Need help understanding CA Synchronous vs Enhanced Asynchronous

[T Downing]

I have been reading through some of the documentation and some posts on here and need to get a few things straight about CA in general and how Enhanced Asynchronous works vs Synchronous

We have 5 EVA 8100's, 3 for production and 2 for DR the we just got. We are trying to get our arms wrapped around CA to make the right choices on how we lay things out and how we do our replication.

First clarification is if all our controllers are running XCS v6.11 then when we select async we are doing enhanced async? On DR groups that have completed the initial Full Copy I can go into Command View or RSM an select Async but it does not specifically say Enhanced Async so I am making the assumption that I am using Enhanced Async due to my XCS version.

From what I understand the only difference between old async and enhanced async is the buffer being written to disk vs cache, thus allowing for better handling if your connection between sites had high latency or being able to handle brief line outages that in the past could have forced a full copy since the cache buffer would have been to small to handle amount of data written to the cache during the outage. Is this correct?

Next once there is an outage and you are in logging mode are sync and async basically the same, where both are logging to disk? Then post outage do they basically operate the same during the merge till the log (at least on the sync side) is empty and the merge is complete?

Though they have basically settled on doing Async due to the potential of some apps/os's not dealing with the extra latency well I am wondering since the distance to the DR is pretty minimal if in someways we would be better off using sync or a mix depending on application. Do most people just use one or the other, I know it partly comes down to distance, the size of the pipe and how much data is getting pushed across.

I am sure I will think of more latter, every answer brings up a new question. Any help is appreciated.

[Del_3 · Accepted solution]

You are correct xcs 6 to xcs 6 is enhanced async. You can not do the old style async on xcs 6.x. With the enhanced async you are alyways writing from the log.

There are many caveats regarding enhanced async. Eg. You must be in sync mode to:

- Create DR groups or add members.
- Do full copies at initial DR group initialization and triggerd by a full log and several other conditions.
- Failover for testing test.

So even if you operate in async there will be times when your link must support a sync tunnel.

[Peter Mattei]

Hi

First to the previous post:
The limitations stated are not true!
You do not need to switch to synchronous in order to do the actions described!!

Then, most customers are using synchronous CA since this is easier to manage (failover) and you are always up to date with your data.

CA Async on EVA8100 is always enhanced CA.
You would want to use it if your link cannot sustain peak write rate or your distance introduces too much latency.

Please have a look at the CA planning guide, current link: http://bizsupport.austin.hp.com/bc/docs/support/SupportManual/c01375142/c01375142.pdf

Cheers
Peter

[Del_3]

I repectfully disagree Peter. I did some digging. From the 2006 CA Enhanced Async presentation at Tech Forum - slide 27 and 28.

Cannot create DR Group in async mode
Changing from sync to async:
Diskgroup must have enough disk space for selected log size on both source and destination
DR Group must be normal (good link/no logged data)
Changing from async to sync:
DR Group must be normal
Actual mode changes to Sync when Async rundown completes
Occurs automatically if full copy is initiated
Returns to asynchronous upon completion of full copy

While in async:
Cannot add or remove DR Group members
Cannot change DR Group log size
Cannot change DR Group member size
96 async DR Group members (temporary)
Continued restrictions
Low bandwidth limitations
Turn off any write acceleration in switches/routers
No exchange based (OXID) routing supported or allowed in switches/routers
While in async:
Cannot add or remove DR Group members
Cannot change DR Group log size
Cannot change DR Group member size
96 async DR Group members (temporary)
Continued restrictions
Low bandwidth limitations
Turn off any write acceleration in switches/routers
No exchange based (OXID) routing supported or allowed in switches/routers

[Peter Mattei]

Del,
I will review your statements and follow up tomorrow!
Cheers Pete

[T Downing]

Thanks for the answers, I appreciate the help thus far.

[Joseph Salzer_3]

Del, Peter is correct. The limitation you are referring to, i.e, that synchronous replication is required when creating enhanced asynchronous DR groups is no longer true. It has not been true since XCS 09xxxx code and I believe even 6.2xx as well.

You're correct as far as the old XCS code is concerned, but it is no longer the case.

As far as adding and removing DR group members; you're correct though. I believe you have to switch from enhanced asynchronous to synchronous before you can add or remove members.