SharePoint 2010 High availability - scenario

Question

I am designing a SharePoint 2010 farm topology to provide high availability for a three server farm (2 SharePoint, 1 SQL). The servers are setup that SQL is using SQL Server clustering, and there are two servers (across two physical locations) which we can use to install SharePoint 2010.

The Technet "Plan for Availability" article describes redundancy across WFE servers, and redundancy across application servers.

Is it possible/a good idea to provide redundancy (using NLB) having two servers, each as a WFE and providing services?

Is there a better way to provide high availability for SharePoint given two servers as described above?

Answer 1

While coming up with the best possible DR and HA plan for SharePoint is highly subjective and beyond the scope based on the information shared , I still have some really short and quick pointers for you.

Request Balancing
Consider the 2 WFEs as "N" , you should think of using N + 1 architecture where you have a cold standby WFE (+1). Make them behind a Hardware based NLB, since Windows NLB supports not more than two systems per group and has the problem of sticky sessions.

From request balancing perspective its like in the event of complete failure N , your SharePoint does not become offline completly. Additionally, during server patching scenarios this topology would likely to help you with a least downtime upgrades incase you have groups of N+1 architecture working together.

Service Application Planning
While you havent mentioned what service applications will be used in the farm, SharePoint by design allows scaling on the basis service applications primarily. You need to plan for adding more servers to the farm, incase you wish to use service applications that are generally more resource hungry like Search, Excel and Word Automation services,etc. If you wish to configure your WFEs to act as Service application servers as well(which is not advisable), make sure you increase the current main memory by 2 times from the present and cpu cores to plus four.

If you just wish to use User Profile , BCS and Managed Metadata services probably these can be configured to run on both WFEs without the need of hardware upgrades as they relatively consume less memory and cpu.

Database
Clustering SQL Server will give you redudancy, however if you are planning this for production you should consider using mirroring for Disaster Recovery site over SAN as well.Additionally, from performance and availability perspective think of separating Report Server from the primary host instance in the cluster if you are planning to use Reporting service over SharePoint Intergrated mode. Clustering SQL wont give redundancy on Reporting Services. So you may need
SQL Cluster + 1 Report Server(which will be also running SharePoint WFE role with reporting services addin as well)].

Answer 2

The minimal high availability topology is indeed possible with a 2-tier implementation composed of a 2 servers dedicated for SharePoint and acting as a WFE & Application Server (all required application including search) under a network load balancer and a cluster SQL Server (or mirrored) to handle the persistence layer (If I understood correctly, you only have a single SQL Server so watch out for that single point of failure in your HA plan)

So to reply your question

Is it possible/a good idea to provide redundancy (using NLB) having two servers, each as a WFE and providing services?

That's possible, you should just be sure that your servers have initial resource to handle all the services and web server role (fortunately 2010 is already quite happy with a proper server with 16gb ram).

Answer 3

In order to leverage a stretched farm (no longer supported in SharePoint 2013), you must have 1Gbps bandwidth and no greater than 1ms of latency between all members of the farm, SQL Servers included.