So here are my top 3 takeaways from the conference:
- Lambda pushes the envelope – The Lambda announcements tops the top 3 pretty easily in my view. Anyone who has ever tried to do anything automated on large volumes of data, while streaming at the same time and not knowing when and how often an action will have to be taken, will be intrigued by Lambda. In the past only database triggers combined with nifty CEP capabilities would come close to what AWS has announced with Lambda. Would love to understand the architecture behind it – not sure if AWS will (ever) share, but basically it must have added a layer to its servers and have a superfast code propagation library for the Lambda code. With Lambda AWS solves a fundamental challenge and problem of 21st century, next generation applications and uptake will be interesting to see.
But the capability comes with a price, it is AWS proprietary, so lock in considerations will be a factor. But in the meantime the capability is pretty unique in the cloud and with that, pure lack of options, don’t raise the proprietary question (yet). To be fair, anything of the scope of Lambda will have to be very much embedded in the cloud architecture of every provider and with that likely becomes proprietary. Notwithstanding the desire of vendors to be proprietary to experience lock in benefits. It will be interesting to see how Google, IBM, Microsoft, Oracle et al respond (alphabetical order, don’t read anything into it, please). And then maybe the OpenStack community will come up with a new project? It will have to do so very quickly.
- Amazon RDS for Aurora a sweet spot move – Already a day earlier AWS announced Aurora, bringing HA capabilities to MySQL. I call it the sweet spot as the ‘owner’ of MySQL, Oracle left a void between what MySQL can do and where the Oracle RDBMS comes in. AWS now smartly goes after that gap, there are probably a 5 digit number of MySQL based deployments out there that can use this capability (purely my best educated guess). The performance numbers of 5 (or was it 6) million writes a minute and 30 million reads a minute are very, very good for a version 1 product. Adoption of Aurora will be interesting to see in the next quarters. And even more the reaction of entrenched HA RDBMS vendors, let’s start with Oracle and IBM. AWS pointed out that Aurora is engineered to work at 10% of the cost of established HA systems, definitively an attention grabbing fact, but the established vendors can reduce prices, too. And then let’s not forget, RDBMS R&D is not trivial work, as Amazon knows and shared, as it took 3 years to get here. More on Aurora here.
- More AWS DevOps Capabilities – Largely on Day 1 unveiled already, AWS now gets a leg up on DevOps automation with the productization of the – so far – internal Apollo tool and more. An important move by AWS to increase developer productivity. But also a challenge for the DevOps vendors, as Amazon largely stayed out of that business, till now. Similarly like Lambda it closes AWS a little more. But then it’s probably a logical choice, as AWS probably did not want to share the internal workings of many of its products. The consequence is to provide tools to ease code deployment complexities, offer better automation (like some competitors already do) and on the flipside become a little more a black box. But if one takes into account what AWS has become – probably a better choice by AWS. At the end of the day enterprises want to build, run and maintain applications on AWS, the inner working becomes secondary, as interesting it is for the techies, geeks and pundits. While I give kudos to AWS for sharing more about the basic inner working (read here) – I cannot imagine a CIO, CTO really caring about how e.g. Lambda code is executed and definitively not how to deploy it. More on all the ALM tools here.
Tidbits
- Amazon EC2 Container Service, Docker of course – No cloud conference without containers, and most likely Docker. No difference here, Docker CEO Gollup on stage, and AWS announced the EC2 Container Service, shown in a pretty long demo, as if the audience needed re-assuring that AWS is serious about containers. More here.
- Engineered systems – For the first time AWS has gotten a hardware maker (in this case Intel) to customize key components, in this case the Haswell processor for the new AWS machines. A sign of the sway that AWS has with the hardware suppliers, but then also a necessity for AWS to do, as some of its competitors build their own chips (e.g. IBM and Oracle). In my view a validation point that engineered systems are proving their point all the way to even otherwise very much commodity hardware oriented providers like Amazon. And then some people will point out that AWS was in that business already – see the massive HDD server AWS designed and shared in James Hamilton’s presentations.
