Data Replication (over Distance)

[The Brit]

Hi Guys,
We are currently investigating Data Replication options (for our Tier 1 (Critical) data). At our site we use HBVS on OpenVMS (7.3-2), and Critical Tier 1 data will be setup so that one unit is presented from an HP XP10000, with the other coming from an EVA8000. Tier 2 is two units from 2 EVA8000's, and Tier 3 is unshadowed, hosted in an EVA6000.
HP proposes a setup where our XP10000 performs remote replication with an SVS200 (which is (to my understanding), a diskless XP shell) which would be at the remote site, and would control "attached" storage (i.e. local storage in EVA's).
Hitachi proposes a similar setup using (at least) 2 x NSC55's (equates to XP10000 with Hitachi S/W & F/W), (but preferably with at least one end being a USP (XP12000 equiv) or a USPV (XP24000 eqiv).

In our case, the distances are quite large (~2400 miles), Baltimore to Reno, (Latency ~50ms, so definitely asynchronous),

We would however, be interested in hearing about any solution currently being used over any distance (greater than say, 5 miles), regardless of storage vendor. The only condition is that it must be certified for OpenVMS (since for us, this is not going away soon).

Within our shop, we also run Windows, Linux (Redhat), and a residual amount of Tru64.

Appreciate any information you can share.

Dave.

[Ron Marchak_1]

Not sure about the VMS component, but to replicate data over a WAN there are third party vendors that handle this quite well which removes the latency. Basically between your routers, you can connect an appliance, which comunicates to another appliance on the other side. The performace increase is quite impressive. The vendors to look at are Riverbed and Blue Coat. These are the two big payers in that area. I am sure there are others! Hope this helps.

[EdgarZamora_1]

We have a VMS environment (like you) and looked at EVA Continuous Access and EMC SRDF and management decided to go with SRDF. Our main datacenter is in Ft. Lauderdale and the DR site is in Kentucky (approx. distance 1200 miles).

We have just put this in production and is working quite well. It is an asynch solution and we "synchronize" every hour.

Let me know if you'd like to hear more and I'll provide email address.

[The Brit]

Thanks for the info Ron. We were already aware of the advantages of using WAN/SAN Extension Gateway's/Accelerators, however as I understand it, these only address the issues related to the actual transportation of the data. We still need a 'front end' for the process, i.e. some type of application for admin/config/sched of the replication process.

It is likely that might have to use accelerators at both ends anyway, so I appreciate the names of additional vendors. For your info, another is "LightSand" in Texas.

Edgar, thanks for the information. I need to find out how my superiors feel about EMC. I may be intouch.

Dave.

[EdgarZamora_1]

Another option you might want to consider is ASCI's RemoteShadow product. It is a host-based asynch shadowing product. They don't support shadowing the system disk though. If that's not a problem for you then you should definitely check out that product.

www.advsyscon.com

[Ivan Ferreira]

I think that you should know that asynchronous replication is not for all kind of data. This is from an HP document:

Traditionally, you can choose from two modes of operation for remote replication of data: synchronous and asynchronous. For long distances, synchronous replication mode may impose unacceptable latencies (response time) in returning write completion to the server. Asynchronous mode may reduce the response time (write completion back to the server), but it puts each of the outstanding writes at
risk if the source storage system is lost between the time the write is acknowledged as complete to the server and it is written to the destination controller.

Because both synchronous and asynchronous replication saturate at approximately the same rate, as shown by Figure 12, in disaster-tolerance planning environments the recommendation is always in favor of synchronous mode, for data protection, when all other factors such as peak I/O rate are equal.

So, my question is, what if you are replicating database data and the storage crashes before sync is done?

[The Brit]

Ivan,
We are not trying to replicate any "databases", just a suite of flat, RMS files. For us, the issues with synchronization apply whether the data is local or remote, so replication, either synch or asynch, doesn't really make things any worse.

Dave.