Zoning Methodology

 
SOLVED
Go to solution
Carlos A. Munoz Lopez
Frequent Advisor

Zoning Methodology

Hi guys, I have a doubt. I know different methodologies can be implemented when zoning a SAN switch, but in this scenario I want to know the differences between HBA and application (i.e exchange, oracle) zoning in terms of performance. As far as I know, zoning should be implemented to control access when different hosts are attached to a SAN.

For instance, in a 2-node cluster environment, by choosing the first methodology mentioned above, you create a single zone per HBA, in this case if you have two HBAs in the first node you will have two different zones configured in the switch, likewise for the other node. At the end you might have 4 different zones in this environment.

On the other hand, you have the same scenario and you choose the second methodology (application) and create a single zone for both nodes (all HBAs in each node should belong to a single zone in the switch).

Which of these two cases provide a better performance? By performance I mean a better utilization in I/O by the fibre channel in which there is no resource competition. Or is the performance/utilization regulated by load balancing software installed in the OS? Let's assume we have and EVA 4100 and 4Gb SAN, with servers running Windows 2003.
4 REPLIES 4
marsh_1
Honored Contributor

Re: Zoning Methodology

hi,

the zoning itself should not affect performance , it is simply there to ensure correct access between endpoints, unlike trunking which aggregates links together for greater performance.
single initiator zoning might help guard against disruption in a zone interrupting a whole application.

fwiw

Amar_Joshi
Honored Contributor
Solution

Re: Zoning Methodology

Carlos,

Let me explain a bit more why single HBA per zone is recommended. This recommendation is also called "single initiator per zone" because host HBAs are capable of sending State Chnage Registration (SCR) and Registered State Change Notifications (RSCN) which may easily interrupt the operations of other host HBAs (if present within the same zone). This doesn't cause any big problem for FC protocol but always recommended to avoid such occurance. By zoning single HBA in one zone will always be a part of best practices.

Also, in case of sinlge zoning for clustered nodes, it's risky that if someone deletes one zone both servers will be disconnected together, so going with multiple zoning is better option for those who are prone to make mistakes (all the humans).

Hope that helps.

my 2 cents worth.
Sivakumar MJ._1
Respected Contributor

Re: Zoning Methodology

Hi Carlos,

The communication between the HBA card(Initator) and Storage Array (Target) is based on the zoning.

Amardeep and Mark has given a clear picture..

The Best Practice in Zone which most of our customers follow is Single Initator Single Target.

In 2 node cluster you have 4 zones out of which 2 zones act as failover when one node is down..

Performance is based on many aspects, Disks in the Arrays, HBA Speed, Switchport speed,the throughput etc..



Carlos A. Munoz Lopez
Frequent Advisor

Re: Zoning Methodology

Now, my picture is clear. The subject came up from a discussion I had with some HP Partners in which they stated that HBA zoning involved a better performance along with the benefits you mentioned. Thank you guys.