Comparing Serverless Performance for CPU Bound Tasks

#1

Post by bosdev » Tue, Jul 10, 2018, 4:03 PM UTC

Comparing Serverless Performance for CPU Bound Tasks
blog.cloudflare.com

Re: Comparing Serverless Performance for CPU Bound Tasks

#2

Post by com2kid » Tue, Jul 10, 2018, 4:52 PM UTC

I'd love to see an honest comparison across other providers, throwing in Google's Firebase Functions and Azure Cloud Functions.

Re: Comparing Serverless Performance for CPU Bound Tasks

#3

Post by handruin » Tue, Jul 10, 2018, 4:57 PM UTC

I've recently been exploring AWS Lambda in a stack which contained API Gateway + Python Flask under Lambda for a task I was working on. I deployed it using Zappa and its purpose was to be a simple REST frontend for transferring files to S3.

After experimenting with uploads from Lambda to S3 I was noticing that the time to upload a tiny 4MB file changed dramatically when I reconfigured the Lambda function's memory size. At 500MB it took 16 seconds to upload the file which is pretty slow. Once I got past roughly 1500MB of memory, the performance no longer improved and the best I could get was about 8 seconds for the same payload.

None of my tests were controlled or rigorous in any way so take them with a grain of salt...they were just surprising to me that the speed changed dramatically with memory size allocation. I'm new to Lambda so I wasn't ware that memory size is tied to other resource performance. I'm curious if this goes beyond CPU and also changes network bandwidth/performance? The Lambda I deployed did not write data to the temp location that is provided, it streamed directly to S3.

I've since moved on from this implementation and now my Lambda function performs a much simpler task of generating pre-signed S3 URLs. I have noticed something else about Lambda that bothers me a little. If my function remains idle for some period of time and then I invoke it, the amount of time it takes to execute is around 800ms-1000ms. If I perform numerous calls right after, I get billed the minimum of 100ms because the execution time is under that. The part that bothers me is I'm being charged a one-time cost that's about 8x-10x the normal amount because my function has gone idle and cold. I'll have to continue reading to see if this is expected. It's not a huge amount in terms of cost but surprising that I'm paying for AWS to wake up from whatever state it is in.

Re: Comparing Serverless Performance for CPU Bound Tasks

#4

Post by sreque » Tue, Jul 10, 2018, 4:59 PM UTC

Assuming the author's tests are single-thread, I'm pretty sure 1024 MB doesn't give you a full CPU core on Lambda. I could be wrong though; I haven't payed attention to Lambda in a long time. Last I remember it was 1.5 GB that gave you a full core. This alone makes the comparison between a mid-range server and Lambda unfair, not to mention the differences between language runtimes.

That said, if you are using Lambda and expecting to not pay extra you have somehow been mislead. Lambda is definitely more expensive per cycle than managing your own instances, and I doubt that will change any time soon.

Re: Comparing Serverless Performance for CPU Bound Tasks

#5

Post by djhworld » Tue, Jul 10, 2018, 5:12 PM UTC

I remember a few years ago we tried to implement a scheduled Lambda that needed to download a bunch of files from an S3 prefix, perform some aggregation on the data and then write the result to a database.

Our EC2 prototype of this on one of the m3 class instances could do the work in about 2 minutes which seemed a perfect opportunity to port to Lambda.

Even on the top memory instance at the time (1536mb), the job just couldn't finish, timing out after 5 minutes. The code was multi threaded, to parallelise the downloads, but not matter how much we tweaked this the Lambda would just never complete in time.

As you don't have visibility of the internal we didn't know whether this was due to CPU constraints (decompressing lots of GZIP streams), network saturation (downloading files from S3) or what.

In the end we gave up. Didn't have the time or resource to keep digging, and just pinned the problem on the use case we were trying to fit was against what Lamba is designed for

Not saying this is an indictment of Lambda, we use it in lots of places, with a lot of critical path code (ETL Pipelines).

Re: Comparing Serverless Performance for CPU Bound Tasks

#6

Post by wolf550e » Tue, Jul 10, 2018, 5:18 PM UTC

I'll copy from Twitter[1]:

@zackbloom @jgrahamc I can't find it in the docs on AWS site, but I've read that AWS Lambda scales CPU linearly until 1.5GB, then gives you 2nd thread/core and again scales linearly until 3GB. If your PBKDF2 was single threaded, Lambda bigger than 1.5GB is wasted.

11:12 AM - 9 Jul 2018

reply by blog post author[2]:

Replying to @ZTarantov @Cloudflare @jgrahamc I can't think of a way to test that within the Node code. The only option seems to be to update the C++ version (or some other language) to use multiple threads.

5:16 PM - 9 Jul 2018

1 - https://twitter.com/ZTarantov/status/1016384547364229120

2 - https://twitter.com/zackbloom/status/1016476314864312321

Re: Comparing Serverless Performance for CPU Bound Tasks

#7

Post by Dunedan » Tue, Jul 10, 2018, 5:22 PM UTC

wolf550e wrote: post #6

I'll copy from Twitter[1]: @zackbloom @jgrahamc I can't find it in the docs on AWS site, but I've read that AWS Lambda scales CPU linearly until 1.5GB, then gives you 2nd thread/core and again scales linearly until 3GB. If your PBKDF2 was single threaded, Lambda bigger than 1.5GB is wasted. 11:12 AM - 9 Jul 2018 reply by blog post author[2]: Replying to @ZTarantov @Cloudflare @jgrahamc I can't think of a way to test…

Yes, Lambda functions use a second core above 1536MB of memory. Back in the past they had it in their documentation, but removed it at a certain point. Also see: https://stackoverflow.com/questions/34135359/whats-the-maxim...

Re: Comparing Serverless Performance for CPU Bound Tasks

#8

Post by gleenn » Tue, Jul 10, 2018, 5:22 PM UTC

djhworld wrote: post #5

I remember a few years ago we tried to implement a scheduled Lambda that needed to download a bunch of files from an S3 prefix, perform some aggregation on the data and then write the result to a database. Our EC2 prototype of this on one of the m3 class instances could do the work in about 2 minutes which seemed a perfect opportunity to port to Lambda. Even on the top memory instance at the time (1536mb), the job ju…

I thought the use case for things like Lambda were more along the lines of rarely used web requests that you'd save money on by not running a full box. I do remember them being slow too.

Re: Comparing Serverless Performance for CPU Bound Tasks

#9

Post by dsl » Tue, Jul 10, 2018, 5:24 PM UTC

The question is, what does Amazon know now, that Cloudflare will figure out in a year or so?

Re: Comparing Serverless Performance for CPU Bound Tasks

#10

Post by thinkmassive » Tue, Jul 10, 2018, 5:27 PM UTC

handruin wrote: post #3

I've recently been exploring AWS Lambda in a stack which contained API Gateway + Python Flask under Lambda for a task I was working on. I deployed it using Zappa and its purpose was to be a simple REST frontend for transferring files to S3. After experimenting with uploads from Lambda to S3 I was noticing that the time to upload a tiny 4MB file changed dramatically when I reconfigured the Lambda function's memory siz…

This is why people using Lambda at scale are concerned with keeping the containers “warm” https://aws.amazon.com/blogs/compute/container-reuse-in-lamb...