VMWare / SAN / System Strategy

[gingerdazza]

Hi

We currently have two buildings (Primary and DR) with a SAN presented inm each location, with 1Gbps link between sites. We have blades in each building connected over fiber to each local SAN, and these are running ESXi 5.x. We currently have alot of systems running HA across the sites, making the "DR" site a live site too really.

I'm considering how best to move forward and am interested to know peoples thoughts and experiences...

1. Do people normally approach the primary site as the location where services should be delievered at all times (i.e don't run services from the remote site unless you have a DR event), or is this active/active scenario pretty standard?

2. Do people normally approach replication at the SAN layer, or at the hypervisor/VM layer. If the latter, what do you use?

Thanks

[john23]

-- Recovery site will live, when DR event happen. Still If you exploring active/active solution, you can target EMC VPLEX solution. It will provide the Distributed environment. VPLEX metro Round Trip time between sites is 5ms and GEO 50ms.

-- You can use Shared DR solution, For multiple DR site only one recovery site.

--SAN layer (Array based replication) is faster, compare to the vSphere replication. It depends what kind of loads are there, Because RTO time may increase if VM layer (vSphere replication, host based replication).

-A

[gingerdazza]

I think my problem is that the DR site currently hosts alot of mature services, with many of the IT services to the business providing resilience via this active/active layout. So, from a DR perspective I don't know whether it is best to bring a whole new philosophy into play and move towards active/standby for these sites, with a failover component designed in (like SRM) [of course, then you have 50% of your SAN effectively doing nowt 99.99% of the year], or whether to continue to pursue the active/active scenario where we have no DR failover as-such, but all services use both sites and reslience to any site loss built-in.

I'd assume alot of VM replication can be of limited use without a layer-2/stretched vLAN layout as recovering those VMs at the alternate site is useless unless you either (a) re-IP that VM in the alternate site or (b) provide that vLAN at the alternate site?