Why not RAID?

Sanify Cluster is more reliable and more versatile than traditional RAID.

No Single Point of Failure

There are dozens of iSCSI or Fibre Channel attached RAID solutions on the market. Although they can withstand various common hardware failures and generally improve reliability, they generally all have single points of failure and many vendors outright lie about it.

Consider that even highly redundant boxes with dual RAID controllers and multiple power supplies all have single points of failure, such as a shared backplane or a shared power bus. If these solutions truly had no single point of failure, they would be sold as two boxes and the cabling would probably cost extra!

Even the redundant parts of dual controller boxes aren't particularly isolated, either electrically or physically. It's easy to recount a wide variety of whole box failure scenarios ranging from destructive power supplies to fumbled screwdrivers. Clearly a convincing solution which truly has no single point of failure needs to be a cluster of separate computers with no hardware in common other than the network which connects them.

That being said, some unscrupulous vendors will try to convince you that two storage devices (which each store a complete copy of your data) is a solution which won't corrupt your data when a piece of equipment fails. Although intuitively believable, this is patently untrue. Consider that two storage devices can run independently of each other by design and it isn't mathematically possible to prevent split brain (two different versions of your data) with only two storage devices: no distributed system can break a tie to determine where your most recent data is without at least three members.

Never Degrade

Traditional RAID 1, 5, 6 and their variants were designed to handle disk drive failures, and they generally do this by degrading redundancy. Firstly, this strategy only addresses media failures and means your data is off line when a controller or it's host fails, sometimes taking weeks to resolve. Secondly, typical RAID solutions aren't configured with hot spares and when drives fail, they run in a degraded mode for extended periods of time, putting valuable data at risk.

In the real world, most problems accessing data aren't media failures: networks are interrupted, the power fails, and controllers die. In particular, highly available network storage clusters are exposed to far more transient faults than to media failures. Transient faults do not substantially increase the risk of data loss with traditional media, and ideally should not increase the risk to redundant storage. This is not generally the case though since typical solutions treat transient faults as media failures.

For example, pulling a functional drive out of a running RAID array, or pulling a network cable out of a cluster node, is fundamentally a transient fault since the media hasn't permanently failed. Anyone who's ever accidentally or even intentionally done this can attest that a typical storage solution responds by degrading, thereby putting important data at substantially greater risk. It is particularly important to consider that data on a degraded RAID array or cluster is actually less safe than if it were all on a single drive.

Sanify Cluster does not take dangerous short cuts. It is immune to data loss under all possible transient fault scenarios even if a storage device fails since it never degrades when faults occur. Given sufficient space it simply redirects io and replicates your data on the remaining working equipment.

Scalability

The size and number of disk drives and controllers in a typical RAID based NAS device is fixed at the time of manufacture and initial configuration. Additional devices, particularly if they are newer hardware, do not benefit your existing volumes and are not an incremental cost over your initial investment. Migrating to a larger NAS device involves considerable downtime and effort to create new virtual disk drives, copy your volumes and reconfigure all of the client connections and security.

With Sanify, adding a node (which can be newer hardware with more capacity or better performance) is a small incremental cost and all existing volumes gain the safety and performance benefits afforded by the additional hardware. The data in the cluster will automatically and effortlessly migrate with zero down time.