SAP Availablity with vLockstep

[Former Member]

Hi,

Has anyone implemented SAP Availability with vLockstep? if so, can you please share your experience? Would be ideal if you could let me know the following,

a. How was the overall experience?

b. As I understand majority of configuration happens at the virtualization layer & no configuration needs to be done at the SAP Application layer, is this correct?

c. Any issues/Risks that were noted

d. Does this in any way compliment VMware Site Recovery Manager?

e. Can we leverage ACC (Adaptive Computing Controller) in anyway?

Would appreciate any information as we are evaluating the solution.

Regards,

[Former Member]

Hello,

allow me to add something from the technical SAP-VMware alliance team.

vLockstep is the technology behind the feature called Fault Tolerance (FT). To enable FT, the Virtual Machine (VM) must not be configured with more than 1 virtual CPU (vCPU). This constrain limits the use cases. In SAP environments though, there is one instance that can easily be run with 1 vCPU: the SAP Central Services for ABAP (ASCS) and / or J2EE (SCS). The Central Services contain message and enqueue server and do not need much CPU resources.

At some point, some customers are afraid of the latency FT will add to the enqueue communication (see [VMware vSphere 4 Fault Tolerance: Architecture and Performance|http://www.vmware.com/resources/techresources/10058] - 3.4. Netperf Latency Bound Case). Our internal tests with an FT enabled ASCS showed ~ 5 % CPU utilization with 200 active SD users. The latency FT will add is forseeable because it is very constant, so the maximum amount of users that can be handled with appropriate response time by your environment is dependant to the given Single Thread Performance and therefore can be determined through testing.

To configure an (A)SCS as a standalone instance, consult the Installation Guides of SAP. On Windows, there are some specialties which we are documenting on SAP Note 1609304. This Note doesn't sound very confident though, but the described configuration is supported by SAP and VMware and we constantly will work on that solution to make it more adoptable.

In the release notes of Site Recovery Manager 4 you can read:

Support for Protecting Fault-Tolerant Virtual Machines.
SRM can now protect virtual machines that have been configured for fault-tolerant operation. When recovered, these virtual machines lose their fault tolerance, and must be manually reconfigured after recovery to restore fault tolerance.

When it comes to ACC / LVM, I don't quite understand how you want to leverage FT in that context. You can not configure FT via the ACC / LVM software. But the ACC / LVM management of FT enabled VMs is the same like it is with all other VMs.

I recommend also reading [VMware KB 1013428|http://kb.vmware.com/kb/1013428] - VMware Fault Tolerance FAQ.

Kind regards,

Matthias

[Former Member]

Thanks Matthias for the information. So to summarise your reply, VMware FT is not yet suited for SAP vide Implementations. The SAP Note 1374671 had 2 options for clustering a. with-in the guest OS like MSCS or b. outside of the OS like VMWare HA. We are not too keen on MSCS but without MSCS and FT, the DB is the single point of failure. Ofcourse we can split the CI & DB and cluster DB but considering we have 10+ production server is a lot more servers.

What is the SAP/VMware's recommendation on HA? HA as in local failover/online failover locally. We already have SRM (Site Recovery Manager) for DR. I would really appreciate if you could point me to any best practices documents on this.

Regards,

Guru

[Former Member]

Hi Guru,

with FT you cannot "cluster" whole SAP systems, only (A)SCS would be an appropriate use case.But VMware has more High Availability methods than only FT. The solution of choice strongly depends on your availability requirements, so you first have to determine which degree of availability you want to achieve.

VMware HA, as a base line, is also very helpful to keep your VM as a whole up and running:

[vSphere High Availability Deployment Best Practices|http://www.vmware.com/resources/techresources/10232]

Next, VMware created an "Application Awareness API" which partners and customers can utilize to develop an own monitoring solution. Symantec did so, offering [ApplicationHA|http://www.symantec.com/business/application-ha] which operates critical applications like Databases and SAP instances and significantly increases their availability:

[Virtualizing Business-critical Applications with Confidence|http://www.vmware.com/resources/techresources/10147]

So you could use FT for your (A)SCS and ApplicationHA for your Database. If this high degree of availabilty is still not enough, you have to go for in-guest clustering. This can be done with MSCS / WSFC as you already stated. There is also [Linux in-guest clustering|http://www.cc-dresden.de/en/whitepaper] for VMware available.

In attached document, which is not yet available for download, we assembled some of the above mentioned combinations. It would be definitely worthwhile to read for you.

Kind regards,

Matthias

[Former Member]

Thanks Matthias! appreciate your reply.