06-19-2014 04:49 AM
My client planning to move some HA products. But they still not sure which one to choose. The requirement is HA should be Active- Active setup in duel DC. They ready to buy some products, I recommend SF cluster but not sure which one is best. Kindly suggest me which combination suits with below products.
Available DC Setup :
1 Active DC ,1 Active near DC and 1 DR location
Products in hand:
1) Oracle & IBM hardware
2) storage (3 * 20K vmak)
3) EMC vplex
4) DB - Oracle RAC (11gR2)
Now which symantec HA clustering topology & method can help to achieve 99.99%. To make active-active in HA any minimum perquisites need to maintain from Network, SAN & storage.
How about IBM P780 , EMC vplex with SFRAC for Active - Active- Standby ?
06-19-2014 05:02 AM
This is pretty architectural question & there are too many of small & big considerations.
My recommendation, IBM P780 is the best from server hw standpoint, storage - you can again have multiple choices, I would again go with IBM XIV or v7000 series. SFRAC + VCS + VVR would be good choice.
So it would be like 2 node A/A or more at primary site & 1 node or more (A/A) in DR site, replicated via VVR
Refer below post for more comparision of Oracle DataGuard vs VVR
06-19-2014 06:52 PM
Thanks Gaurav !.
My management want to use AIX + RAC with data gaurd + vplex. Is there any best solution match for this requirement ?
06-19-2014 08:05 PM
My previous comment still stands same, pSeries with AIX with RAC+ODG & vplex can still go with solution mentioned above. Some good points about ODG can be seen in link posted above.
06-19-2014 11:42 PM
Let us which SFRAC version you would try if atall you choose , we can give more recommendations.
06-20-2014 12:15 AM
I can choose latest version. Since my OS is aix 7.1 & DB is 11gr2. I hope vplex also whould be newer version.
06-20-2014 01:31 AM
A few points to help determine your solution:
99.99% availabilty means you can have 52.5 minutes downtime a year which should incorporate planned as well as unplanned downtime
For unplanned downtime the following are ranked with longest to failover first:
However you need to also consider unplanned downtime - so for example Oracle RAC may (I am not an Oracle expert) require more downtime to upgrade than stand alone Oracle in a cluster.
All of these should give you 99.99% availability so if you don't need 99.999% availability then you can probably avoid the high cost of RAC.
If your Active-Active cluster is SAN connected then you can use standard SF mirroring which is a standard option of SF at no extra cost (although Enterprise version is preferable to give fast resync of mirrors)
If your Active-Active cluster is ONLY IP connected, the you can use Symantec Flexible Shared Storage (FSS) Mirroring which is a version of SFCFS which mirrors over IP, and also gives you faster failover - this requires Enterprise Licence
I am sure not sure of the capabilties of EMC Vplex, but almost certainly, both array would need to be EMC and probably the same model, but with SF, you could change one or both arrays and SF mirroring will still work fine (you can even migrate to new array, completely online with no downtime)
You can use VVR, Oracle DataGuard or EMC hardware replication - (use link Gaurav provided for comparison of VVR and ODG)
06-20-2014 03:40 AM
Thanks Mike !
The think actually I require now is all (Aix + SFRAC + Vplex) all three could work fine ? RAC doesn't need to understand vplex.but,SFRAC should know (Am i right )? Just heared the avaliability needs to 99.999. All this three support each other ?
After many search over internet I found document in EMC site talking about vplex with SF RAC in linux. also their support matix can do for AIX 7.1 with sfrac. In symantec site i only found the agent for vplex and not able to get any other reference documents for case study.
The same case in Oracle website, no info about SFRAC db support.
06-20-2014 04:05 AM
From a Symantec point of view you can use AIX + SFRAC + Vplex Metro configuration.
I don't know much about Vplex but https://sort.symantec.com/agents/detail/5384 says
VCS supports the EMC VPLEX Metro configuration in VCS global cluster and replicated data cluster environments; you do not need to install an agent for this support.
If you don't need an agent, then vplex is probably transparent to the host.
However, in your opening post you said:
1 Active DC ,1 Active near DC and 1 DR location
Which I read as 2 sites close together with synchronus copies of data in a single RAC cluster and a 3rd DR location using async replication,
Are you planning to use VPLEX Geo to replicate to DR and are you planning on using VCS at DR cluster connected to primary site using GCO (Global cluster option)?
06-22-2014 08:21 PM
I guess no plan for vplex Geo . But vplex metro with recover point.
06-23-2014 01:55 AM
Ok - there is a VCS agent for recover point, so this is fine. As with VVR, hardware replication, like recover point, will replicate more data than ODG so will require more bandwidth.