Hi

I am in a process of designing the SharePoint 2013 farms and considering high availability and disaster recovery model, using both data center and the office servers together.

Client decided to go with a streamlined topologies and have chosen Denver data center to have the SP 2013 farm. Farm consists of 2 WFEs Network load balanced and a single DB Server with a strong hardware setup. I believe based on our SP 2010 usage and number of users  less than 750, this is fine.

For the Failover scenarios, we have one VM setup for WFE and one DB server(physical) with the same production db configuration via mirroring. I believe log shipping approach we are taking.

Now here are 2 scenarios:

1. Data Center Database fail:

When primary DB failed, we want to switch to mirrored database (we dont have budget for witness server, so automatic is not an option) with minimal or no downtime. I worked with networkadmin and its less than 1ms latency and 1GBPS bandwidth. So switching from primary to mirror manually is very minimal downtime.

1. When whole datacenter is down:

Need to redirect requests to the office VM(WFE backup server) and switch mirror Database.

My thought to consider the stretched farm scenario, so the 3^rd VM(WFE) will be part of the Network load balancing servers. If Data center is down, automatically all requests route to only 3^rd VM(wfe).  We just have to switch the mirror db server, their by having minimal downtime.

The distance between data denver Datacenter and my office Servers(either Chicago or San Francisco) is more than 1000 miles.

My questions are:

• By including my office VM in a NLB setup leads to any performance issue on a everyday requests or is it fine as long as the latency is less than 1ms?
• If I go with a automatic switch from primary db to mirror db, witness server is must or is there any other script/timer approach we can switch?
• Considering all these details, what other best practice approach you would recommend?

Many thanks

Shri

1) A witness can be a small file server (e.g. 40Gb vDisk, 1GB RAM, 1 vCPU).

2) This would be performed by networking equipment or a DNS reconfiguration.

However, it is simply not possible to stretch a farm 1000 miles. In a vacuum, 1ms one-way is a maximum of 186 miles (the distance light travels in one millisecond, again in a vacuum, at the speed of light). Since we don't live in a vacuum, and networking equipment, as well as the physical medium which the packet travels over (copper, fiber) introduces latency, you would need to be < 186 miles in order to stretch a farm.