Does AWS S3 Intelligent-Tiering save money? Spoiler alert: usually, it does!
Firstly, what is it? Just announced by AWS, S3 Intelligent-Tiering is a new storage class. It delivers automatic cost savings by moving data between two tiers (frequent and infrequent access) based on access patterns.
For a monitoring and automation fee (based on number of objects), the new offering checks for access patterns. It then moves objects you haven’t accessed for 30 consecutive days to the infrequent access tier and back.
Importantly, the S3 Intelligent-Tiering storage class has no retrieval fees. If you access an object in the infrequent access tier, it is automatically moved back to the frequent access tier. No additional tiering and data retrieval fees apply to retrieve or move objects between tiers.
So how much money can an enterprise save by moving to S3 Intelligent-Tiering?
For purpose of the following analysis, let’s assume:
- We host 1 petabyte of storage composed of 1 billion objects;
- The average object size is 1MB; and
- The amount of “hot” data (i.e. frequently accessed data) is 10%
We’ll also assume that, on average we:
- Access hot objects 4 times a month;
- Modify hot objects once per two months;
- Access cold objects 0.05 times per month (that is, once every 20 months); and
- Never modify cold objects
Based on these assumptions, each month:
- Every month we read hot objects 400 million times;
- Every month we read cold objects 45 million times (for a total of 445 million read requests); and
- In addition, we perform 50 million object updates every month
(Note: for the sake of simplicity, we’ll ignore the slightly higher price for the first 500 TB).
Let’s do the math!
Now, let’s calculate the monthly costs for our system, utilizing information from Amazon’s Cloud Storage Pricing Page. Does S3 Intelligent-Tiering provide cost savings?
So, is S3 Intelligent-Tiering for me?
In our sample scenario, S3 Intelligent-Tiering beats both “Standard” and “Infrequent Access” by a nice margin.
If you use the “Standard” storage class…
The Monitoring & Automation fee becomes more negligible as your average object size grows larger. So, as a rule of thumb, if you keep large volumes of data with an average object size of 1MB or larger, and most of your data is “cold,” you should probably use S3 Intelligent-Tiering. In our example, this saves you a cool 23 percent in your storage costs – an amazing $5,000 per month.
If you keep your data in the “Infrequent Access” class…
You’re still likely to save money by moving to tiering. In our example you would have saved 7.6 percent in your storage costs, or over $1,000 a month. These savings are more modest, and in some rare cases, the Monitoring and Automation fees may result in somewhat higher costs than “Infrequent Access.” However, you shouldn’t ignore the risk that’s embedded in the Infrequent Access pricing model. Namely, exorbitant retrieval fees should you need to retrieve a large percentage of your data due to unexpected circumstances.
The most attractive property of S3 Intelligent-Tiering compared to previous offerings from AWS is that the pricing model is simple and the fees are easily predictable. This lets you sleep better at night, and who wouldn’t want that?