Practically Speaking about Cloud Storage – Part 2
The marketplace for cloud storage is getting more and more crowded each and every day. Whether it is targeted towards individuals, SMBs or Enterprises more new players are trying to offer service value.
As in the case with Dropbox. There are at least a dozen others in the limelight. They are: Box.net, SugarSync, Apple Mobileme, Carbonite, ZumoDrive, iBackup, LiveDrive, Mozy etc… For the business and enterprise section, in addition to Amazon S3, there are Nirvanix, MozyPro, Parascale ( now with HDS), Vembu, Caringo, Microsoft Windows Azure, IronMountain, EMC Atmos etc.. Practically end users have to make their choice depending on what value they get from these service providers.
Amazon S3 and Amazon RRS: Amazon S3 or Simple Storage Service is a pioneer offering with its full set of features turned towards web oriented enterprises. It was launched in the year 2006 with a bold statement as ‘Storage for the Internet’. And Amazon RRS or Reduced Redundancy Storage is a choice within S3 for storing secondary data storage.
We can take up five key components of Amazon S3 service:
Bucket : At the highest level of storage, buckets contain objects in S3.
Objects: Objects are the fundamental entity of storage. Inside Bucket, objects are identified with keys and versioning
Keys: Key is the identifier of the object.
Regions: US Standard, US ( North California), EU (Ireland) and APAC (Singapore)
Data Consistency Model: The updates are made atomic; when loading data, the retrieval brings either the old data or new data.
S3 Service Level Agreement: As of now, Amazon S3 stores more than one hundred billion objects with about a hundred thousand operations per second. To know the fine print of these exclusions, one has to not only look at the SLA of S3, but also AWS Agreement. The Amazon S3 SLA or Service Level Agreement provides us with a service commitment of 99.9 % calculated for a monthly billing cycle. There are three basic terms defined in the SLA. They are Error Rate, Monthly Uptime Percentage and Service Credit. For a given billing month, if the service commitment goes below 99.9% up to 99%, a 10% service credit percentage is given; if it goes below 99%, a 25% service credit percentage is given. Service Credit is defined as the payment of the month. The service credit given as part of the SLA, is adjustable in the subsequent month bills. There are exclusions for the calculation of un-availability of service.
S3 Pricing: Amazon S3 pricing slightly varies between different regions. Pricing is based on how much is stored and how much is accessed. We have seen the service availability term in the SLA. Another term that is added here in the pricing section is durability. Durability can be simply explained as a period, stored object is available after a period of one year. Supposing let us say that we mention as 100% durability, that means that object will be available for access 100% or always; 80% durability means that there is a chance of that object getting lost one in five times. For the S3 the durability is 99.999999999 (99 followed by nine 9 s in point) and for RRS it is 99.99%. Practically speaking we need not worry about this durability as if we store around 1000 objects – with this percentage it will not be lost for several hundred years!
The storage cost per month per GB is starting at $ 0.15 for about 50 TB/ Month; this cost gets lesser for higher TB/ Month. All data transfer is free until Nov. 2010. RRS or Reduced Redundancy Storage is about 30% lesser in cost.
From the grapevine blogs in the web, we can say that Amazon gets revenue of about $ 60 million a year from S3 Service compared to $ 260 million from EC2 Service.
By Glenn Blake