Twitter | Search | |
Alex DeBrie
Engineering Manager . AWS fan. Creator of . DM's open!
2,433
Tweets
531
Following
2,165
Followers
Tweets
Alex DeBrie 9h
Yep, makes sense. Ok, let me look into this and see what I can do!
Reply Retweet Like
Alex DeBrie 11h
Replying to @mthenw
What!!!! Yessss
Reply Retweet Like
Alex DeBrie retweeted
Serverless 12h
Take a look at some of the latest details around supported languages and runtimes with a post from !
Reply Retweet Like
Alex DeBrie 13h
Replying to @alexbdebrie
Reply Retweet Like
Alex DeBrie 13h
There are a couple overflow rooms for my Data Modeling with DynamoDB (CMY304) talk at reInvent -- come join! Will do an overview of DynamoDB and then a deep dive on a single-table data model example. A few different hotels available, so it will fit well wherever you are!
Reply Retweet Like
Alex DeBrie 13h
Hey Tom, just following up on this -- I saw there are some overflow rooms for my session. Full list here:
Reply Retweet Like
Alex DeBrie 16h
Agreed, we need something better there. Is it mostly in the functions block, with functions and events? Trying to see a good place to start.
Reply Retweet Like
Alex DeBrie Nov 20
The total package is still way cheaper, it's just when you look at the individual pieces and you're like .... wait, 70% of my spend is going *where* ? Makes you hope it could be even better 😬
Reply Retweet Like
Alex DeBrie Nov 20
Absolutely not! I mean, I'm still a huge user of APIGW + Lambda b/c I still think it's better overall. My big point is that 80% of Lambda web app users are way overserved by the APIGW offering and would be happier to use a service that is less feature-rich but much cheaper.
Reply Retweet Like
Alex DeBrie Nov 20
The crazy part is that APIGW is likely to be the biggest expense. Not uncommon to see APIGW + Lambda + DynamoDB service where APIGW is >70% of the total cost, which is nuts considering what it's doing compared to the rest of the stack.
Reply Retweet Like
Alex DeBrie Nov 19
It's all part of the rope-a-dope 😉
Reply Retweet Like
Alex DeBrie Nov 19
Somebody turn up the concurrency on the AWS blog! 😬
Reply Retweet Like
Alex DeBrie Nov 19
Ha! I hope it's less a battle than a discussion 😬 Rick's the master and the main reason I started to really love Dynamo so much.
Reply Retweet Like
Alex DeBrie Nov 19
Yep, and that's totally valid. That's the best-case scenario, when you have a fairly simple service that won't be changing. Use single-table, and it's probably not too complex. The rubber meets the road with a complex service that may be changing.
Reply Retweet Like
Alex DeBrie Nov 19
The tradeoff is around iteration speed -- will you need your data model to evolve and, if so, how hard will it be to evolve? You may decide that slower application performance (200 ms instead of 50 ms responses) is worth it for some faster iteration performance.
Reply Retweet Like
Alex DeBrie Nov 19
Totally agree, and I would expect single-table reads to be faster! The whole point of single-table is to make a single request to DDB, which will be faster than multiple, separate requests.
Reply Retweet Like
Alex DeBrie Nov 19
Absolutely! Will try to find you at reInvent as well
Reply Retweet Like
Alex DeBrie Nov 19
Let's do it!
Reply Retweet Like
Alex DeBrie Nov 19
Yep! Just use the new runtime label in your runtime property: runtime: java11 runtime: python3.8 runtime: nodejs12.x
Reply Retweet Like
Alex DeBrie Nov 19
Agree generally with this, and people should at least learn and understand the principles behind single-table design. For some folks, it makes sense to deviate from the pure best practices that are recommended for very high-scale applications. Just know what you're getting into!
Reply Retweet Like