AWS re:Invent 2017 in Las Vegas attracted over 40,000 attendees. From the vendors to the conference itself, there were themes of predictive...
You are here
Event Diary: AWS Summit, NYC 2018
Held on July 16-17 at New York’s Jacob Javits Convention Center, AWS Summit NYC was one of a series of 34 free, end-user events Amazon is delivering in major tech capitals this year. Complementing the week-long annual re:Invent (held in November in Las Vegas) and taking the same format in miniature, AWS Summits provide a day of affordable half- and full-day “bootcamp” trainings for AWS certification exams, plus a second day, entirely free of charge, featuring top-tier executive and partner keynotes replete with new-feature and -service announcements, a full show-floor and partner expo, rapid-fire third-party product demos, fully-automated instructor-led and self-guided hands-on labs, birds-of-a-feather sessions, plus continental breakfast, lunch, and post-show networking.
Yes, one might cynically note that this large investment in end-user outreach, community building, and education probably pays off handsomely for AWS in motivating customers to expand and elaborate their use of hosted cloud services. But AWS Summit NYC -- presumably a good example of what these events deliver elsewhere -- was also one of the best events I’ve attended, so far, this year. Like pretty-much everything bearing the Amazon brand, it delivered lots of easy-to-consume, practical value, at large scale. Here are some highlights:
Fortnite Triples Down
Speaking of scale, it’s hard to imagine greater than that involved in delivering Fortnite, Epic Games (creators of Unreal Engine) battle royale game, which pits live groups of 100 players on multiple platforms in a last-person-standing free-for-all. Introduced midyear last and growing quickly to challenge (and more recently, surpass) Player Unknown: Battlegrounds as the leader in its kill-or-be-killed casual game genre, Fortnite, hosted on Amazon EC2, was reputed to be supporting between 2 and 3 million simultaneous users in the first quarter of 2018 (the somewhat-smaller PUBG is hosted on Azure), and churning revenue on the order of $125+ million USD per month.
The headline-grabbing announcement at AWS Summit NYC -- delivered by Amazon’s CTO, Werner Vogels, and Epic Director of Platform, Chris Dyl -- is that Epic Games is expanding its use of Amazon services to create a machine-learning and analytics pipeline, using Amazon SageMaker ML model training/hosting/production deployment. The artificial intelligence will be used to track and understand player preferences, improve player skill matching; and likely for more ambitious projects, such as directing the behavior of huge groups of NPC opponents (intelligent automata) in realtime PVE (Player Versus Environment) combat, as suggested by Epic Games CEO, Tim Sweeney, in a Medium post, last year. The Fortnite team will also be shifting towards use of Kubernetes-hosted containers at scale, hosted on Amazon Elastic Containers Service (EKS).
Cloud9 IDE and DevOps
AWS bought SF-based “virtual IDE” startup, Cloud9, in 2016, and, last November, introduced this entirely cloud-based, multifunction integrated development environment to AWS users. AWS CTO Werner Vogels presented an informative intro to Cloud9’s impressive features, which are quickly being built out as part of a complete infrastructure-as-code/DevOps solution. Cloud9 can now consume code from GitHub, Amazon CodePipeline, and similar repo services; downline it for build/test using Amazon CodeBuild fully-managed build service (or Jenkins, with other solutions in the pipeline), and then deploy many different ways, including Chef-based deployment with AWS Opswork Stacks. All these CI/CD options can, of course, deliver and lifecycle-manage production builds on Amazon hosted substrates and services. They can also work with premise bare-metal or cloud infrastructure; so Cloud9 and its integrated DevOps services may end up being the only IDE your organization needs.
SageMaker Streaming Algorithms, Channel Synthesis
Amazon SageMaker is a hosted framework for designing, training, testing, and deploying machine learning solutions at scale, that works with an expanding palette of familiar tools like MXNet, TensorFlow, and Chainer. Unlike some Amazon services that emphasize and enable use of commodified infrastructure, machine learning is highly resource-bound: requiring serious CPU/GPU hardware for fast processing and often significant storage and uptime for iterative training. So working with SageMaker used to require data scientists devote much attention to optimizing cost/benefit, particularly in the training phase. At AWS Summit: NYC, Dr. Matt Wood, head of Machine Learning at Amazon, announced availability of new algorithms (initially for TensorFlow, soon for other frameworks) enabling streaming of training data direct from Amazon S3 storage to SageMaker models. Along with other batch-processing innovations, streaming algorithms can, in principle, lower training costs up to 90%, making it affordable to apply SageMaker to whole new classes of (less dramatically- or easily-monetizable) problem.
Wood also announced upgrades to several of Amazon’s ML-based computational linguistics services, increasingly popular with call and contact centers (including virtual contact centers, engineered with Amazon Connect), like Amazon Transcribe hosted automatic speech recognition (ASR). It will soon be possible to submit multi-channel (i.e., multi-speaker) conversational audio recordings via the Amazon Transcribe API, receive back transcribed dialogue text for each speaker, then submit the individual transcriptions to Amazon Comprehend to generate realtime sentiment analysis for each speaker.
EC2 instances for AWS Snowball Edge
In terms of cloud-service architecture, one of the more impressive announcements at AWS Summit: NYC was about Snowball Edge: an edge-network device/platform that lets you move various kinds of virtualized compute closer to mobile endpoints or into customer facilities, plus network services that facilitate transport of that remotely-sourced data back to centralized AWS services at faster-than-internet speeds. Snowball Edge was launched with the ability to deploy Lambda serverless functions on the remote endpoint devices. Announced at AWS Summit: NYC was a new feature: ability to deploy EC2 VM workloads, giving you much more flexibility to create different types of sophisticated, highly-responsive processing. The new feature was discussed in detail by AWS evangelist Jeff Barr, in a blog released during the show.
While a lot of the action revolved around keynotes and hands-on lab sessions, I found the expo fascinating as well. One obvious trend, appropriate for the New York market in which healthcare and finance play such a large role, was the presence of many third-party providers of security and audit technologies for AWS: stuff that purports to make Amazon Web Services safe for valuable and regulated data, including AlertLogic and Sophos. Also present were several Opsview partners, including our friends at ServiceNow and VictorOps. Opsview, of course, can monitor all the most important AWS services, including EC2, S3, and RDS.
Meanwhile, if your career revolves around, or seriously intersects with Amazon Web Services, you owe it to yourself to make time for AWS Summit events in your area. They’re solidly produced, upbeat, super-informative, and of course, you can’t beat the price.
More like this
Amazon Web Services has become a go-to provider for cloud computing efforts because it simplifies the process of setting up servers and services...
New support for AWS services and regions helps IT teams monitor their cloud and on-premise environments within a single product.