AWS re:Invent 2018: The Biggest And Best Yet
Cloudreach Head of EMEA, Chris Bunch, recently returned from AWS re:Invent 2018, the Cloud platform's headline event in Las Vegas. Here, Chris shares his thoughts on some of the highlights from the event.
Writing this at 2am on Sunday morning, having ruined my sleep pattern in Nevada for another year. It was worth it. I don’t want AWS to get big heads, but quite simply they crushed it this year. An amazing event, from logistics and scale, through to an almost unbelievable selection of new releases and enhancements. Here’s a quick recap of highlights for those that didn’t go, or who have jetlag fogged heads and have forgotten….
How was that crystal ball?
Pretty sharp! I’d argue I got 4 of the 5 main predictions I made in advance right, although AWS not simplifying their billing mechanisms blocked a full house. Enough of the trumpet blowing, here are the big things you need to know:
1) re:Invent theme of the week: Machine Learning
The overall theme was definitely 'Machine Learning'. Andy Jassy pushed this hard during his keynote, releasing a plethora of new services. These included:
- SageMaker Ground Truth - for classification of data via Mechanical Turk, or other sources
- 150+ Machine Learning algorithms being made available in the AWS Marketplace. This will grow and grow
- SageMaker RL - an awesome step forward, allowing reinforcement learning and training of models through rewards over time. This, of course, led into DeepRacer, bringing RL to the masses through a commodity remote controlled car. This is similar in concept to the equally awesome Microsoft AirSim, which tackles the challenge from the air
- There were also jumps forward for text recognition, forecasting, and recommendation engines. Whilst they may seem smaller in comparison to the bigger announcements, it bears remembering that some companies used to have entire business models built off the back of these...
This is just a snippet of the most interesting. Overall, there were 13 new ML services - all listed here. A big step both for AWS and the democratisation of complex technology.
2) Serverless, serverless, serverless...
AWS got the drop on the world in late 2015 when they launched Lambda. They continued to race ahead last week with the announcement of a variety of new features.
Support for new languages such as Ruby and C++ was added, before dropping the bomb of being able to add any language yourself. This is a game changer, make no mistake. Lots of other languages are being worked on, including Cobol (yum) - finally opening up a path for migration for older mainframe code.
Possibly the most exciting introduction here is Layers, providing an effective construct for pulling in remote libraries. You can expect this to grow over time, but the usual early adopters like Epsagon, IOpipe and Serverless are already providing functions you can include. This is a huge opportunity for companies to make money by providing pluggable software components that can be called dynamically and paid for while in use only.
3) Hardware (yes, hardware!)
You probably weren’t expecting that. But AWS did release news here last week - and it’s pretty big.
First up, they’re no longer relying solely on Intel - who in previous years had been joyously speaking on stage in the keynotes. AWS' bigger push this year was ARM based (i.e. non-Intel architecture) chips of their own creation. They look cost-effective, and are intended for Cloud scale apps, i.e. for scaling out effectively. Almost certainly, this comes off the back of the acquisition of Annapurna Labs a couple of years ago.
Next up, and this causes me a bit of pain to write it, came AWS OutPosts. Whilst I know AWS have to do this to fight competitive pressure (AzureStack, traditional private Cloud vendors, etc), and to make sure they absorb as much of the Cloud market as is possible, the continued move towards private Cloud isn’t one that excites me. I still regard it as a distraction from the technologies (including Lambda above) that will really give your business a competitive advantage.
The tl;dr summary here is running AWS infrastructure on-premises - yes, actual AWS provided hardware. We’ve seen AWS putting their tech at the Edge already with Greengrass, and now they’ll be in your data centres as well. This is bad, bad news if you’re Dell or HPE.
4) Databases galore
A big chunk of Werner’s keynote was devoted to Aurora - a combination of Oracle bashing, talk of the world’s best Cloud-native database, plus confirmation of new features including the new fully managed global option.
In addition to a new managed blockchain service for those looking to build applications based on distributed trust, the Quantum Ledger database (QLDB) was announced (serverless, of course) for those looking for centralised trust-based solutions. QLDB democratises something pretty complex, and provides a solution for businesses needing a 100% proven audit trail as it’s append-only/immutable (can’t be edited) and cryptographically proven (definitely hasn’t been edited). There’s an explanation of some use cases here.
The other big news in database-land was TimeStream, a new time series based database that should prove a godsend to anyone working with large streaming data volumes (think IoT data collection for ML purposes). This was possible to set up before, but time-consuming. This is a simple and cost-effective (serverless, of course) offering.
5) Pricing of storage
Ok, yawn, it’s storage, but I still felt it made the top five. Why? Glacier Deep Archive is outrageously cheap - charging $1 per month to store a TB of data. $1. With "13 9’s" of durability (they won’t ever lose your data).
Awesome. AWS put the hammer down and put serious pressure on the competition last week. "Competition" doesn’t include Oracle of course, who nonetheless took a beating, as I also predicted,
The only real watchpoint on the above is to check dates on the newer services, as they’re not all available until next year in some cases.
I’d also reiterate that I’ve only covered a small section of what was released. It was an incredible week, with everything from a simplified data lake option, to a new Well-Architected tool, to a Managed Kafka service being released.
Sometimes one thinks life in the Cloud might slow down. On this evidence, not in the near future.
Chris Bunch is Head of EMEA at Cloudreach.