How #AWS #Lambda handles failures: automatic batch retries for stream-based event sources, two automatic retries for asynchronously invoked functions (then delivery to a “Dead Letter Queue”), a 429 error for synchronous invocation
on 02017-05-29Chat on HN about using #AWS #Lambda for a “serverless” webapp platform called “JAWS”.
on 02015-08-17Werner Vogels announcing #AWS #Lambda.
on 02015-08-17HN discussion of #AWS #Lambda from when it was introduced in 2014-11. Initially it was #Node only. Mentions iron.io as an alternative.
on 02015-08-17#AWS #Lambda functions can be up to 50MB compressed, get 500MB of /tmp, can open TCP/IP sockets (except to TCP port 25!), have a minimum of 128MB RAM, and can run up for up to 60”, though the default kill timeout is 3”. They talk about “thousands [of requests] per second”. Default concurrency limit is 100Hz and 100 concurrent executions per region.
on 02015-08-17you can run #Node #JS and #Python code in #AWS #Lambda sandboxes, not just JVM stuff!
on 02015-08-17Hmm, so now there's an #AWS resource to compute things with a billing quantum of less than a CPU-hour, and without having to wait for Linux to boot. “Billing is metered in increments of 100 milliseconds” ☺ US$0.20 per million requests, plus US$0.00001667 per GB-second, although the lowest price they mention on the page is for using 128MB. This works out to US$43.84 per gigabyte-month. #AWS #Lambda
on 02015-08-17