What is the real benefit of zoning?
Way not just open all the connection?
In my 3 years old environment there wasn¿t any zoning and everything worked fine.
Way do I need it with EMC?
What is the real benefit of zoning? Way not just open all the connection? In my 3 years old environment there wasn¿t any zoning and everything worked fine. Way do I need it with EMC?
A few reasons I can think of off the top of my head:
1) Registered State Change Notifications (RSCNs) get propogated to every device on your san when you reboot a server. When you use zoning RSCNs only get propogated to devices within that same zone.
From Brocade Design Deployment and Management Guide page 124: Be aware of mixing different HBA vendors in a single zone. Each vendor HBA responds differently to RSCNs, a method to notify an HBA for device discovery, and may cause one of the HBAs to lose the zoned device. In addition it is recommended to have single initiator zones, that is one HBA per zone. Separate HBAs from each other for clustered hosts. Allow each HBA to see the same storage but not each other. Once again, RSCNs, may cause the clustered host HBA to lose the storage array. www.brocadejapan.com/resources/tl/pdf/DDM_v3_0_53-0000366-03_C.pdf
2) SCSI Bus resets can be propogated from an upset server and spam all visable targets. This can cause every target to abort any current I/O with any server. Bus resets against a fibre tape device can cause tape rewinds and tape corruption, this is also one of the reasons EMC doesn't recommend sharing disk and tape on the same HBA.
3) Logon history tables for disk arrays would be a mess (as already mentioned).
Security and administration are two major points that come to my mind without EMC in mind at all. And it's not just EMC. If you worked with clustered solutions or mix of tape/disk environment or even mixed OS then zones are something you must think about.
Way do I need it with EMC? because emc designs the systems for single intiator zoning (i believe since they push that requirement). this also helps them troubleshoot the connectivity issues by isolating things. or maybe they just like it to be that way
zoning comes from the SCSI legacy. SAN basically emulates a SCSI transport over fibre. old SCSI systems were used on a standalone machine where one SCSI controller would connect to many scsi devices (targets). This setup assumed that there would be only one controller on the bus to control the devices.as discussed somewhere on the forum earlier, this is referred to as single initiator zoning.
if you open all the connections, all HBAs (SCSI controllers) see all the targets and even HBAs...this can lead to connectivity issues if the HBA drivers or targets (sp ports) run into any logical semaphore whhich they havent been desinged to handle.
practically we have seen issues while detecting storage or intermittent connection issues.
to overcome this and to maintain the expected environment for eveything, zoning helps...even otherwise, this isolates traffic between irrelevent nodes.
to add on that, with help of zoning we can isolate host and arrays at switch level for more control over interconnection.
there is a logical limit to how many HBAs will login to a single storage port. by opening all the connections, all hbas try to login to available ports and chances are that you will run over the limits soon. zoning allows you to prevent that.
Zoning prevents aggresive initiator HBAs from trying to log into where they shouldn't. In other words, an FA port on a DMX should not login to another FA port. That would cause problems if that happens. Zoning keeps all the devices in line and seeing only what they are supposed to. Not to mention that there are numerous security issues this solves too.
All the replay here doesn¿t really justify zoning. I have used HP EVA on the same SAN. And we haven¿t use ZONING. The overhead zoning is adding to the management task is tremendous. What will happen if I simply allow all traffic on the FC-switch? Why masking isn¿t enough? Have anyone try it? I have tape at the SAN and with it we are using switch port zoning. I can argue a bought security and troubleshooting but I don¿t want to, at this thread.
Take for example EMC Celerra. FA ports that are used for connecting Celerra to DMX do not have VCM bit enabled. That means that there is no masking done on those FA. I am sure you are familiar how windows tries to grab any disk it can and write its signature to it. So imagine without zoning, my windows boxes would start logging in to my FAs and starting corrupting my Celerra drives.
imagine you have just a clariion/symmetrix and 16 hosts with 2 hbas each. in this case, every HBA will log in to each array port. i.e. 32 initiators on a single port from array view. i believe this is a upper limit. now if you add one more host, it will not be able to register itself on the array.
this was specific to EMC in a very basic setup, but i am sure other vendors also have these limits set.
Ok. Celera is a good zoning needs, I don¿t have Celera. More over, I have another site with IBM DS-4300 (fast600) and EVA 3000, I don¿t use zoning on this site. It has 45 hosts and 70 HBA. I don¿t use zoning on it. I use windows and ESX 2.5.3. Every one (host and storage) ¿see¿ the others it¿s working ok for the last 15 months.
for a small setup like this where "Every one (admin) ¿know¿ the others", zoning may not be required. zoning is not a mandatory step but it is optional for tidy things in big shops.
My setup has few hundred hosts and more than two handful number of arrays...such setup cant rely on the zoneless approach...we need to limit who sees what and thereby reducing mess-up when someone accidently deletes or works on incorrect devices.
JasonBailey
147 Posts
1267
0
Posted April 6th, 2007 17:00
Way not just open all the connection?
In my 3 years old environment there wasn¿t any zoning
and everything worked fine.
Way do I need it with EMC?
A few reasons I can think of off the top of my head:
1) Registered State Change Notifications (RSCNs) get propogated to every device on your san when you reboot a server. When you use zoning RSCNs only get propogated to devices within that same zone.
From Brocade Design Deployment and Management Guide page 124:
Be aware of mixing different HBA vendors in a single zone. Each vendor HBA responds differently to RSCNs, a method to notify an HBA for device discovery, and may cause one of the HBAs to lose the zoned device.
In addition it is recommended to have single initiator zones, that is one HBA per zone.
Separate HBAs from each other for clustered hosts. Allow each HBA to see the same storage but not
each other. Once again, RSCNs, may cause the clustered host HBA to lose the storage array.
www.brocadejapan.com/resources/tl/pdf/DDM_v3_0_53-0000366-03_C.pdf
2) SCSI Bus resets can be propogated from an upset server and spam all visable targets. This can cause every target to abort any current I/O with any server. Bus resets against a fibre tape device can cause tape rewinds and tape corruption, this is also one of the reasons EMC doesn't recommend sharing disk and tape on the same HBA.
3) Logon history tables for disk arrays would be a mess (as already mentioned).