There were a bunch of AWS announcements this week that I’m super-interested in trying out.
“Serverless” Aurora could be a big one for us. We have a bunch of MySQL databases currently hosted on RDS that spend an awful lot of time idle. Essentially they’re only ever used when we’re generating reports, or when a client is accessing those reports. I’d estimate that about 75% of the time they’re doing nothing, and most of the time when they are doing something it’s not super-heavy.
They had to be scaled up from bare-minimum to get full-disk-encryption and to handle the occasional spikes. So we’re paying quite a bit for resources that are only used occasionally.
Thus if the “severless” Aurora can spin up quickly from idle, it’ll be a very interesting possible replacement…
Similarly the T2 “unlimited” instances. Right now we host a lot of c4.xlarge EC2 instances because when we need the cores, we need the cores, and our application has been trying to live in a world where we do SaaS but sometimes clients want on-premises. This is changing, we’re more inclined to tell clients who want on-prem to set up a VPN to an AWS VPC and get over themselves (regulatory requirements are generally not the problem here), so offloading much of the compute-intensive work that needs those cores to Lambda, and putting the rest on T2 with the ability to occasionally burst up beyond the CPU credits, well, that’s interesting.
The “severless application repository” may be of interest, but I’m not sure that much of the stuff that’ll be made available will be all that relevant to us. We’re in a pretty niche business, and there’s a reason we had to write our own applications. But I’m more than happy if that initial evaluation turns out to be wrong: anyone going to throw up a HTML parser at least as good as HTML::TreeBuilder that spits out JSON? Or a good CSS parser that does likewise?