Who Is Hosting This

Even as Google enters the cloud storage fray, Amazon, the industry stalwart and runaway leader, is not taking the matter lying down. Even before Google could finish its announcement Amazon made one of its own, saying that it was offering a new form of storage for its S3 product, this one cheaper though, by design, less durable.

The idea is something of a strange move. In an era where people are questioning the safety and reliability of cloud storage services, like Amazon S3, what Amazon has done is create a new tier of redundancy, one that is less reliable, designed to be cheaper for services that don’t need high levels of reliability.

The goal is to create two tiers of data, the first kind requires the highest level of backup and security, the second involves less risk if it is lost or somehow destroyed. The first tier pays full price, the second tier doesn’t.

Is it an idea one might be interested in? It really depends on the data one is trying to host.

A Word on Durability

According to Amazon, their regular S3 storage has a durability of 99.999999% or “11 Nines”. This means, by their math, if you host some 10,000 objects with them they might lose one in every ten million years or so. In short, for all practical intents and purposes, your data is safe on those servers and, according to Amazon, their system can survive a complete, simultaneous failure at two locations.

By comparison, Amazon’s Reduced Redundancy storage offers a durability of a “mere” 99.99%. This means the same account with 10,000 items might lose one item every year. Still an extremely low failure rate, but one that pales in comparison to Amazon’s main service.

The question then becomes, what data belongs in which tier? The answer depends on how long the data is expected to stay on the server and how important it is that it not disappear. Data that is merely transitory and can be easily recreated from other sources definitely belongs on the reduced redundancy tier while long-term data that can’t be trivially replaced belongs on the main tier.

So, for example, your Twitter avatars and other photos that you should have backed up elsewhere are fine on the reduced redundancy service where your will or accounting documents would be better serviced with the highest level of backup.

Durability is Relative

One of the classic uses of Amazon S3 is for backups of local data, such as a hard drive. So which tier would this likely fall under? The answer is, most likely, the reduced redundancy tier.

The reason is that 99.99% durability is still higher than just about any commercially available hard drive and the odds of both a hard drive and an Amazon S3 back up failing, even with the reduced redundancy, are extremely slim.

In fact, even if we assume that a hard drive has an extremely low durability rate of 10%, the odds of losing BOTH your hard drive data and your S3 back up is .001%. This means that, combined, they have a durability of 99.999% or 5 nines. This is much higher than what is practically necessary for a backup system and it easily beats other common methods, such as external hard drives.

But even if Amazon does eat your most recent backup, you can always go back to your previous one, meaning almost no data is actually lost.

While it is true that 4 nines is less than 11, it is still more than what most applications will ever need.

Save Money With Reduced Redundancy

So what does one get by agreeing to have less redundancy on their S3 account? The cost of storage, though not transfer, is drastically reduced.

The average savings is approximately 33% of the price. At the lowest tier, for example, the price for storage drops from 15 cents per GB to 10 cents.

While these savings won’t be very large for Webmasters using S3 as a CDN for their small site, I would only save about 10 cents per month, those who run large-scale Web services, could clearly benefit from having their storage fees cut by a third. Backupify, for example, may well benefit from this.

However, anyone who uses their Amazon S3 accounts to backup their hard drive or store large volumes of data may want to look at this, especially if 99.99% durability is more than adequate for their purposes.

Bottom Line

The Amazon announcement had two very critical points:

1. The Stated Durability of the Main S3 Service: Amazon had been previously tight-lipped about exactly how reliable S3 was. Now they’ve gone on record with a stated durability of 11 nines.
2. Providing Choices on Durability: More importantly though, Amazon is now providing customers a choice on durability and letting them pay only for the redundancy they need. It is a smart move that will greatly benefit its customers, especially those who store many gigs of data.

The question, however, is whether many will switch down to the reduced redundancy service or be enticed to try Amazon S3 by it. The latter, sadly, seems to be untrue. Price has not been a major obstacle to using S3 for most but fear of reliability has so introducing a less reliable service may seem like an odd move.

Then again, this wasn’t a move aimed at the technically-illiterate who don’t understand how reliability works on the Web. Instead, it was aimed at developers of Web-based products built on S3 that may want to create services and offerings cheaper than before.

In the end, we will all be rewarded by cheaper services and better tools to work with but those benefits will, most likely, be at least some time away.