Storage Replica in Windows Server 2016

What is Storage Replica?

Block level synchronous or asynchronous, volume based data replication.

What does it do?

Replicates storage data from a disk attached to one instance of Windows Server to another disk attached to a different instance of Windows Server. Because it works at the block level, it doesn’t care about open files. It operates way down the driver stack:
storage-replica-driver-layering

It allows you to create stretch clusters (clusters that don’t have shared storage), or you can just use it to replicate data from one place to another.

How do you get it?

Storage Replica is new in Windows Server 2016. You need the Datacenter edition, but your Hyper-V/vSphere hosts are probably licensed for that anyway, right? It’a available in the “full” editions (i.e. with desktop or core) and also nano. It’s just a Windows Feature, so enable it with Server Manager or PowerShell:

Install-WindowsFeature -Name Storage-Replica

Synchronous vs Asynchronous

It’ll do both, you just need to pick the appropriate one. Firstly, is your network latency between the two servers greater than 5ms?  In which case you shouldn’t use Synchronous replication, the performance impact will be too great. If it’s 5ms or less, you can choose either, based on your needs.

Synchronous replicationstorage-replica-synchronous-replication

This works pretty much the same as oldskool SAN replication, kit like the EMC Clariion’s MirrorView feature does exactly the same thing (but for somewhat more than the cost of a Windows Server licence and some cheap server hardware – which is entertaining as certain models of Clariion actually ran Windows Server internally!). The key point with Synchronous replication is that the write is not confirmed on the source storage system until it has been written to the log disk on both the source and destination storage systems. This gives you peace of mind that your data is safe, at (usually) the expense of performance.

Asynchronous replication

See diagram for Synchronous replication, but confirm the write after step 2 – in other words, once the data has been written to the source log disk. The data is written to the destination log disk separately.

How Storage Replica changes your disk IO

On a volume that is being replicated via Storage Replica, all write activity happens to the Log disk. This data is then “destaged” to the Data disk, which thus only ever handles Because the Log disk is a log, the writes to this are sequential. Despite this, it’s recommended that you make the log disk an SSD – writes tend to slower than reads, and Windows (and applications, e.g. SQL Server, Exchange) can cache data for read operations in RAM. Write operations have to be securely written to disk, and so if the write speed of the log disk isn’t fast enough it’ll become the point of contention.

Is it safe to use now?

Probably – Microsoft has had customers running this technology on production systems since 2014. I plan on using it right now.

Want to know more?

Watch Ned Pyle’s excellent, informative, hilarious Ignite session video (where I stole the diagrams above from!). Hipsters beware…!

This entry was posted in Storage, Uncategorized, Windows and tagged , , , , , , , , . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s