AWS re:Invent 2020 officially opened

The AWS re:Invent 2020 in virtual format is officially opened. The next three weeks are going to be filled with virtual keynotes, leadership sessions and virtual breakouts covering the world of IT from A to Z.

And as a suitable kickoff to the event, CEO of AWS brought his keynote mayhem to all of us. Almost three hours full of industry leadership thoughts and naturally, a bunch of announcements of new services and features.

Jeff Barr was also liveblogging the event, and you can find his notes here: External link.

As it is not productive to reproduce what Jeff has already composed, let’s concentrate more on the major themes, that were covered this year.

The state of the union

One very important and current topic is touched upon before diving into the technology realm; equality is a major topic and Mr Jassy acknowledges that Amazon and AWS are still on their way on that topic. Being a responsible and sustainable company in the new normal of the world as we know it is certainly something that needs to be a part of every company’s agenda.

Where AWS is today and how it got there? We learn that today, AWS is $46B+ revenue run-rate company and has had 29% year-over-year growth. With the standard clause of how hard it is to get big percentages in Y/Y, if you are a multibillion run-rate company. It took AWS 123 months from founding to get to $1B, and in absolute numbers, the growth has been large, even if for a starting and a small company, to get to Y/Y growth numbers in the hundreds is easy, the absolute growth numbers of AWS are staggering. From a competitive landscape, according to Gartner, AWS holds 45% of the IaaS and PaaS cloud market, and AWS still holds the Gartner pole position against the competition. But also interestingly enough, it is noted that still only 4% of IT spend is on the cloud. That is a major driver, why all of the relevant players are turning their heads towards cloud migration stories and offerings.

The global situation with the ongoing pandemic has also impacted the business landscape of many companies. But it is acknowledged that the situation has also been pushing companies towards cloud platforms.

We move on to identify working thesis for companies how to drive and enable change – so in other words, how to re-invent yourself constantly for keeping yourself relevant in the changing market. The eight theses he presents are as follows:

  1. Leadership will invent and reinvent
  2. Acknowledge that you can’t fight gravity
  3. Talent that’s hungry to invent
  4. Solve real customer problems
  5. Speed
  6. Don’t complexify
  7. Use the platform with most capabilities & broadest set of tools
  8. Pull it all together with aggressive top-down goals

The next steps for compute

Compute landscape has stabilized to three different compute options: instances, containers and serverless. That is the de-facto situation, and everyone has claimed their seat in the virtualized table – pun intended.

Compute future in EC2 instances is all about efficiency and price. The world is embracing ARM more and more, especially after Apple paving the way with their recent announcement of Apple M1 silicon. The AWS Graviton has proven to be efficient both in performance and in cost, where there have been cost savings to up to 50% with Graviton2 based instances when comparing similar workloads to x86 based ones.

Clearly one main theme is purpose-built silicon. It has been a theme on instance types for a while now – we’ve had memory optimized instances, storage optimized instances and so on. But the trend is continuing on the silicon level for more and more specific instance types. Especially there are new announcements for ML enabled chips, like Habana Gaudi, which is new upcoming ML custom cpu developed by Intel owned Habana Labs. AWS Trainium ML training specialized chips are also coming later in 2021.

AWS container offering is supporting hybrid cloud model moving forward – the major announcements are ECS Anywhere and EKS Anywhere, which allow the installation and management of ECS / EKS fleets of containers in on-prem. That will make the management of the fleets uniform operation both in the cloud and at on-prem. Additionally, AWS is open sourcing EKS, which is very exciting news for the open source communities.

Serverless part brings the gap between containers and serverless functions closer together with the announcement of AWS Lambda container support. Keep in mind that the realities and limitations of Lambda still are in effect, so not all of your container workloads are valid candidates for “Lambdafication”, but the cases fitting the parameters (like for example the 15min execution max time) can be good candidates. Additionally, one seemingly small thing is that Lambda execution billing will go from 100ms increments to 1ms increments. That is a huge cost savings opportunity to architectures using a lot of small lambdas.

The final major feature of this segment is AWS Proton – automated management tool for serverless and container environments. AWS Proton enables infrastructure teams to define standard templates centrally and make them available for developers in their organization – which allows infrastructure teams to manage and update infrastructure without impacting developer productivity. Proton is in Preview in your “standard first wave” regions. Though it seems that Proton might be still a bit rough on the edges, so consider it as an MVP product for now:



Moving to different aspects of data. Getting the basics out of the way first: faster storage for EC2. We get gp3, which also allows scaling IOPS independently from capacity. And io2 Block Express, which is intended to compete with the performance and scale of on-prem SAN systems. Also, a possible candidate for virtual SAN on proprietary tech stack, like NetApp, if that is what you need to support your hybrid cloud or migration strategy.

On the database front, the move to more serverless paradigm is also in the move. For sure, Aurora Serverless has been around since 2017, but it has had it’s shortcomings, what comes to the speed of elasticity and performance. Aurora Serverless V2 (what a fine name) is tackling those issues with “instant” scaling. Support for MySQL out of the gate, and PostgreSQL later next year.

