Behind the Scenes at AWS – DynamoDB UpdateTable Speedup
![]() |
We frequently discuss in regards to the Tempo of Innovation at AWS, and share the outcomes on this weblog, within the AWS What’s New web page, and in our weekly AWS on Air streams. As we speak I wish to speak about a barely completely different sort of innovation, the type that occurs behind the scenes.
Every AWS buyer makes use of a special mixture of companies, and makes use of these companies in distinctive methods. Each service is instrumented and monitored, and the workforce accountable for designing, constructing, operating, scaling, and evolving the service pays steady consideration to all the ensuing metrics. The metrics present insights into how the service is getting used, the way it performs underneath load, and in lots of circumstances highlights areas for optimization in pursuit of upper availability, higher efficiency, and decrease prices.
As soon as an space for enchancment has been recognized, a plan is put in to put, modifications are made and examined in pre-production environments, then deployed to a number of AWS areas. This occurs routinely, and (thus far) with out fanfare. Every a part of AWS will get higher and higher, with no motion in your half.
DynamoDB UpdateTableIn late 2021 we introduced the Commonplace-Rare Entry desk class for Amazon DynamoDB. As Marcia famous in her publish, utilizing this class can scale back your storage prices by 60% in comparison with the prevailing (Commonplace) class. She additionally confirmed you ways you can modify a desk to make use of the brand new class. The modification operation calls the
UpdateTable
perform, and that perform is the subject of this publish!
As is the case with nearly each AWS launch, clients started to utilize the brand new desk class immediately. They created new tables and modified current ones, benefiting from the decrease pricing as quickly because the modification was full.
DynamoDB makes use of a extremely distributed storage structure. Every desk is cut up into a number of partitions; operations comparable to altering the storage class are finished in parallel throughout the partitions. After taking a look at quite a lot of metrics, the DynamoDB workforce discovered methods to extend parallelism and to scale back the period of time spent managing the parallel operations.
This transformation had a dramatic impact for Amazon DynamoDB tables over 500 GB in measurement, decreasing the time to replace the desk class by as much as 97%.
Every time we make a change like this, we seize the “earlier than” and “after” metrics, and share the outcomes internally in order that different groups can study from the expertise whereas they’re within the course of of creating comparable enhancements of their very own. Even higher, every change that we make opens the door to different ones, making a optimistic suggestions loop that (as soon as once more) advantages everybody that makes use of a specific service or function.
Each DynamoDB consumer can benefit from this elevated efficiency immediately with out the necessity for a model improve or downtime for upkeep (DynamoDB doesn’t even have upkeep home windows).
Incremental efficiency and operational enhancements like this one are finished routinely and with out a lot fanfare. Nevertheless it’s at all times good to listen to again from our clients when their very own measurements point out that some a part of AWS turned higher or quicker.
Management Rules
As I used to be interested by this variation whereas on the point of write this publish, a number of Amazon Leadership Principles got here to thoughts. The DynamoDB workforce confirmed Buyer Obsession by implementing a change that may profit any DynamoDB consumer with tables over 500 GB in measurement. To do that they needed to Invent and Simplify, developing with a greater strategy to implement the UpdateTable
perform.
Whilst you, as an AWS buyer, get the advantages with no motion wanted in your half, this doesn’t imply that it’s a must to wait till we determine to pay particular consideration to your explicit use case. In case you are pushing any side of AWS to the restrict (or wish to), I like to recommend that you just make contact with the suitable service workforce and allow them to know what’s occurring. You is likely to be operating right into a quota or different restrict, or pushing bandwidth, reminiscence, or different sources to extremes. Regardless of the case, the workforce would love to listen to from you!
Keep Tuned
I’ve a protracted checklist of different inner enhancements that we’ve got made, and will probably be working with the groups to share extra of them all year long.
— Jeff;