AWS re:Invent 2019 – Podsumowanie

AWS re:Invent 2019 — Summary

Lukasz Dorosz
Ponad 14 lat w branży IT. Konsultant i architekt projektów Amazon Web Services. Entuzjasta rozwiązań serverless. Współtwórca AWS User Group (3500+ osób). AWS Community HERO. Masz pytanie, napisz do mnie.
pl flag
en flag
Voiced by Amazon Polly

As every year at this time, there was a conference organized by Amazon Web Services, AWS re:Invent.

The event, which brought more than 65,000 participants to Las Vegas this year, was full of various interesting attractions and incredible information.

During these 5 days (01-06.12) there was a huge number of technical sessions and other activities such as workshops, competitions and the possibility of passing certification exams.

One of the most important points of the program is, of course, Keynotes Andy Iasi (CEO AWS), and Verner Vogels (CTO Amazon). In the course of these speeches, the most novelties are usually announced.

I this year for the first time had the opportunity to watch these presentations not from the screen of my computer, but the 3rd row just opposite 😎

Let me therefore share with you my brief summary and my conclusions.

AWS still number 1 in the world

Looking at the division of the market, I still and probably for a very long time AWS will be the number 1. Of course I deliberately added in the title in the world, because this division is different in different markets.

In any case, AWS now has 47.80% of the market in terms of the IaaS segment (Infrastructure as a Service) and for the 9th consecutive year Gartner's IaaS leader.

Although this advantage is gradually melting, because there are other suppliers. However, it does not decrease as much as it seems to some. In the post from the year ago, the market share was 51.80%, i.e. by 4% for a year.

However, it is worth remembering that the global share of the cloud is still at a very early stage. At the moment, only 3% of IT expenditure is spent on the cloud.

There is, therefore, a lot ahead of us in the next years. Many companies are still at the stage of considering whether the cloud is what can give them value. Looking at the development and opportunities it offers, one should expect that the distribution of expenditure on IT will change.

Machine Learning, Machine Learning everywhere

If I had to specify the guiding password this year, machine learning will be that.

Andy Iasi in his presentation, announced quite a lot of different novelties. In fact, compared to Verner's presentation, it was practically the only keynote to announce anything.

In the IaaS area, among other things, new generations of machines driven by AWS Graviton2 processors have appeared. Some improvements related to networks and their connectivity have been added, such as Ingres Routing in VPC or inter-region peering in Transit Gateway.

But Andy spent most of his time on solutions that allow building or are based on machine learning. Amazon SageMaker, a service that allows you to train ML models, is enriched with more “fichery” from year to year.

This time there was a kind of IDE (Integrated Development Enviromnet,) for machine learning (ML) - Amazon SageMaker Studio. This service allows you to manage the entire machine learning workflow from one place. Based on some other services that have also been presented in the course of this speech:

Amazon SageMaker Debugger

Amazon SageMaker Model Monitor

Amazon SageMaker Autopilot

Amazon SageMaker Experiments

In addition to everything around SageMaker, Andy announced a new Amazon Fraud Decetor service (at the moment in preview), which will allow without much knowledge of machine learning to build solutions for monitoring and detection of potential financial fraud.

Another interesting service (also for the time being in Preview) is Amazon CodeGuru service, which will automatically view the application code using ML and indicate potential problems or places that require excessive resources to be executed. It's about the ability to optimize the code for performance and power demand.

Whether and how exactly it will work in practice, time will tell. Although we know well what it's like when the application runs “slow”. It's always easier to add power than to look for optimization in code 😉

Andy showed that, on their example, they managed to get a lot of resource optimization using this service.

Amazon Kendra is another service that can use ML to search data contextually. It can be a great application in applications that process users' questions and with “understanding” they have to know what they are looking for and return the right information.

The cloud even closer to you

A year ago on re:Invent, AWS Outpost was announced, which is a hardware solution that allows you to have AWS in your own server room. However, it was only this year that the service actually saw the light of day. You can already order AWS Outpost (unfortunately not yet to Poland) and “stretch” AWS cloud.

Currently, Outpost can run virtual machines (Amazon EC2), together with EBS disks. Configure Virtual Private Cloud network or container environments based on Amazon ECS and EKS. Over time, more services will be added, such as Amazon RDS databases.

The second such solution to be closer to us and closer to our customers is AWS Local Zones.

This is a kind of new type of location, designed for applications that require very little delays. Local Zony will be built in big cities and will offer compute (EC2), storage (Amazon FSx) and network services (Application Load Balancer and VPC). At the moment, the first such wife is available in Los Angeles.

Containers, but more serverless

So far, if you happen to look further beyond virtual machines and Infrastructure as a Service, you probably have seen two slightly rival directions. One of those who build applications for containers, often placed on various clusters of machines, orchestral services such as ECS, Kubernetes, etc.

The second direction of those who say that you already need to go a step further and build applications at serverless end, using the services provided by the providers.

AWS for a long time, you will see that it is turning quite hard in this direction to remove from its clients server management. There is, of course, a lot of services that we do not manage machines. However, when you put the application in containers, these machines always appeared somewhere.

At first it was Amazon ECS, a management service, container environments, where normal virtual machines were placed in the cluster. But then there was something called AWS Fargate, a cluster where there are no servers, and it's a kind of virtual entity that fires containers.

Then came the Kubernetes service - Amazon EKS, where part of the so-called Master Nods was within the service, however, nody worker continued to install and configure the client. Until now...

.. when Andy in his session announced that at Amazon Fargate, he is also available in EKS.

I think this is the second important element behind the ML topics that was discussed on re:Invent. Andy Iasi, and then in Verner's session, there was a quantum of time devoted to the EKS in Fargate variant.

At this point, we can put up infrastructure for Kubernetes, where we do not need to manage any virtual servers.

The second huge advantage is greater flexibility of scaling. Comparing statistics for the classic variant on EC2 and Fargate machines, in the latter we achieve better flexibility. The environment scales faster, thus maintaining high performance, and can be more cost-effective.

At the end of the

For me this year's re:Invent was under the slogan of Machine Learning and Containers. During the conference itself there was a lot of other interesting topics - you can find a catalogue of sessions and presentations on the AWS YouTube channel. Of course, there was a lot of other novelties, which especially for you, we collected with Przemek even before re:Invent. You'll all find HERE.


I also invite you, at AWS re:Invent re:CAP, an ONLINE meeting within the AWS User Group PL community. We will be discussing all the news from AWS re:Invent 2019.

I'm also curious what you liked most about this year's re:Invent and the news that were announced. Drop your conclusions in the comment or write to me.

And one more thing...

Watching the presentation of both Andy and Verner, he made me a certain conclusion. AWS is a leader and sets directions in certain areas that others follow.

Why is this happening?

Perhaps because they have

A large part of the solutions they build apart from being the result of customers' needs are also the need for them. Having such a big business as, they themselves are a source of different needs/ideas that can address new solutions. At the same time, they have a training ground where these new ideas and solutions can implement. Cheers 🥃