Chris Mellor from The Register (http://theregister.co.uk) recently wrote an article about cloud storage starting to savage the market share of traditional/legacy storage vendors. Link to the article is here.
It is predicting that due to the economies of scale and custom built hardware, out of commodity parts, cloud providers will be able to provide far cheaper storage than any traditional array vendor. An estimated market value graph was included, with cloud intersecting legacy around 2024.
- Cloud Storage v Legacy Storage
So how do businesses start to embrace and utilise this cost effective storage, which is protected with multiple copies and redundancies, better than any legacy storage array located in a standard data centre? All this with making sure data is obfuscated for compliance and security reasons.
One option is to re-architect your applications and use RESTful APIs to send, fetch and modify data within a cloud provider…and this works extremely well for a lot of business (Netflix are a great example). But is this approach for everyone? I’d say probably not at this stage, not many companies can put the development and DevOps effort into an application to ensure it’s robustness like Netflix can.
The other option, which is one of the things I’m focussed on with Microsoft, is a storage array that resides on premise and can integrate with the cloud to provide the best of both worlds. Provide local, enterprise class storage, to your existing servers and applications, and be able to utilise the efficiencies of the cloud without having to re-write your applications. The local storage array is StorSimple, which was acquired by Microsoft in October 2012; more info can be found here http://www.storsimple.com/.
Suddenly you have enterprise on premise arrays with the below features on premise
- auto tiering
- flash optimised (as a tier)
- dedupe of production data
- compression of production data
- application integration
- snapshots without performance impacts
Then, as blocks of data grow cold, encrypts this data and migrates it, at a block level to Azure. This means you could have constantly accessed VM hot blocks on SSD, warm blocks on SAS and then the cold blocks residing on Azure. With more than 80% of data generally being cold data this suddenly makes a lot of sense.
Not only does this remove a lot of cost around storage and complexity around archiving it also opens another door about how you can mange your backup. Once this data resides on StorSimple it is possible to take a “cloud snapshot”. A full copy of the data (encrypted and optimised) is taken to Azure on the first snap…after this only changed data, at an optimised block level, is copied to Azure for the next snapshot. Suddenly you can keep hourly, daily, monthly and even yearly backups/archives within Azure and significantly reduce your backup window, operational expenditure and effort around backup, as well as increase your ability to restore data quickly and efficiently.
Finally, from a compliance perspective, all this data that sits within Azure is encrypted (256 bit AES encryption), three copies of the data are made in the primary data centre and a fourth copy is made to the Azure partner data centre. This is the case for the cloud snapshots and for any data tiered to Azure.
I think it is pretty cool tech and will be writing more on StorSimple and Azure in the future.