Blogger social buttons

Wednesday, 22 August 2012

Cost Comparison Amazon Glacier vs S3

Experiment results: AWS Glacier storage is around 90% cheaper than S3 
and a good long term data-archive option 

Yesterday AWS announced their dedicated service for data archiving on the cloud. This new service, called Amazon Glacier, builds on the tremendous success of S3, which has been used to store over one trillion objects since it launched in 2006.

PlanForCloud supported Glacier from launch, and we're already seeing many of our users get cost forecasts for it. On average, each Glacier storage block, called a Vault in AWS terminology, is 3.2TB in PlanForCloud. That's just a monthly baseline; we're also seeing many users apply Growth Patterns to their Glacier storage to get more accurate and long term cost forecasts. One of the most popular patterns is to increase storage size by 10% per month.

Elasticity Growth Patterns in PlanForCloud
Using patterns to model 10% growth in PlanForCloud
Let's take this example and forecast costs of using AWS Glacier vs AWS S3 (Standard and Reduced Redundancy). Read more about AWS Glacier or S3 Standard/Reduced Redundancy storage services if you are not familiar with them. Note that Glacier's annual durability is 99.999999999% per-archive, which is the same as S3 Standard, but S3 Reduced Redundancy is only 99.99%. There are also other differences between these storage service, for example, Glacier has a retrieval time of around 4 hours whereas you can download data from S3 instantly.

For this cost comparison example, we only concentrate on costs for storage. The data transfer costs are exactly the same in both scenarios, so they have also been left out of the graphs.

The following graph shows the 3-year cost forecast for our example storage (3.2TB baseline and add 10% every month. i.e. first month we would use 3,520GB). The final results:

  • Amazon Glacier = $11K
  • S3 Reduced Redundancy = $86K
  • S3 Standard = $114K

That makes Glacier around 90% cheaper than S3 and a great long term data-archive option if the slow retrieval time is not an issue for you.

You can run this experiment for your own deployments on PlanForCloud.com,
no cloud credentials required and 100% free!

Amazon Glacier vs S3 results on PlanForCloud
Results of Amazon Glacier vs S3
Next blog post: AWS Reserved Instances vs On-Demand: Breakeven point

-- Ali

7 comments:

  1. ok if all there was to storage was cost per capacity, concur, glacier would be cheaper however that is like comparing a hard disk drive (HDD) to SSD, or even tape to HDD.

    For the target use cases of glacier (e.g. inactive, archive, reference data, etc) concur that the lower cost is a step in the right direction with room for improvement.

    However rest assured as the sun rises in the east and sets in the west, its not if, rather when or how soon somebody jumps into glacier and then complains it is too slow for their videos etc...

    Cheers gs

    ReplyDelete
    Replies
    1. Thanks for your comment Greg,
      Agree with you that Glacier would be used for a different business purpose, but it is also a valid comparison for enterprises who have used S3 as their archiving solution up to now (since that's what was available on AWS).

      Cheers,
      Hassan

      Delete
    2. Apt name - Glacier. The data retrieval process is as slow moving as a "glacier."

      Delete
  2. im not sure how can we make money out of cloud. can anyone here explain to me. cloud is very expensive, when you turn off your cloud server, you only pay for the hard disk. i hope there's some group of wealthy individuals out there that are willing to spend hundreds of dollars so that people like me can work as a salesman/reseller for their cloud. if i only have my own cloud resources, i would sell windows and linux vps. every reseller program that i checked is kinda rip off. i can't imagine how we can profit from doing that.

    ReplyDelete
  3. Cloud provides you Amazing scalability, flexibility and Performance and ease of Use. There are many ways you can capitalize from cloud computing. Gary B (Netset Software)

    ReplyDelete
  4. Ali, this article is nice, but calculating just the storage cost is relatively simple.
    However, Glacier seems to have a really confusing method of charging you for bult restores. Just see this question https://aws.amazon.com/glacier/faqs/#How_will_I_be_charged_when_retrieving_large_amounts_of_data_from_Amazon_Glacier
    If PlanCloud could somehow simplify this, that would be awesome.

    ReplyDelete
    Replies
    1. Thanks for the feedback! We can currently have Glacier's prices for data storage and data upload/retrieval *requests* but not the data retrieval itself, which is shown as free on AWS with a note saying:
      "Glacier is designed with the expectation that retrievals are infrequent and unusual, and data will be stored for extended periods of time. You can retrieve up to 5% of your average monthly storage (pro-rated daily) for free each month. If you choose to retrieve more than this amount of data in a month, you are charged a retrieval fee starting at $0.01 per gigabyte. In addition, there is a pro-rated charge of $0.03 per gigabyte for items deleted prior to 90 days."

      We are thinking about how to include these "extra charges" in PlanForCloud deployments such that users understand what they are and how significant they can be to the 3-year cost forecasts. What are your thoughts on this?

      Delete