Extended RAC: ASM vs VPLEX

Hi Bruno,

Have you ever tried this one on powerlink (Customer Technical Presentation: EMC VPLEX Family Technical Review):

http://powerlink.emc.com/km/live1/en_US/Offering_Technical/Presentation/VPLEX-Tech.pptx?mtcs=ZXZlbnRUeXBlPUttQ2xpY2tDb250ZW50RXZlbnQsZG9jdW1lbnRJZD0wOTAxNDA2NjgwNjIzNDlmLGRvY3VtZW50VHlwZT1wcHR4LG5hdmVOb2RlPTBiMDE0MDY2ODA0ZGEyMGVfR3JpZA__?

Please go to page 15 of the presention for detail.

Wish you good selling.

Jean

Bruno,

VPLEX+ASM is the EMC recommended approach for extended RAC. you can read the following White Paper for more information. http://powerlink.emc.com/km/live1/en_US/Offering_Technical/White_Paper/h8930-vplex-metro-oracle-rac-wp.pdf?mtcs=ZXZlbnRUeXBlPUttQ2xpY2tTZWFyY2hSZXN1bHRzRXZlbnQsZG9jdW1lbnRJZD0wOTAxNDA2NjgwNWZkMWRkLGRhdGFTb3VyY2U9RENUTV9lbl9VU18w

VPLEX has huge advantage especially on cross-site RAC configuration. Cross site HA software/hardware has to be used along with ASM in extended RAC 11g R2 environment.

Hi,

the Oracle recommendation is this:

http://docs.oracle.com/cd/B28359_01/server.111/b28282/configbp005.htm

this doc explained how to create a extended RAC only with Oracle features (basically with ASM).

but, what I want is show how VPLEX can create a better environment than this.

I want to prepare a PPT (pros and cons) comparing an extended RAC only with ASM and without VPLEX and an extended RAC with ASM and VPLEX.

for example:

- No host CPU cycles consumed on mirroring with VPLEX.

Have somebody more ideas about pros and cons?

Thanks in advance.

Hi Runo,

Pros and Cons for "No host CPU cycles consumed on mirroring with VPLEX" is: VPLEX is host-independent, so it does not consume CPU cycles.

And for the same reason, other benefits/advantages showed on page 15 of that customer-ready presentation (as I mentioned earlier) include:

• With VPLEX, users also do not have to configure failure groups, path preferences, or remote quorum disks.
• DBAs do not have to manage storage – everything can be done by the storage administrator.
• From a performance perspective, VPLEX delivers all reads locally – no more having to define whether the reads are local or remote in ASM. 
• In addition, these reads will take advantage of the VPLEX cache, shortening the latency of the solution.  
• And because all of the replication between the sites is managed and executed by the pair of VPLEX clusters, there is a significant off-loading from the host CPU – giving you more room to grow, or even the possibility of employing smaller servers.

Hope these pieces of information help.

Best Regards,

Jean

Copy/Pasted from my blogpost....

So what are the advantages, from a business perspective?

1. I would say, the most important business reason to run stretched clusters on VPLEX is: reduced risk. VPLEX hides complexity of storage replication for the database (and therefore, for the storage or database administrators) and therefore reduces risk of data corruptions, problems with servers not failing over when you most need them, you reduce performance issues in both normal operating mode, failure state and when recovering from failures. Could you do this without VPLEX? Certainly, if your administrators don’t make configuration mistakes, if the clusterware does not have bugs or design flaws, if your storage system behaves as expected, and if you make all configuration changes (adding servers or storage, upgrades, etc) correctly. Theory and reality often don’t match.
2. You also make it fairly easy to perform failover testing (potentially while the cluster keeps running, and ideally without even breaking the remote replication). This means you’re protected against disasters even during D/R testing. Caveat (but I must be honest): this is future functionality in VPLEX (and I will be glad to push our engineers to drive this functionality if I can give them feedback from customers who really need this).
3. It is application independent. I focused my blog posts on Oracle RAC for two reasons. First, it is my job to deal with Oracle technology. Second, Oracle RAC is (as far as I am aware) the only database that can be deployed as truely active/active clusters (meaning all cluster nodes can access the same database data simultaneously). If there would be another database or application that behaved in an active/active manner (and there must be some around) they will probably benefit from VPLEX in the same way. But even active/passive cluster configurations can benefit from VPLEX. For example, Microsoft clusters (MSCS) can automatically fail-over based on the quorum disks defined in VPLEX – where normally they would need to be manually restarted in case of certain failures that cannot be resolved with the standard quorum disk or node majority mechanisms. In particular, VPLEX allows VMWare environments to benefit from the access-anywhere behavior of storage volumes. This was initially targeted more to cloud application mobility, and in a lesser extent with high-availability clustering. But VMware offers clustering features as well (VMWare HA and DRS, on top of VMotion). So VPLEX allows building a stretched VMWare cluster where failure of a complete datacenter results in all failed virtual machines to be automatically restarted on the surviving datacenter location. Without manual intervention and without risk of split-brain issues.
4. It opens the possibility to create application level consistency between multiple environments with what we call “Consistency groups”. Application level consistency is a tricky topic and most of the people I talk to are not fully aware of the implications. So keep me to my promise that I will write another blog post on this sometime.

Disadvantages? Probably. Here are the objections I expect from my customers:

1. Vendor lock-in. EMC is the only vendor who can do this. Big problem? I don’t think so. It’s not like any business application that once rolled out, you can never get rid of. If, for reasons I cannot imagine, you would not like EMC as a trusted vendor, you could get rid of VPLEX and continue without stretched clusters or implement it using another (IMO, inferior) technology.
2. Price. Well – that’s a good one. Of course, VPLEX isn’t free – I praise the fact that I’m not in the sales department and therefore not aware too much of our pricing. I bet VPLEX is not completely free. But in relation to network bandwidth between the locations? Oracle database and apps licensing? And the potential performance benefit? If you do the math and include everything, I would be surprised if you cannot justify the investment.

By the way, what is the business value of risk reduction?

Several things to talk about here....

Challenges with ASM mirroring

1) Host Mirroring Consumes Valuable CPU cycles that often translates in requiring more CPUs and more Oracle licence cost

2) Admins need to configure failure groups, path preferences and other ASM mirroring functions

3) Cross connect between SANs and host become very complex

4) Scaling out is just another added dimension of complexity

4) I think the biggest reason I would adopt the vplex approach opposed to ASM is the solution is not limited to an Oracle database only. It can encompass the entire IT portfolio supporting that specific application. In order to ensure a smooth transition to any availability issues, you need to maintain all the systems consistent and in sync.

In addition, you need to be careful in defining your HA solution since it will mandate a lot of your DR solution.

Oracle best practice recommends that any mirroring should be done on the storage and ASM mirroring should be used when that option is not available.  Look into to ASM Technical Best practices document in Oracle support.