:Ben Metcalfe Blog

Amazon S3 cost savings and the future of utility computing services

Amazon have been doing some amazing things in the utility computing sector with S3 (storage) and EC2 (virtual servers).

It’s touted as being an economical platform way to run your start-up/company/website/whatever from as it makes use of the space capacity Amazon owns from it’s e-commerce platform (so, er, what happens over Christmas when that platform is at it’s peak?).

I think it very much depends on what you intend to use EC2 and S3 for, but photo sharing site Smugmug has been a champion of S3 for some time – claiming it’s saving them over $500k in the first year. At this point in time they only use Amazon S3 for storage of the images

Michael Arrington called them on their numbers during Web2.0 Summit, and so the Smugmuggers have come back with some real numbers on their blog. The conclusions are:

  • Total amount NOT spent over the last 7 months: $423,686 [by not buying IDE disks, RAID controllers and single CPU servers for their SAN]
  • Total amount spent on S3: $84,255.25
  • Total savings: $339,430.75
  • That works out to $48,490 / month, which is $581,881 per year. Remember, though, our rate of growth is high, so over the remaining 5 months, the monthly savings will be even greater.

I was talking to an interesting start-up on Friday who were offering a product that could essentially be leveraged as a ‘commodity’/’utility’ computing product. I can’t really say what their vertical is but they were marketing it (some-what understandably) with a specific implementation so that it could be leveraged as a turn-key product. Their business plan was based on the percentage their product had on revenue.

I felt that by wrapping an interesting utility product into a single means of implementation they were shutting many doors to ways their product could be used in other areas. My overarching point to them was that what Amazon was really trailblazing here is the pricing model – and rather than charging based on the somewhat fuzzy impact it had on revenue (which could easily have been 0 or negative, not just positive – and potentially off putting IMHO as it requires companies who might be private to disclose revenues, etc) they could charge based on usage. $0.10 for 1000 calls to their service, etc.

EC2, Amazon’s flexible virtual server product, in many ways is even more fascinating. I’m not convinced it’s the cheapest way to run a server continually – not over a 12 month term at least. And its performance over the Christmas period is a bit of an unknown. But the ability to suddenly double or triple the number of instances of an application server, especially for a short time during a serving peak has a real definite value. The emergence of ways to automate the ramping-up and ramping-down of service over the course of a day/week/month cycle are particularly exciting.

It’s definitely an interesting area and I’m curious to see how the industry adopts it.

Your own utility computing service?

My final point on all this, I guess, is look to see whether you have a potential utility computing product in your inventory or as part of your start-up. Even if it’s not your company’s core product it might be something that’s part of your platform. Amazon is still a book and CD retailer after all, and it’s just utilizing affordences in their serving platform.

Anyway the point is not only that you can offer something, but that you can easily monetize it in perhaps the most scalable and lowest risk way – pay as you use. So start thinking whether you have something like gateways (eg api< ->SMS), data processing (sorting, ranking, etc) or database capacity.