A recap of the AWS re:Invent keynotes
We’re returning from three and a half days at the re:invent conference. It was an extraordinary event both in terms of the number of people attending but also in terms of the calibre and diversity of the attendees. I had some conversations where people reminisced about how re:invent has come to be the IT conference and has taken over the mantle from VMWorld. It also helps that the brands making presentations and sharing their knowledge are acknowledged leaders in their industry viz Netflix (entertainment), Merck (life sciences), FINRA (finance), Twilio (messaging), Mapbox (mapping), enterprise IT (VMWare and Workday), McDonald’s (fast food) and more. Every presentation in the keynote has slides filled with logos of major brands in every industry you can imagine. Each of the keynotes—the first by CEO Andy Jassy and the second by CTO Werner Vogels—were about two and a half hours long. Each was filled with product announcements and interspersed with customer presentations, all of which were illuminating and funny at times as well. The Snowmobile announcement, for one, was sheer showmanship (it was driven onto stage) and remains memorable for every attendee. It was also parked right next to our booth in the exhibit hall and had a constant stream of people coming by to take selfies and pose with it.
While both keynotes were interesting from a technology perspective, I felt that it would be worthwhile to condense it down and explore how some of these announcements could be applicable to the healthcare domain.
Possible healthcare impact from major product announcements made and
EC2 instance sizes
EC2 instances that are bigger, smaller and programmable were announced, to satisfy the least and the most compute / memory / storage hungry software. Some of these were geared to keep AWS way ahead of it’s competitors and others such as LightSail (the preconfigured virtual private server - VPS) seemed to be directly targeted at niches serviced by companies like Digital Ocean.
This is important for healthcare and life sciences. As I wrote about before, one of the primary use cases that the healthcare industry is focused on is big data analytics, which necessarily implies large scale data analysis which in turn requires pretty large scale machines. Given that elasticity is also possible with these machines, things such as genomic analysis and outcomes analysis is likely to benefit substantially. At the other end of the spectrum, smaller instances will allow innovation and other groups to dip their toes into the cloud or try out ideas quickly and cheaply.
AI and ML as an API
Rekognition (image analysis and detection), Polly (text to voice with intelligence) and Lex (voice based AI and interaction), all accessible via an API were announced. Machine Learning (ML) via an API and UI was announced earlier with these new products now being able to be used in concert. While some believe that Google has a lead in the AI space given it’s history and access to vast troves of data, AWS seems to be focusing it’s efforts on enabling specific use cases that customers might have in specific domains and offering up the technology it uses internally in products such as Alexa / Echo. All of this seems to be in response to Google’s announcements earlier about the availability of its solutions, it does seem that in conjunction with the data movement and compute solutions, AWS might be in a position to catch up with Google in this regard. It does seem like this is an attempt to also make the technology more ubiquitous and thus the de-facto. I wouldn’t be surprised to see ML and AI models developed by customers made available to others on the AWS marketplace in a little while.
AI and ML is, of course, extremely applicable to healthcare. Reams of data is available within EHRs, PACS, PBM systems, pharmacies and claims management systems. However, analyzing all this data requires the right set of people. Both from a deployment and ongoing system management perspective but also, even more importantly, from the data science perspective. Other industries are also jumping into the data fray which makes it very challenging to staff up on both the devOps as well as the data science roles. The above mentioned solution is an attempt by AWS to try and simplify access to this somewhat esoteric of fields. I would predict academic hospitals with its set of medical informaticists, medical students and researchers, seeing immediate use for this capability. Specifically Rekognition and the ML API…
On the consumer side too, we can expect to see Lex (with voice based commands to summon the nurse, control the temperature, raise and lower the bed, capturing patient feedback and many more) being put to use in various ways in hospitals and Polly on the web or inside clinical applications. The possibilities are quite staggering.
Embedded IoT software aka Greengrass
As a preloaded piece of software onto IoT chips and with pre-built security, AWS Lambda links, local storage and orchestration amongst devices, Greengrass was a major announcement that is very timely. Especially given the recent DDoS attacks initiated via IoT devices and also that the IoT trend is not slowing down anytime soon.
Devices already exist within health systems and have been recognized to be very insecure. The capabilities described can make Greengrass the foundation for another generation of more intelligent and secure healthcare IoT devices.
VMWare Cloud on AWS
This had to have been one of the most significant announcements re enterprise IT. VMWare cloud (vMotion), vSphere (compute), NSX (networking) and vSAN (storage) are all deployed on bare metal machines on AWS. Enterprise IT teams used to working in the VMWare world can now move seamlessly from one to the other. This is likely to speed up adoption of the AWS cloud significantly. There was no mention of exclusivity from VMWare so we can expect to see similar announcements coming soon from Azure and Google.
Healthcare and life sciences are still in the early stages of the move to the cloud but this VMWare cloud on AWS could be just the thing that could act as the catalyst to accelerate that movement. However, the issues around compliance (HIPAA, HITRUST and GxP) remain open and need to be addressed. We do see this as an area of opportunity for us.
Cloud-first data management tools
Step Functions (workflow management), Glue (ETL), Athena (S3 querying) were all geared towards enabling and simplifying data movement and massaging needed to enable analytics and machine learning capabilities. With expanding capabilities, it does seem that AWS could be considered competitive to companies such as Mulesoft.
This is pretty significant in that this is billed only a use basis, which implies that you no longer have to buy expensive upfront licenses but only as needed. This can have an immediate impact for healthcare startups as well as for the innovation groups in healthcare institutions.
Automatic DDoS protection
Shield was a great announcement which essentially provides all AWS customers with core DDoS protection (volumetric and TCP) automatically and for no extra charge. Advanced versions are also available of course at a higher price. But this free offering is a great first step to address the concerns that some customers have raised with us as well. While it makes marketing sense, it also makes sound business sense to AWS because DDoS attacks on any of their customer accounts will have a cascading impact on other customer’s workloads and secondly, there are bound to be issues with the final bill post attacks.
This further removes the security concerns felt by CIOs and CISOs in health systems and plans especially given that basic protection is automatically put into place and leverages the cumulative man-years of knowledge that AWS has in protecting it’s own and it’s customer’s infrastructure.
Easier data transfer to the cloud
I wrote about the challenge that a lot of the customer’s continue to face in moving data to the cloud from their datacenters in the previous post. This has been addressed by using much larger capacity Snowballs with some enhancements and compute ability and for very large data transfers measured in exabytes, a tractor trailer sized device, called the Snowmobile.
Two key points.
With a revenue run rate of $15 Billion—and a projected growth rate of 55% YoY that as a business size dwarfs the next 14 competitors put together—the AWS business is enormous.
The move to the cloud, while not in it’s infancy, is only just in it’s pre-teen years.
We can expect to see AWS as the market leader and the de-facto choice of many of the CIOs in healthcare. Given Amazon’s history of expanding into related spaces, it is not surprising to see it expand into domains led by companies by New Relic, Digital Ocean and Mulesoft. While AWS’ offerings might not be as full featured as the established, we can safely predict that in a year or less, they are likely to be at feature parity which should be a source of concern for those companies.
The healthcare industry can leverage a lot of these new capabilities. We expect to see immediate uptake on the larger machine sizes for big data and analytics, short term uptake on the ML and AI toolsets especially in academic healthcare and medium to long term adoption of the Greengrass toolkit for innovative device development and use.
It will be interesting to see how Microsoft Azure, Google, IBM SoftLayer, and others do to counter next year.