Cloud economics: Measuring up the Indian cloud threat

Last month, Indian IT outsourcing giant Tata announced the launch of InstaCompute, the Indian answer to Amazon's EC2 cloud. 

Backed by an industry giant and hosted at low cost in data centres located in India and Singapore, the arrival of InstaCompute has no doubt set off a few shockwaves through the ranks of American giants Amazon and Microsoft, who have spent a great deal of capital betting their businesses on offering public cloud computing services.

But is the Indian giant going to be that much cheaper than its American rivals? Today we look across a broad spectrum of options [all in US dollars for ease of comparison] to find out.

At the low-end

On face value, those fears are understandable - Tata's InstaCompute offers entry level pricing of 5 cents per hour for an unlicensed instance - well under Amazon's 8.5 cents for its "small instance" on a Unix/Linux operating system. 

But to deliver that, Tata had to shave off a few features that come standard with Amazon. Tata's entry instance runs on 1GB memory with 20GB disk space while Amazon offers 1.7 GB memory and includes 160GB of "instance" storage.

Tata's Windows licensed instance meanwhile is on par with Amazon at 12 cents per hour, but again offers less storage. 

Another US-based challenger, Terremark, has also pared back Amazon's small instance for its VMware-based vCloud, starting at 3.6 cents per hour, but the price is based on a single virtual processor with 512 MB memory and no storage. 

Speaking of storage

Amazon also trumps the Indian newcomer when it comes to paying for data storage, Amazon coming in at 10 cents per GB and Tata at 15 cents per GB.

Microsoft's Azure cloud platform, which doesn't offer unlicensed or Unix/Linux pricing, costs the same as Amazon's Windows instance. But Microsoft has bumped up its pre-packaged disk space to 225GB (versus 160GB on Amazon). 

Exactly how much "instance storage" matters to the average punter is debatable. Amazon described it as "ephemeral" storage that will disappear if an instance is deleted or if there is a failure. 

Persistent storage, such as Azure Storage and Amazon's Simple Storage Solution "S3", would be required for most cloud applications and presents an additional cost. (See storage cost table below).

 At the big end of town

At the big end of town, requirements for additional storage, memory and processors drives a wedge further between Tata and Amazon's pricing.

Amazon's "large instance" includes eight processors, 15GB memory and 1.69TB instance storage. At 68 cents per hour for a Unix/Linux system, it costs around 30 percent less than Tata.  

At 100 percent utilisation, a customer running a basic large Linux/Unix instance would face a $490 monthly bill from Amazon compared with $712 at Tata. 

A similar difference is noted when you compare each provider's Windows system instances. Amazon came in at $691 per month while Tata was $852. 

Tata, however, is also still fleshing out its higher end computing capabilities. While it has outlined pricing. the company is yet to launch enterprise products, and no doubt it will look to Amazon as a yardstick for pricing just as it has at the small end of town. 

The one important variable where Tata did undercut Amazon was (Asia Pacific) data transfer.

From November 1, Amazon will be charging 10 cents per GB for inbound data, while Tata is not charging for it. Tata also charged 17 cents per GB for outbound data, compared to Amazon's 19 cents after the 1GB per month free allocation has been used up. 

Amazon still cheaper

Tata's target is definitely the low-end of the market - configuring an entry level computing package that is $20 per month less than Amazon.

But whether Tata's edge is maintained depends on storage and memory requirements. If storage is bumped up to Amazon's 160GB with comparable memory, a customer would end up paying $30 more to Tata. At the big end of town, Tata's pricing edge eroded entirely, with Amazon offering more storage, processors and memory for at least $200 per month less than Tata.  

That said, public cloud is one of many enterprise computing options - and some analysts suggest that for many applications, a public cloud is not only riskier but also more expensive than deploying servers in-house.

See Part II for a discussion on this topic.