As you probably know, Transloadit has been running its production infra on Amazon's cloud since day 1 - our block stores, encoding servers, loadbalancers, the full monty.
Except for 1 often forgotten, yet essential part of infra: our name servers.

Yesterday however, we succesfully completed the switch to Amazon Route 53 name servers.

This should result in increased availability as transloadit.com's zone is now hosted across 4 different TLDs:

• ns-1849.awsdns-39.co.uk
• ns-1264.awsdns-30.org
• ns-633.awsdns-15.net
• ns-471.awsdns-58.com

So basically half the world could disappear and (assuming you're with the lucky half) you should still be able to resolve Transloadit's hostnames to IP addresses! Now isn't that nice : )

Furthermore we'll be able to scale a bit faster because - thanks to their API (and a nice wrapper) - we can further automate launching and cleaning up after our encoding servers. Or drones as we like to call them.

So that's it, all of Transloadit's infra is now proudly running on AWS - no exceptions - and we hope you'll sleep even better knowing this.

Best wishes,
Kevin

After a few recent complaints from our customers, it became clear that our x264 encoding configuration was way too slow. We had missed some critical optimizations when configuring the library, as well as not using optimal parameters for the encoding itself.

Today we are happy to announce that we have fixed all of these issues:

• We are now configuring our x264 library to take advantage of CPU-specific (assembler) optimizations
• Our presets have been updated to use a slightly faster encoding strategy that still produces great quality.
• Things are now configured to take full advantage of our new 8 core machines

So how much of a difference does this make?

In one especially slow scenario, we are now seeing a 20.5x speed increase over our previous configuration. While your milage may vary, you should very likely see a similar speedup in the neighborhood of 10x or more.

That being said, we will continue to improve our performance over time, so if you feel something is not fast enough, just let us know and we'll investigate.

--fg

Last night we switched our default machine type from Amazon's c1.medium (2 core) instances, to x1.large.

The new machines are featuring 8 cores, 7 GB memory, better I/O performance and 64 bit architecture.

If you are performing assemblies that are heavy on encoding, especially those involving multiple output formats, you should see a significant boost in encoding speed.

Unless something comes up, our plan is to only use 8 core machines in production going forward. And who knows, the Cluster Compute Instances (cc1.4xlarge) are only one step away now : ).

--fg