We had the opportunity to attend Amazon AWS reinvent user conference, held from November 26th till 30th 2018, across Las Vegas. With over 50k attendees, reinvent has become the yearly come and get together of the IT industry, a unique position that AWS first achieved in 2017 and has more than well defended in 2018.
Prefer to watch – here is my event video … (if the video doesn't' show up – check here)
Here is the 1 slide condensation (if the slide doesn't show up, check here):
Want to read on? Here you go:
AWS announces AWS Outposts – An area where AWS risked falling behind against the rest of the IaaS competition was the area of workload portability from cloud to on-premises. The next generation computing platform as I call it has been described in the Market Overview that can be found here… And AWS was holding out likely the longest being the public cloud market leader. With the Outposts announcements, together with VMware, AWS will now bring key services to on premises, giving customer a choice between an AWS or VMware control plane. It will be available in the 2nd half of 2019 – so still a little time out. But probably the most impactful announcement from AWS – ever. It catapults AWS into the hardware maker and tech stack maker market… something you could not see coming a little back.
AWS takes AI to hardware – An area where AWS in our view was behind as well, was in custom hardware for AI loads. It has partnered with Nvidia as many others – but not brought forward a custom design… that changes with the new Amazon EC2 P3dn instances, that run a custom / optimized version of Tensorflow that increased efficiency from, 65 to 95%. They also support (vanilla?) Tensorflow, PyTorch, Apache MXNet, Caffe, Caffe2, Microsoft Cognitive Toolkit (CNTK), Chainer, Theano, Keras, Gluon, and Torch. Not enough, AWS announced a custom chip, AWS Inferentia, that will become an alternative to GPUs, and will be optimized for high throughput and low latency. Last but not least, Amazon added Amazon Elastic Inference - making it more efficient for customers to provision just the GPU istances they need for their AI / ML worklodas. All four announcements are key make AWS more competitive vs Google, that still has a 2-3-year lead in custom AI hardware. But it is good to see AWS (finally) engaging here, more competition is good for enterprises and fosters innovation as well as price competition… all good trends for enterprises.
AWS launches AI apps – Making it easier for business users to use AI apps is key these days and the strategy for vendors is to help them with specific, targeted apps. With Amazon Personalize, Amazon Forecast (both re-using and building on expertise and technology from Amazon.com), Amazon Textract (an OCR replacement)-and Amazon Comprehend Medical (Healthcare use case with NLP) there has been a broad push into AI apps. We will have to see how well practical adoption will pan out, but interests were big at reinvent.
Database Innovation is alive and well – Amazon execs know how important databases are and how much critical enterprise load is in relational databases. The strategy of AWS is to combat the generic, all-purpose databases with more dedicated databases, all in a suite of database products. Last year Amazon unveiled AWS Neptune, it's graph database, this year it unveiled Amazon Quantum Ledger Database (QLDB – a distributed ledger database, but what a confusing name), what a confusing name) and Amazon Timestream (a time series database). A good strategy that seems to get more traction in enterprises. But AWS knows that entrenched vendors are the challenge, and Oracle specifically has almost become an obsession. Multiple references in Jassy's and Vogels' keynote as well as almost 30 minutes on the topic in Vogels' keynote are almost too much attention.
AWS open sources Firecracker – Influencing industry standards via open source has been the trend of the last two three years. Google has shown the way with Kubernetes and TensorFlow… so AWS did not want to stand back and open sources its serverless framework (Amazon likes to call it virtual machine manager) Firecracker… with Google focusing a level higher currently at the service mesh level, with Istio, this is an open space for anyone. And AWS has good chances given that Firecracker runs Lamba at AWS, and Lambda is the leading serverless platform in the cloud (from a load / uptake perspective). Now we must see if other IaaS vendors will support Firecracker and how successful AWS is at stewarding a strategic open source project. CxOs have accepted open source by now, so it really comes back to show traction in the offering from a roadmap and consumption perspective.
Amazon QuickSight Renaissance – Two years ago Amazon debuted Amazon QuickSight a product that targets to replace the usual BI tools (remember Business Objects, Cognos, MicroStrategy?). The most promising capability was its automatic capability to visualize data. A key vision and step towards user free software, ahead of its time back then. Then it got quiet around Amazon QuickSight… but AWS does not give up easy and came back with a new attempt to win this market, and the value proposition remains strong. Time to look and evaluate this product for CxOs. True to cloud DNA, the pricing is now no longer per seat, but per session (with a maximum of $5 per reader user per month) – an innovation on the licensing side.
