Live data from Hacker News

Serverless Map/Reduce

tothestars.io

111–120 of 161 posts

Re: Serverless Map/Reduce

#111
post #57

I do not agree with term serverless. Amazon Lambda is a service, therefore there is a server involved. It's like saying deathless meat, because someone else killed the animal you are consuming.

Exactly. The cloud is just someone else's computer. It's a genuine business case, as running a lean and secure server farm is no child's play, but it's also oversold so much that my eyes hurts. Actual serverless computing would be implemented as p2p. Possibly interesting in a typical EvilCorpesque large corporation. Besides, Seti@home (Boinc) have been doing serverless computing for almost 20 years and it wasn't the…

P2P would be using peer computers as servers, so it wouldn't be serverless by your definition.

Re: Serverless Map/Reduce

#112

Earlier quoted context omitted.

With AWS API gateway + lambda + dynamodb, I maintain exactly zero servers. Zero physical servers. Zero server OSes. Zero sysadmin. It's perfectly reasonable IMO for that to be called "serverless".

With my car dealer, I maintain exactly zero factories. Zero physical factories. Zero factory workers. It's perfectly reasonable IMO for that to be called "factory-less car".

If originally you had to precure time at a factory to produce your car and now you can go to a dealer and buy a car, I would call that new way factory-less — the factory has been taken out of your personal car buying equation.

Re: Serverless Map/Reduce

#113

Earlier quoted context omitted.

It's less the total size of the data I'm worried about and more the concurrency. For example, say I had a process that retrieved 1000 tiny records (using index query) and ran some cpu-intensive calculation on them, and I wanted to run 1000 of those processes simultaneously to reduce into a final result. This would require tuning dynamo to thousands of concurrent reads (and maybe writes, depending on the process), the…

Just use RDS and S^3 for the blobs. RDS can do tens of thousands of index lookups a second. If you only need one index, then just name your s3 document by the compound index value and call it a day. Otherwise, just use RDS for everything.

From the RDS FAQ:

> In order to maximize your workload’s throughput on Amazon Aurora, we recommend building your applications to drive a large number of concurrent queries.

Perrrrfect. Thanks!

https://aws.amazon.com/rds/aurora/faqs/

Re: Serverless Map/Reduce

#114
post #77

I wonder if Amazon will ever open Lambda up to any Docker image? (I know it's possible to run binaries, but its a bit of a pain to compile with the Amazon AMI, etc.) Being able to have a bunch of `docker run` with any image would be pretty powerful.

Check out Hyper. https://www.hyper.sh/ "With HyperContainer, we make the performance/overhead of virtualized container similar to linux container --- 130ms launch time, and sharing the read-only part of memory (kernel and init)." -gnawux https://news.ycombinator.com/item?id=12873089

Yep, and "hyper func" in the roadmap

Re: Serverless Map/Reduce

#115

Earlier quoted context omitted.

Yes, I also think that's a much more accurate term that better represents the concept.

Great. We now know what color to paint the bike shed. Hoorah, progress!

Yep, that's how progress is made, one step at a time.

As seen in: human history, evolution, etc.

Re: Serverless Map/Reduce

#116

Earlier quoted context omitted.

With my supermarket, each time I buy a chicken I maintain zero farms. Zero physical farms. Zero farmers. It's perfectly reasonable IMO for that to be called "farmless chickens".

It's just a term. Don't worry about it.

Don't worry about roman numbers! It's just a representation.

Don't worry about them. Just keep doing your math in roman numbers. It's fine.

It bothers me that some people don't like my roman numbers. They work! Look!!

I + I = II

MMMCMXCIX - CXXIII = MMMDCCCLXXVI

There's absolutely no way of making this better through representation! Anyone who says otherwise is a bike shedder!

In fact, also don't worry about words either. Instead of serverless, let's call them catgiraffetablechaircthulhu. It doesn't matter!

Re: Serverless Map/Reduce

#117

Earlier quoted context omitted.

With my car dealer, I maintain exactly zero factories. Zero physical factories. Zero factory workers. It's perfectly reasonable IMO for that to be called "factory-less car".

If originally you had to precure time at a factory to produce your car and now you can go to a dealer and buy a car, I would call that new way factory-less — the factory has been taken out of your personal car buying equation.

So I am not paying for the factory infrastructure, workers, etc. when I buy my car? wow! that's great. We've came a long way with these factory-less cars!

Re: Serverless Map/Reduce

#118
post #50

OT: I teach computational methods and even as much as I dislike teaching/conflating it with web dev, I have included "let's build a web app" because students like building and deploying a thing, and because Heroku has a free tier. I've considered the possibility of having students do things on AWS (beyond web dev), including Lambda, and just expensing the costs. It seems feasible to quickly set up every student with…

AWS has a free tier for a year

It becomes a problem when you're a student and people expect you to use AWS's free tier for different projects in two different years.

Re: Serverless Map/Reduce

#119
post #26
post #25

Earlier quoted context omitted.

Some are, some aren't. For example, my experience has been that it's harder to get the 100 S3 bucket-per-account limit raised.

That used to be a hard technological limitation but they fixed it a few months back. You should be able to get more now (but you have to have a good justification for it).

What would be a good justification? Is there anything you can do with 101 buckets that you can't do with folders one bucket?

Re: Serverless Map/Reduce

#120
Note: the underlying comparison to other systems is from a 2014 blogpost [1] which suggest they used the m2.4xlarge series of EC2 VMs (which were Nehalem class parts from 2010). Nehalem vs Haswell or Broadwell (the likely parts underlying Lambda) is a pretty big jump.

Disclosure: I work on Google Cloud, but I'm just pointing out a fact ;).

[1] https://amplab.cs.berkeley.edu/benchmark/

Post reply on HN