Mirroring data from one storage controller to another has been with us for some time. For many organisations it is the key to their Disaster Recovery policy.
Most storage vendors implement hardware mirroring using one of three possible methods (often offering all three). So how do they implement it? Lets assume in every case we have a production site (site A) and a secondary site (site B).
With synchronous mirroring, every block written at site A is mirrored to site B. Writes sent to the storage at Site A will not be acknowledged by the storage at site A, until the storage at site B has also acknowledged them. Ideally to avoid impacting performance, synchronous mirroring needs a large cross site pipe with low latency (think dark fibre and WDMs).
In traditional async mirroring, every block written at site A is mirrored to site B, but the writes to B lag behind the writes to A. How far behind depends on the change rate and the size and speed of the cross site link. Ideally the two sites are almost always in sync. I call this sync-aspirational mirroring, in that you really want Synchronous replication but your faith in the size of your cross-site pipe is not strong enough.
Asynchronous mirroring with snapshots
We can enhance asynchronous mirroring by using Snapshots to track changes between the site A and site B volumes. By enabling change block tracking at site A, changed blocks can be identified at fixed points in time and then mirrored from site A to site B. This means less traffic is sent from A to B than traditional async mirroring (or sync mirroring), especially if your applications over-write the same blocks on a regular basis (we call these over-lapping writes).
Asynchronous mirroring with WAN acceleration
If we take either form of async mirroring and add a WAN acceleration device to the link, we can speed up the mirroring process. This is because WAN accelerators can use a cache at Site A to track which blocks are being sent to site B. If this device can identify duplicate data (it depends on a local cache at site A to do this) it can try to cut down on cross site traffic, while also compressing the traffic being sent.
Dedup Async: Lets try something new.
The benefit of using snapshots with async mirroring is that we don’t mirror over-lapping writes. The benefit of using a WAN accelerator is that we start to reduce cross-site traffic by performing some de-duplication and compression. Actifio DeDup Async technology takes this a step further by performing a global de-duplication of the changed blocks captured using regular snapshots, ensuring that the only blocks ever sent from site A to site B, are blocks that are not already present at site B. This is genuine global multi-site de-duplication. And by compressing the unique blocks before transmission, we now have the most efficient possible use of our cross-site bandwidth. Actifio also encrypts the data during flight for good measure.
If this is not enough, Actfio can mirror between two heterogeneous storage controllers, for example, 3PAR to EVA or DMX to AMS2500. Actifio takes backups of your production data without the need to use traditional backup software and then replicates only new unique blocks to the remote site, optionally incrementally re-hydrating those blocks to create a DR image of your production data. Actifio can act as both your backup, recovery and copy-management storage at Site A and your DR copy, long-term backup storage and copy-management storage at site B. Plus there is no need to match the disk vendor or disk product at either site.
Sound interesting? You can contact Actifio here.
You can also keep up with what is happening at Actifio by following our blog.