AWS Tooling everywhere – Last year Amazon announced its own IDE, a key strategy to get closer to developers. But switching IDEs is not easy, so this year AWS is bringing the AWS Toolkit to PyCharm, IntelliJ (Preview), and Visual Studio Code (Preview). It's important to help developers build apps fast and bringing the toolkit to the 'living room' (aka IDE) is the right strategic move.
In general AWS is impressive when it comes to delivering innovation across the board, across the product set. An astonishing feat that the AWS product teams manage to repeat year over year. And that raises the bar for innovation for the overall industry to keep up with.
On the concern side, the AWS portfolio is broader and deeper than ever before. AWS keeps staying with the Chinese menu philosophy, and it means massive choice… but staying with the analogy, when a Chinese restaurant adds a full menu every year, even the most loyal customers will have a hard time to keep an overview of the menu – and manage to repeat-order the same dishes (aka products) at the next visit (aka implementation project). I was not around to ask Andy Jassy my traditional question (Is AWS getting too complex?) but he came prepared (very Amazons style) and negated the question (no surprise). Customers though are struggling to find repeatable success across the vast AWS services portfolio. And cross portfolio QA and testing is getting close to impossible. So, simplification by grouping services, testing them together, aligning roadmaps, versioning and more … must be in AWS future – at some point. To be fair at re:Invent, AWS has made steps to make it easier for customers to run the large product portfolio - with new offerings like Landing Zone (AWS Control Tower); Security (AWS Security Hub); and, Data Lakes (AWS Lake Formation), customers should be faster and have an easier time to setup, operate and maintain AWS offerings. That is a good step in the right direction.
But overall respect to AWS to a massive, well organized and amazing event, sharing the breadth and depth of the AWS services portfolio and innovation. Here is to another year taming the dragon.
Here is the 1 slide condensation (if the slide doesn't show up, check here):
Want to read on? Here you go:
AWS announces AWS Outposts – An area where AWS risked falling behind against the rest of the IaaS competition was the area of workload portability from cloud to on-premises. The next generation computing platform as I call it has been described in the Market Overview that can be found here… And AWS was holding out likely the longest being the public cloud market leader. With the Outposts announcements, together with VMware, AWS will now bring key services to on premises, giving customer a choice between an AWS or VMware control plane. It will be available in the 2nd half of 2019 – so still a little time out. But probably the most impactful announcement from AWS – ever. It catapults AWS into the hardware maker and tech stack maker market… something you could not see coming a little back.
AWS takes AI to hardware – An area where AWS in our view was behind as well, was in custom hardware for AI loads. It has partnered with Nvidia as many others – but not brought forward a custom design… that changes with the new Amazon EC2 P3dn instances, that run a custom / optimized version of Tensorflow that increased efficiency from, 65 to 95%. They also support (vanilla?) Tensorflow, PyTorch, Apache MXNet, Caffe, Caffe2, Microsoft Cognitive Toolkit (CNTK), Chainer, Theano, Keras, Gluon, and Torch. Not enough, AWS announced a custom chip, AWS Inferentia, that will become an alternative to GPUs, and will be optimized for high throughput and low latency. Last but not least, Amazon added Amazon Elastic Inference - making it more efficient for customers to provision just the GPU istances they need for their AI / ML worklodas. All four announcements are key make AWS more competitive vs Google, that still has a 2-3-year lead in custom AI hardware. But it is good to see AWS (finally) engaging here, more competition is good for enterprises and fosters innovation as well as price competition… all good trends for enterprises.
AWS launches AI apps – Making it easier for business users to use AI apps is key these days and the strategy for vendors is to help them with specific, targeted apps. With Amazon Personalize, Amazon Forecast (both re-using and building on expertise and technology from Amazon.com), Amazon Textract (an OCR replacement)-and Amazon Comprehend Medical (Healthcare use case with NLP) there has been a broad push into AI apps. We will have to see how well practical adoption will pan out, but interests were big at reinvent.
