Hacker News new | past | comments | ask | show | jobs | submit login

Wait...their model is an EC2 instance per customer? The normal limits Amazon imposes are 20 reserved or on-demand instances and 100 spot instances per region. You can request more, but will Amazon really accommodate a one instance per customer model?



Amazon is happy to. The limits you cite are merely the point at which you need to have a conversation with Amazon staff. They are quite happy to accomodate _much_ heavier usage from customers.


When I asked for a raise to my limit they denied me, on the basis that my usage was insufficient. Of course, my usage was low because I hadn't launched my product fully because I didn't have enough instances to serve a lot of customers ... catch 22.

So I had to build out a rather convoluted architecture that used the loophole of deploying to multiple regions and failing over to whichever region would give me an instance ... which gives me up to about 80 instances ... just barely enough for me to get going with a trial beta program.

Which is all just to say, it is slightly more than just a "conversation" that you need to have to get a higher limit.


Sorry you had a bad experience. I have gotten nothing but superb support from Amazon staff. I'm surprised they weren't willing to accommodate you, doubly so if you were ready to pay.

One thing that surprises me is when people talk about utilizing multiple availability zones in EC2 as some sort of burden. It's very clear from their documentation and architecture that you need to be capable running in at least 2 availability zones regardless if you want any sort of availability.


> people talk about utilizing multiple availability zones in EC2 as some sort of burden

My use case isn't for an ongoing server where you require availability. It's purely about compute power - I don't care where the compute power comes from but preferably I want low latency to my customer. So ideally I would just get all instances for any given customer from a single region.

I did find in the end that, as you say, I would sometimes not be able to get an instance in a region even when I was below my 20 limit for reasons internal to Amazon, so the failover work was going to be something I had to deal with anyhow ... but it just added complexity to my life earlier than it would have otherwise.

Edit: I would also mention that I certainly don't think of it as a "bad" experience. I think it is something of a small miracle that Amazon offers the service they do in the first place and I certainly understand why they have caution about handing out large limits to just anyone. I only made my comment above as a kind of caution to not just assume you're going to get a raised limit from Amazon immediately and especially don't leave talking to Amazon about it until the last minute if you're planning to launch something.


Thanks, that's an angle I haven't seen since I've mostly used EC2 as a hosting service. What little batch work I've done on it hasn't involved instance counts where I ran into limits. (Now, EBS i/o limits and other things... :/ )


"Can I please speak to your supervisor" works sooo often in these situations.


I've always assumed that the limits are there to make sure that no one can order 1000 instances and pay with a bad/stolen credit card. In other words, if the customer does a dine and dash, how much are we willing to lose?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: