Microsoft® SQL Server® 2012 Bible

74 6

Part V: Enterprise Data Management

High Availability versus Scale Out

Clustering in Windows and with SQL Server 2012 is primarily designed around high avail-

ability rather than scalability. (Although you can get some scalability from the use of Read

Only AlwaysOn, which is built on top of Windows Clustering.) So what is high availability

and what is scalability?

High availability provides you with levels of redundancy to increase SQL Server’s uptime.

Although uptime of 100% is highly desirable, this is virtually impossible. Some level of

downtime is required to perform certain tasks such as server or database patching, and

there is always the chance of hardware problems that would impact that uptime number. A

high number uptime number would be the oft quoted “fi ve nines.” This is an actual uptime

of 99.999% over the course of a year. That means your SQL Server would need a total down-

time not exceeding 5.26 minutes. Only having downtime of fi ve minutes in a year is a huge

challenge. Frequently, larger servers with a lot of memory can take that amount of time

just to reboot once. If you had regular patching that required reboots, there is no way to

achieve that number. This is where clusters can assist in greatly reducing that downtime.

In a simple cluster you may have only two physical servers (nodes) that host a single-

clustered SQL Server 2012 instance. This instance has the facility to freely move between

the two nodes and be available on either one. (An instance may only be up and running on

a single node within a cluster at any given time.) This provides you with a level of redun-

dancy and fault tolerance from a hardware perspective that you would not have otherwise.

You may think of a cluster as providing you with a hot-standby physical server that you

can go to if another one goes down. As a bonus with clusters, you get automated failover

and invisible client redirection (with no special options required within application con-

fi guration strings).

With the Enterprise Edition of SQL Server 2012, you can have as many as 50 instances run-

ning on up to 16 nodes. You can only deploy 50 instances on a cluster when installing to

SMB shares, otherwise you will run into a 25 instance limit (one instance per available

drive letter in Windows).

The same clustered instance limit also applies to the Business Intelligence and Standard

Editions of SQL Server 2012; however, you are limited to only two node failover clusters.

With clustering at this level, you still have a single point of failure in your disk infrastructure because it is a shared

resource. Multisite geographically diverse clusters can be built, but they generally require the use of specialized SAN

replication techniques and third-party vendor tools.

Scalability is provided when you have data on multiple individual SQL Server 2012

instances. In this situation you may have an application that uses internal logic to write

data to a particular server and then have some kind of data replication techniques to

ensure that the data is pushed out to the other nodes. Alternatively, you could have a farm

c29.indd 746c29.indd 746 7/31/2012 9:51:58 AM7/31/2012 9:51:58 AM

http://www.it-ebooks.info