Database Innovation is alive and well – Amazon execs know how important databases are and how much critical enterprise load is in relational databases. The strategy of AWS is to combat the generic, all-purpose databases with more dedicated databases, all in a suite of database products. Last year Amazon unveiled AWS Neptune, it's graph database, this year it unveiled Amazon Quantum Ledger Database (QLDB – a distributed ledger database, but what a confusing name), what a confusing name) and Amazon Timestream (a time series database). A good strategy that seems to get more traction in enterprises. But AWS knows that entrenched vendors are the challenge, and Oracle specifically has almost become an obsession. Multiple references in Jassy's and Vogels' keynote as well as almost 30 minutes on the topic in Vogels' keynote are almost too much attention.
AWS open sources Firecracker – Influencing industry standards via open source has been the trend of the last two three years. Google has shown the way with Kubernetes and TensorFlow… so AWS did not want to stand back and open sources its serverless framework (Amazon likes to call it virtual machine manager) Firecracker… with Google focusing a level higher currently at the service mesh level, with Istio, this is an open space for anyone. And AWS has good chances given that Firecracker runs Lamba at AWS, and Lambda is the leading serverless platform in the cloud (from a load / uptake perspective). Now we must see if other IaaS vendors will support Firecracker and how successful AWS is at stewarding a strategic open source project. CxOs have accepted open source by now, so it really comes back to show traction in the offering from a roadmap and consumption perspective.
Amazon QuickSight Renaissance – Two years ago Amazon debuted Amazon QuickSight a product that targets to replace the usual BI tools (remember Business Objects, Cognos, MicroStrategy?). The most promising capability was its automatic capability to visualize data. A key vision and step towards user free software, ahead of its time back then. Then it got quiet around Amazon QuickSight… but AWS does not give up easy and came back with a new attempt to win this market, and the value proposition remains strong. Time to look and evaluate this product for CxOs. True to cloud DNA, the pricing is now no longer per seat, but per session (with a maximum of $5 per reader user per month) – an innovation on the licensing side.
AWS Tooling everywhere – Last year Amazon announced its own IDE, a key strategy to get closer to developers. But switching IDEs is not easy, so this year AWS is bringing the AWS Toolkit to PyCharm, IntelliJ (Preview), and Visual Studio Code (Preview). It's important to help developers build apps fast and bringing the toolkit to the 'living room' (aka IDE) is the right strategic move.
MyPOV
Another monster re:Invent with progress across the board. If there are areas where AWS was weaker, or even behind (on premise support, AI hardware), AWS has reacted and is making product available. The AWS Outposts announcement is changing the market for hardware fundamentally. Given AWS readiness to commit to aggressive pricing and ability to manage tight margins successfully, it is bad news for the Dell, HPs et al… and good news for CxOs. It's a great reward for long term AWS customers, who now can move loads from the AWS cloud to on premise – should they need or wish. On the hardware front AWS is correcting its miscalculation from a few years ago when it thought Machine Learning is all about… linear regression. With the adoption of Tensorflow across the board AWS acknowledges that MxNet has not won, but is also not giving up on the toolset yet. Both are important course corrections and innovations that are key to keep AWS in the overall IaaS market leadership position.In general AWS is impressive when it comes to delivering innovation across the board, across the product set. An astonishing feat that the AWS product teams manage to repeat year over year. And that raises the bar for innovation for the overall industry to keep up with.
On the concern side, the AWS portfolio is broader and deeper than ever before. AWS keeps staying with the Chinese menu philosophy, and it means massive choice… but staying with the analogy, when a Chinese restaurant adds a full menu every year, even the most loyal customers will have a hard time to keep an overview of the menu – and manage to repeat-order the same dishes (aka products) at the next visit (aka implementation project). I was not around to ask Andy Jassy my traditional question (Is AWS getting too complex?) but he came prepared (very Amazons style) and negated the question (no surprise). Customers though are struggling to find repeatable success across the vast AWS services portfolio. And cross portfolio QA and testing is getting close to impossible. So, simplification by grouping services, testing them together, aligning roadmaps, versioning and more … must be in AWS future – at some point. To be fair at re:Invent, AWS has made steps to make it easier for customers to run the large product portfolio - with new offerings like Landing Zone (AWS Control Tower); Security (AWS Security Hub); and, Data Lakes (AWS Lake Formation), customers should be faster and have an easier time to setup, operate and maintain AWS offerings. That is a good step in the right direction.
But overall respect to AWS to a massive, well organized and amazing event, sharing the breadth and depth of the AWS services portfolio and innovation. Here is to another year taming the dragon.