My re:Invent 2020 Highlights
Jon Hammant, AWS, Head of Compute UKIR
What a three weeks it's been! I've been spending a great deal of time watching a whole bunch of content from re:Invent 2020 and have been absolutely delighted by some of the customer stories we've been able to highlight, as well as the announcements we've been able to share.
This is my first re:Invent working at AWS, and whilst I would have loved to have been in Las Vegas, I feel that the remote format has enabled us to reach a huge audience of people from all over the world who wouldn't normally be able to see the amazing progress going on. It's mind blowing how much useful information we've shared out for free. For me this is exactly what cloud stands for, the democratisation of information and resources to a the world.
It amazes me the power available today to build by combining simple services with deep capabilities - each time my home security system (homebrewed!) uses AWS Rekognition to greet me I'm amazed at what I've managed to build by stringing together some simple devices, python and a powerful yet easy to use cloud service. Over the next few years we'll see small, nimble teams inside large & small organisations using and combining new services to create massively powerful and useful functionality that could not have been created by any one team in isolation. I feel the cloud is the closest thing we have to building Turing's Cathedral right now.
We are at an extremely exciting point in the genesis of Cloud right now, as Andy Jassy mentioned in his keynote we're only at 4% market penetration - over the next few years I expect this number will grow at a vastly increased rate as enterprises move into the scaling phase of cloud adoption. I made my own personal move to Amazon because I honestly believe there is no better place to be to have the scope and impact on changing the world than working at AWS.
I could (and happily would!) talk at length over why I think extremely well placed during this change, however instead I'll give you my view on my top three highlights & themes of re:Invent 2020!
1. Cloud, Everywhere
A number of key announcements dealt with a key change that I see happening across different areas of our product portfolio. As cloud matures we're moving to an increasingly distributed deployment, whether that be through the location of the infrastructure itself or the technical stack.
We announced EKS-Anywhere & ECS-Anywhere (as well as the OpenSource EKS-D) - allowing customers to extend workloads between AWS and on-prem data centres. I think we'll see pretty widespread usage of this over the next few years as people sweat their on-prem investment prior to sunsetting and moving all-in on Cloud. I'm extremely proud we're giving people the ability to run the same market leading software driving containers in their own data centres as we're running in ours.
The growth of Outpost into increasingly small footprint units also allows us to increasingly serve our customers specific use cases (data residency, latency, regulation) that need AWS services in customer locations, we're dropping the barrier to entry here and allowing smaller deployments.
We announced our growth of Local Zones, announcing 3 more in 2020 followed by 12 more in 2021. As use cases for cloud change and complex processing increases there will be a growing ubiquity and need for low-latency, high bandwidth connections to nearby cloud zones. In addition we highlighted expansion and progress in our Wavelength service, offering AWS zones within the 5G providers own networks - which will start to power all kinds of interesting new 5G services (one to watch).
Cloud, in more locations, powered by containers, facilitated by advanced services, allowing our customers more choice and power, without increasing cost for them will enable some amazing use cases - autonomous vehicles, Cloud Robotics, IoT, VR/XR, High performance rich AI/ML leverage multiple sensor networks, VTOL transport networks, HPC-Compute-on-Demand, and far, far more.
2. Heterogeneous Computing
For many years I've been predicting a rise in the world of Heterogeneous Computing - using the right hardware, for the right job, in the right location, at the right time. We've seen x86 winning the tech war for quite some time, with the Tick/Tock style evolution of x86 (Clock Speed/Die Shrink) keeping it ahead of everything else. Over the past few years we've seen the rise of GPUs, FPGAs and now ASICs - however as great and as powerful as these have been, we've not really seen broad usage until recently (and even that is massive limited in comparison with x86). I think this trend is going to reverse very, very rapidly with people starting to use a far more mixed set of compute nodes.
