Hacker News Reader

Comment by johansch | Hacker News Reader

I don't understand why Amazon doesn't launch a clone of App Engine. This is the holy grail in NoOps. AWS Lambda + API Gateway is a bureaucratic exercise in pain, true to AWS form, while App Engine is a pleasure to use. Anyone who have used both services will understand me.

It seems to me (from reading HN) like most people are avoiding using App Engine because Google tortured and killed their pet, err, I mean cancelled Google Reader three years ago.

(And I guess, the inference is that because Google killed a tiny unprofitable service once in the past, you can not realistically depend on them to continue to provide services they are actually putting real money into because it is of strategical interest. Yeah, that makes total brogrammer sense, let's go with that line of thinking.)

View on HN

I've been using App Engine for many projects since 2008. Like any tool, it's not the right tool for every job, but for its intended use cases, it's great. Being able to click Deploy, and not have to deal with sys admin is really convenient.

The one area where App Engine falls a bit short is lack of support for some widely used libraries like Numpy. It would be nice if The Google would add support for those (support for some transcoding libraries would be nice as well). Even better would be an interface to TensorFlow.

I've also been using GAE daily since 2008.

Brian - the new "Flexible Environments" (aka Docker aka Managed VMs) give you Numpy, python3, native libraries and more. I took my latest project into production with it, and it's been pretty painless.

https://groups.google.com/forum/#!searchin/google-appengine/...

G's habit of killing services doesn't explain the extent of AWS's dominance. There must be some way that AWS is actually superior for popular uses.

Or maybe Silicon Valley is a fiscally incompetent echo chamber, because, well, they don't tend to worry much about money until it's too late.

Let's stipulate that it is. That's still not a complete explanation.

Well, they had a pretty big head start. And I do remember not understanding why people were going for AWS 4-5 years ago instead of self-hosted clouds of rented actual servers at 1/5 the cost.

Besides that, when starting last summer to evaluate both services, I didn't find much that was in the benefit of AWS, compared to GC.

(Besides AWS China. GC does not currently operate in China. I really, really want them to fix this.)

Oh, and in case you wanted an explanation: it's called groupthink and echo chamber. In (the more frugal, because less VC) Europe we were all amazed at how much american startups would pay for hosting.

For the kinds of things medium-big spenders spend money on (mostly, largish amounts of network storage and the network to connect it to app servers) AWS was the first major provider to be decently competitive. I remember pricing this out at the time, a lot of what you're paying for on EC2 is proximity to EBS, which other providers at the time either didn't have an equivalent to or charged several multiples of AWS prices.

If it's 2010 you're used to the costs of a hosted or coloed system with tens of TB of NetApp storage starting your next project on AWS was compelling in a way discount providers was not.

Even without an AppEngine style PaaS, AWS had a lot of cloud services available before Google (EC2 and S3 are notable services AWS had long before Google had anything comparable.)

I think a lot of the popularity advantage AWS has is a result of being first mover in many of these areas,

AppEngine had the misfortune of being too early.

I remember checking out both in 2007, right when AppEngine first launched. EC2 was just Linux; you had a familiar virtual machine to work with, and you could port your existing code over with virtually no effort. AppEngine required learning a whole new API, which was completely non-portable and locked in to Google, and that was a complete non-starter.

Also, being a PaaS, there was a lot more to get right with AppEngine, and the API changed a lot over the years. I remember looking at it in 2007 and thinking "There is no possible way I can get work done with this." In 2009, it was "Well, it might work, but it's way too much hassle." In 2010, it was "Nope, still too hard." Finally in 2012, I tried it again and it was like "Woah, finally this stuff is usable, and it's actually pretty pleasant." But by then, the outside world had Heroku, it had Beanstalk, it had Parse & Firebase, and there's a huge ecosystem around AWS.

It's a pretty good example of path dependence. If you're just coming to cloud computing now and have never run your own software before, AppEngine is actually a fairly nice alternative. But in 2007-2009, when a lot of us were first checking it out, it was an overly-complicated, vendor-lock-in mess. And vendor lock-in mattered a lot more then; now many developers have just accepted that they're going to be locked into certain platforms, but back then, building on anything that didn't have a publicly-specified API was a huge mistake.

This matches my experience. Right now, AppEngine is really nice (pleasant).

And if you use Flask/WSGI interfaces + Cloud SQL there isn't even a lock-in.

The only big issue I can see: no AppEngine in China. :/

At the moment I'm resigned to use app engine for the world-minus china, and some AWS EC2 inside China, running the same code plus a a lot more admin work.

Elasticbeanstalk is the answer to App Engine and has been around for many years.

It's not really. Elastic Beanstalk is about dynamic provisioning of virtual machines from AMI images. It doesn't solve the security maintenance issue, for one.

Well, if you take it that way sure, but the goal of Eb is to provide a similar solution like GAE and customers should be replacing instance as often as possible.

Reply on news.ycombinator.com