Duplication is primitive, archaic, dépassé for large storage.

For the longest time data duplication through mirroring or other schemes was regarded as the safest form of redundant data protection. What reenforced that view was the fact that traditional RAID systems using parity based protection are an all-or-nothing type of deal.
That is, with traditional parity based RAID systems, if the said systems are unable sustain and recover from a particular failure level then all data is lost. This chronically negative aspect of traditional RAID systems has lead to the doctrine of “RAID is not backup”.

Mirroring is simple.

Here is an example of a mirroring system:

And here is an example of a parity based system:

As you can see, it is much simpler to implement a mirroring scheme than a parity one. Mirroring involves simply copying the data to two disks where parity based schemes involve algorithms for computing the parity and for where to write the data and parity.

Looking beyond simplicity… understanding the importance of tolerance level

Yes, mirroring is simple an inherently easier to make robust. However, parity based systems are just as robust when properly implemented. If all schemes have the same robustness when properly implemented, then we need to focus on the real differentiators. The biggest factor when it comes to data protection is tolerance level, which is the number of failure the system can tolerate and successfully recover from.

Mirroring has a tolerance level of only 1

The typical mirroring system can only tolerate a single failure as there typically only exist two copies of the data. Lose both copies and you have lost it all.

Mirroring is expensive

Mirroring requires double the amount of space used to store the data. This turns, for instance, a $1k system into a $2k system, and things add up really quick.

Mirroring makes a poor choice when your data requires more than one disk for storage

Mirroring makes perfect sense if you only have two disks. Or rather, we should say that mirroring makes perfect sense if your data can fit on a single disk, which you mirror out for protection.

Once your data requires more than one disk for storage, mirroring becomes a poor choice for data protection.
Let’s go through an example and say you have 4 disks and that your data requires two of those disks for storage.
This leaves two extra disks for protection.

Your choices are:
– Two RAID 1 sets of two disks each
– A single RAID 6 volume
– A FlexRAID double parity array

Now, let’s say you lost two disks:
– If both disks are from the same RAID 1 set, you will lose data
– In the case of the RAID 6 array, you will recover from the double failure just fine
– In the case of double parity in FlexRAID, you will recover from the double failure just fine too

So, past a single set, mirroring becomes impractical or rather a poor choice.

Use this as a rule: if your data can fit on one disk, you should go with RAID 1 and mirror it. Else, go with a parity based RAID that lets you lose any X disks (where X is the tolerance level).

Evolution in data protection

Parity based schemes can be extremely cost efficient. For instance, if you have a million 2TB drives, it only takes one additional 2TB drive to provide protection to all those 1 million drives.
For all the benefits that parity based schemes provide, their core drawback has always been and will remain a hard pill to swallow: you lose all your data if things go wrong.

logo
We have evolved, and now comes FlexRAID to the rescue.
FlexRAID gives you the benefits of parity based schemes without the risk of losing all of one’s data unless you lose every single one of your drives. That is, if something was to go very wrong in FlexRAID, your data loss would be limited to only the failed disks as the data on the surviving disks would remain intact and would be readable and writable.
Effectively, the worst case scenario in FlexRAID is a minimized loss as opposed to a complete array loss as when using standard RAID.

FlexRAID allows you to choose and pick your protection level going from 1 to infinity.
That’s right, your protection level under FlexRAID is limited only by your choice and by your own hardware.
Once the protection level is chosen, any X disk can be recovered in the event of disk failure.

Conclusion

Do not waste your time with any product that relies on duplication as a protection scheme for your precious data.
Great technological progresses have been made and you should no longer be doing things the hard and expensive way.
When it comes to data protection, do it the FlexRAID way. :)