Previously been a number of reason for the lack of progress: increased complexity of development for "non-standard" had nullified any real world benefit (previously when I was involved in development of trading algo's using FPGAs - we often found that x86 development won, mainly due to the speed and richness of the development environment). In addition actually getting unique hardware on demand, in the right place was nearly impossible. I remember well the challenge of getting liquid immersed servers and Xeon-Phi cards approved for usage in a data-centre. These issues stopped the adoption before it could ever really start.
We announced new EC2 Graviton2 ARM instances. We're currently seeing fantastic benefits available from porting environments to ARM, I think we'll very much see this as a trend over the next year, with some really impressive cost/benefits available for a very good value investment in platform migration. Also - we'll increasingly see Graviton2 underpin other services (eg RDS), enabling customers to take advantage of the benefits with very limited changes needed on their side. This increased layering of technology will allow teams to take advantage of the right hardware in a way that was never before possible.
Our investment in custom hardware with AWS Nitro will also increasingly give some unparalleled advantages in Cloud security and innovation, again heterogeneous compute is in use here without customers needing to worry. You just benefit from a more secure, more available systems - as well as allowing us to deliver Mac O/S machines in a cloud environment (yay!).
From a machine learning perspective we had a number of announcements on the hardware side that were pretty interesting, including AWS Trainium to speed the training of ML, partnering with AWS Inferentia to dramatically drop the inference costs on processing those models. I was also really excited around AWS Panorama - a hardware device that adds powerful computer vision onto onsite cameras - there's some really interesting possibilities that will be driven here. I was also really happy about the change in in granularity of Lambda charging (100ms to 1ms) - this will see real decrease in cost and increase in usage of the technology.
Making extraordinary technology feel routine will really accelerate the pace of innovation and transformation for our customers and allow billions of people to benefit from new services and technology.
3. Technology, Open to All
Finally I was extremely excited to see the increasing democratisation of technology that we're pushing on, allowing increasingly sophisticated technology to be used be all.
This is extremely interesting to me in one area - Automation.
Automation is one of the dirty secrets of technology - I constantly see top down approaches pushing automation down to technical teams and service providers. However I actually see very few organisations that have managed to put platforms or methods in that really scale to enable automation to really deliver value. We've taken a specific slant on this problem and want to democratise the automated delivery of services in a sensible managed way.
This is done with probably my favourite announcement from the whole of re:Invent - AWS Proton. This allows you to easily create scaleable services that can be used by multiple teams across an organisation - it fits perfectly into the increasing used Platform Team model of technology. The layout of tech teams are changing from the traditional infrastructure aligned model (Unix Team, Networking Team, Storage Team) into a far more service oriented, platform approach - where smaller teams are responsible end-to-end for business and tech services, with individual users/dev-teams actually responsible for the spin-up/usage.
For example an ELK team could be responsible for the underlying ELK functionality, patterns and support - however individual teams would invoke instances of it. Previously - this has been a pretty difficult thing to do and manage at scale - one that required significant investment to get right. Proton allows this in an easy and simple way - I think out of all our announcements we'll see a huge amount of customer usage here. I also think we'll see a number of our partners investing time and effort in creating blueprints and accelerators in Proton that are rolled out to large amounts of people - ending up with better value all round.
There's a huge number of examples of us opening up advanced technology, however the announcements from SageMaker in the forms of Data Wrangler, Clarify (amazing for helping detect bias an increasingly key topic), Edge, Jumpstart, Pipelines, Feature Store, Debugger (and many more) were by far the stand-out for me. These combined with SageMaker Studio - the online Machine Learning IDE - will help bring advanced Machine Learning to many, many more people. Try this out if you haven't already - it's worth it.
Also on this note - we offer a free tier on AWS. If you haven't already sign up, then go and use some things for free. It's pretty amazing.
Democratisation of advanced technology is key to us helping everyone take advantage of the positive changes in the world and something we need to work increasingly hard to deliver on.
Summary
It's been an amazing event and there's a whole bunch of content to watch. Let me know if you have feedback or questions.
I'd love to know what people liked most and what they thought the best sessions to follow were - we're here to help and make sure everyone comes on this exciting journey!
Business Leader | Managing Director | Advisor | Digital Transformation
4 年Great summary.