Some of the known players get their share of banter – but not to the extent we have actually come to expect. In any case – there is an interesting new thing called Babelfish for SQL server, which is in essence a transparent translation layer (thus the name) to stand in between of DB clients talking in Microsoft SQL and the DB backend, which is actually PostgresSQL. The only question that I have is that where is the Babelfish for Oracle?

One final thing in this segment is the AWS Glue Elastic Views, which also acts as the segway towards Machine Learning segment of the keynote.


Machine Learning

As there is separate keynote reserved for next week, we can only speculate, that we are discussing the tip of the iceberg here. But it is clear, that machine learning will be one of the strong themes this year as well.

The breadth of ML services is targeted from ML experts to everyday developers and data scientist and that out of that necessity leads to very wide offering. Most of the ML practitioner specific offering is packaged under the SageMaker offering and the announcements this year make no difference. There is Sagemaker Data Wrangler, Feature Store and Pipeline are the ones we hear of today – and probably more to come next week.

On a higher abstraction level, we have few interesting ML enabled services included in the roster. DevOps Guru is an advisory algorithm, which is learning from your environment setup and suggesting remediations and adjustments based on the data. QuickSight Q is a very interesting new feature, which promises to use Natural Language Processing to understand your questions in free form to get answers based on your data. That is basically like a Enterprise bridge computer in making – very interested to see what the real world results look like. Naturally – I think it is fair to assume, that the language which Q speaks is in English.

A lot of love is given to Amazon Connect, which is effectively a helpdesk or customer service solution as a service on AWS. These are features, which Amazon and AWS are supposedly using themselves for utilizing ML in enhancing the customer service processes. The idea of the features is that any process, where ML can assist the actual customer agent, the heavy lifting is done by the algorithms. The solutions vary from predicting what answers are applicable to the customer to identify the customer from their voice profile.

Next, we are talking about industry solutions. Monitron and Lookout are solutions for sensor analysis and predictive maintenance for use in industrial applications – engines, factories and such.

The future of Hybrid Cloud

The vision, what is Hybrid Cloud in 2020 is different what was meant a few years back when talking about hybrid cloud – which essentially was a division between the on-premises data center and IaaS/PaaS in a public cloud. Now, the holistic vision includes edge with Outposts, 5G networks with Wavelength, and only at the other end – the cloud.

Outposts will also be available in small configurations, starting from 1U and 2U. Interestingly enough, apparently, IT is now so far away from on-prem world, that it makes sense to explain, what a rack unit is.

Local Zones get updates on local availability, but unfortunately at this point – local means just US. I guess the old world needs to wait a bit longer on that.

Visit our AWS re:Invent event page to see more: External link.

Announcements during the first keynote

The full list of announcements which were specifically mentioned during Andy Jassy’s keynote:

  1. Habana Gaudi – Intel Habana Labs based custom processor for ML
  2. AWS Trainium – ML training chip custom designed by AWS to deliver the most cost effective training
  3. Amazon ECS Anywhere – schedule ECS workloads on-prem
  4. Amazon EKS Anywhere – schedule EKS workloads on-prem
  5. EKS open source
  6. AWS Lambda – per ms billing
  7. AWS Lambda – container support
  8. AWS Proton – automated management for serverless and container deployments
  9. EBS gp3
  10. EBS io2 Block Express – The first SAN built for the cloud
  11. Amazon Aurora Serverless V2 – Scale in a fraction of seconds!! 350,000 DBs were Migrated to AWS
  12. Babelfish for Aurora PostgreSQL – open sourced realtime translation library API from MSSQL to Postgres
  13. AWS Glue Elastic Views – Copy Data from each data store to a target store!!
  14. Sagemaker Data Wrangler – Data preparation tool to simplify data scientist job and prepare features, with 300 built in automated transformation
  15. Sagemaker Feature Store – centralized place to store Data and features that can be used with one or more model or with more data scientist
  16. Sagemaker Pipeline – first and easy to use CI/CD and MLOps
  17. Amazon DevOps Guru – uses ML to identify and detect operational issues and recommends actions
  18. Amazon QuickSight Q – Natural language processing and answers any question in seconds for your data
  19. Amazon Connect Wisdom – machine learning to listen to recording transcription and finds answers from all relevant data repositories
  20. Amazon Connect Customer Profiles – able to deliver faster and more personalized customer service
  21. Real-Time Contact Lens For Amazon Connect: Uses ML to detect customer experience issues during calls
  22. Amazon Connect Tasks – follow up tasks easier for agents and managers can automate tasks entirely
  23. Amazon Connect Voice ID – real-time caller authentication base on ML analysis on the customer’s voice
  24. Amazon Monitron – E2E equipment monitoring system for predictive maintenance
  25. Amazon Lookout for Equipment – predictive maintenance, identifies early warning signs of maintenance
  26. AWS Panorama Appliance – adds computer vision to your existing onsite cameras
  27. AWS Panorama SDK – enables manufactures to build next gen computer vision equipment
  28. AWS outposts – 1U and 2U outposts
  29. AWS local zones – Regional expansions (all in US)AWS wavelength – Regional expansions

Links to AWS announcement posts External link. External link. External link. External link. External link. External link. External link. External link. External link. External link. External link. External link. External link. External link. External link. External link. External link. External link.