- AWS Service Catalogue provides more control – Long awaited, finally here – the tool for CIOs to enable certain services, on certain resources to control a finite cost budget and satisfy security needs. Both key steps for AWS to become more attractive (and viable) for enterprises to use. More here.
- Security remains paramount – As the market leader for public cloud, AWS keeps also carrying the security torch. The new AWS Key Management Service was one of the most interesting announcements for security minded (European) customers. With AZs in Frankfurt and AWS Key Management a European CIO / CTO can convince an executive board concerned about security and NSA activities to consider AWS. More on Key Management here.
And as a logical next step AWS Config (in preview) will help to give a certain level of confidence to control the ‘human factor’ on the security side. Even though AWS does not announce roadmaps, it’s good to see the steps it’s taking to become more viable in the enterprise. And then classic ISV, startups and other users will approach the better monitoring, too.
- Wait – no price cuts? – This marks probably the first re:Invent conference with no announced price cuts. AWS loyalists were quick to point out that there are reasonable cost reductions in many of the new available and announced products – but it’s different than saying we are reducing service x by y%. And if memory doesn’t fool me we also did not hear about this is the xth price cut in AWS history. In my view it means two things: Price cuts are no longer event, but competition driven and AWS is becoming more a value than a cost play. Both are good for the market and cloud customers.
MyPOV
AWS has certainly moved the yardstick, good move of adding more transparency. And while last year I was surprised that AWS was moving into more higher level services (e.g. Amazon WorkSpaces, Zocalo) it is now clear that behind the scenes it was working hard on building capabilities now shared like Aurora and Lambda. It’s the opposite nature of SaaS vs PaaS apps, that SaaS apps can be built faster than key PaaS capabilities, but then SaaS adoption takes longer than PaaS adoption (once there). We see the same in the industry at Oracle (typed on Wifi free plane so check my OOW14 MyPOV here).
And a good moment to remind ourselves that Amazon is also the retailer, even though the AWS folks are likely to deny it – Lambda is huge for Amazon.com. So always good to spend a moment to think what technological capabilities Amazon.com needs and overlay its AWS product roadmap with that.
Moreover, it’s clear that AWS is more and more becoming a PaaS. In the past AWS was a ‘Welcome everybody’ place and while the Welcome sign has not been taken down, previous products that were and are running on AWS are likely to run in other clouds at some point in the future. Enterprise customers know that likely it will be better to run Microsoft products on Azure, Oracle products on the Oracle cloud, IBM product on the IBM cloud etc. In that sense Aurora is a necessity for AWS, as with no RDBMS offering it’s tough to get enterprise apps load, the NoSQL movement non with standing. Aurora is actually a great proof point that the RDBMS as we know it isn’t dead for quite a while.
And PaaS vendors attract developers and load by unique features, Lambda being the prominent example of re:invent 2014. And in most cases these unique features come with lock-in and the longer the absence of standards to provide the same persists, the more likely market success and no other alternative to do the same. The next quarters will show what the competition will do to respond to Lambda, my best guess is Google being the ‘first responder’ – both from an architecture and mindset perspective.
And lastly – when you have unique feature – you need to be less transparent to enable it – both to protect IP and hide complexity. A tradeoff I expect to be accepted by AWS customers, the more the longer the vendors who are playing catchup in the cloud game take to respond with attractive and viable alternatives.
So in the last consequence AWS must build its own technology stack – it has its own unique demands as e-tailer, most of today’s large technology vendors running on AWS will move to their own clouds, all serious AWS competitors have their own technology stack anyway – so from that perspective Aurora, Lambda (and more) are logical steps. What could come next? 21st century BPM, ‘liquid’ RDBMS / NoSQL capabilities, AWS ‘universal apps’, mobile development kits, electronic ad capabilities, credits and payment, development tools etc. are all areas where AWS can (and likely) will go. But don’t expect a confirmation or a roadmap on this.
In the meantime congrats to AWS for having moved the yardstick once more, and being the vendor enterprises simply must evaluate for public cloud projects and benchmark against for any private cloud plans.
