10.

I moved all the virtual servers (20) to this new storage, from my old DS3300 iscsi 12 disk system.

And things started running like crap. answer times vere above 30.000 ms!!!!!!! Especially the raid 5 vm's was extreme slow.

I migrated half of the servers back to the old storage, and things normalized again.

I contacted HP, who got some logs, and found out that there vere media errors on one of the disks in the raid 5, so i replaced that disk today.

But still, it is not satisfactionary that my brand new expensive storage system is not running fast!

HP made some recommendations, one was to install dual port hba in each servers, they claim it will run faster, and that i need dual ports hba to use load balancing. They also stated that i two switches would give me more performance - but in my head i have a hard time understanding why?? I got 8 GBit on each port, and a backplane of 24*8 Gbit. So thats why i ask you guys - imo performance would be the same with one or with two switches.

Setting a single iniator zoning - 1 host HBA zoned with all storag HBAs

Does that mean i should zone each HBA with all storage ports? - and what will i gain from it? (in iscsi enviorement network packets and broadcast will collide and create bad performance - is FC the same princip?)

Thanks

0 Kudos
m1kkel84
Contributor
Contributor

I have 8 luns.

currently i have set path to fixed, and i have fixed paths to 2 luns on each controller port.

0 Kudos
m1kkel84
Contributor
Contributor

I found this:

Single initiator zones are the way to go. If there’s no need, and for ESX there isn’t, for initiators to be able to communicate with each other then they shouldn’t be able to. Not only is this more secure, because initiators can’t communicate with each other, it also cuts out a lot of rubbish on your fibre. Rubbish as for instance “Registered State Change Notifications”. Although RSCN storms don’t occur that often anymore as they used to it’s still a risk of contention and should be avoided when possible. So if you’re doing a design or preparing for one keep this in mind: Single Initiator Zones are the way to go!

I will try to set this up.

Would i gain performance by using 2 hba's in each server - and selecting round robin path policy? And your point of view on two switches please.

THanks

0 Kudos