Week 19, 2019 - Transit Gateway Direct Connect; DeepRacer Available; S3 Batch Operations

By (4 minutes read)

Last week was the AWS Summit in Sydney and as I attended, it’s clear that my mind is focused on AWS. It helps that they released a number of new features, though not announced at the summit. I’ll focus on Transit Gateway integration with Direct Connect, Deep Racer, and S3 Batch Operations.

Transit Gateway Integration With Direct Connect

When Transit Gateway was introduced at re:Invent, the one glaring missing feature was integration with Direct Connect. This has now been rectified.

In case you want a quick recap, Transit Gateway is the new networking service that lets you connect all your networks together and manage the connections between them through a single point. If you’re familiar with the concept of Transit VPCs, which usually consisted of lots of expensive Cisco routers, this is a managed version of that. Direct Connect on the other hand offers you a dedicated connection between your private data centre or office and your AWS environments1. Having the two together, just makes sense2.

We’ve all known this was coming, as it was promised at the introduction of Transit Gateway. What we didn’t know is how it works, which this answers. In short, there is a new type of virtual interface for Direct Connect Gateways named Transit. You can only have one of these VIFs per Direct Connect Gateway, and it is in addition to the existing limits for your VIFs. Once you’ve created this, you can then associate the Direct Connect Gateway to your Transit Gateway3. Of course, all of this assumes that it’s actually available in your region, as right now it’s only available in the 4 main US regions.

DeepRacer Available

The Friday before the summit, I got the email I’d been hoping for for a long time. I finally got invited to the DeepRacer preview!

Unfortunately I didn’t have time to train any models before the summit, but I was still really happy to get this access and felt special. Until a couple days later when it was made available for everyone. Now, despite losing my status as being special, I’m happy that it’s finally available and for once I don’t even complain that it’s only available in us-east-1. Because I really don’t care in which region my racing car models run. While there are some changes4, in general it’s still very similar to what I wrote about it back in December so I’ll just refer you there if you want to read about it. But obviously, what you really want to do is use up the free tier and train your own racing car and enter it into the virtual racing league.

S3 Batch Operations

The last thing to discuss today is S3 Batch Operations. As the name says, this allows you to run batch operations on your S3 bucket. It’s pretty straightforward, you basically start a job from an inventory5 and tell it what to do. You can do this from the Console or the CLI. As always though, the interesting part is actually what you can do.

While there are a number of regular things you can do, ranging from copying the files to adding tags, the really interesting one is that you can make it invoke a Lambda function6. While there are many possible use cases for this, personally I think this can solve a recurring problem. Triggering Lambda when new files are uploaded to a bucket and have it run some things is something most of us will have done at some point7, but if you do this with an existing bucket you still need to deal with the existing files. But now you can run all of those same files against the Lambda function automatically. Which sounds pretty useful to me.


  1. If you want to learn a bit more about Direct Connect, we had someone explain it all at our last Melbourne meetup. [return]
  2. ☀️🏖⛄️ [return]
  3. Yes, that’s a lot of gateways. And you can’t do any of this with CloudFormation. [return]
  4. More tracks for one. [return]
  5. It’s possible you never created an inventory before, I know I never needed one. [return]
  6. Yes, once again it’s all about Lambda. [return]
  7. Even if it’s just playing around. [return]
comments powered by Disqus