I recently helped out with the inaugural Serverless Conf, and gave the closing remarks. The event was held in a warehouse in Williamsburg, New York. Organiser Ant Stanley of ACloudGuru did a great job of organising his first conference.
There are plenty of smarter people than me explaining what “serverless” is – Paul Johnston, who spoke at the conference, has a great blog series on the topic. To me however serverless computing is primarily interesting for three key reasons:
- On demand computing is taken to its natural conclusion – paying at the point of customer value. AWS Lambda, the first mainstream serverless platform, runs compute services in response to events – you write and upload code to AWS Lamda, associating it with particular AWS services, but what is kind of mind-blowing is that you don’t pay for the services until they are actually invoked. A startup can build out an entire infrastructure, without paying anything for it until the customer does.
- The atomic unit of compute is the function. Just as we begin to understand the implications of container-based computing, in which the atomic unit of compute is the application, comes a new model to consider, in which the individual function is the service. As with microservices this kind of approach creates coordination overheads, but it also offers a great deal of flexibility.
- You don’t need to think about running containers, any more than you do running an operating system when developing to a PaaS. Unlike containers, which have been associated closely with microservices, serverless is more like PaaS in abstracting all infrastructure required to run a service. Serverless certainly doesn’t obviate the need for ops, but it does exemplify a more PaaS like model.
The chart above suggests a couple of interesting things about the discontinuity in 2007. Deal size shrunk, while average number of deals dramatically increased. Why? The obvious answer is the launch of AWS, and the attendant creation of Y Combinator and the accelerator model, rather than traditional VC-driven company creation. Given that serverless hasn’t exploded so far, it might seem overblown to draw comparisons with the introduction of AWS. And yet. And yet. There is no doubt that today smaller teams can make a bigger impact than ever before. Consider Instagram or WhatApp, which sold to Facebook for $1.9bn with only 55 employees. Serverless attempts to industrialise developer impact- if we’re going to see the first single employee billion user multi-billion dollar valuation startup it’s likely they’ll be building on something like AWS Lambda. Could we see another similar discontinuity to the chart above, with deal size cratering again, and many more flowers blooming? It’s a reasonable scenario. Certainly Amazon’s serverless implementation would hope to enable that. Serverless is fundamentally about permissionless, the ongoing direction of travel for all IT decision-making.
As Sam Kroonenberg, CTO of ACloudGuru put it:
“AWS Lambda is to compute what s3 is to storage.”
AWS is pretty confident about serverless advantages. Its Serverless GM Tim Wagner made it pretty clear that he expects serverless to put severe pressure on orthogonal markets – such as Docker-based infrastructure, partly because he expects the model to drive more testing into the cloud. Serverless is not solely an Amazon phenomenon. IBM has adopted an open source model with OpenWhisk, while Google is building a set of services around Firebase, originally a database as a service it acquired. Microsoft has taken a hybrid hosted/open source approach with Azure functions.
Competition has been sufficient that Amazon took the surprising step of announcing it was to open source a serverless framework called Flourish. On that more details as they emerge.
Joe Emison works at DMGT building B2B applications and services. He gave a great presentation which was frankly pretty scathing about AWS.
“In this case i don’t think Amazon has the best option for any of the services out there. It’s about the front end. AWS serverless is largely about back end processing, which we have largely outsourced. To me, the advantage of serverless is the key is faster iterations, and a move to fatter clients”
30 day trials of SaaS products that expire before you’ve had a chance to schedule some time to actually try them. Such an anti-pattern.
— Stef Lewandowski (@stef) July 1, 2016
Surprised you didn’t mention how serverless rewards massive operational efficiency of hyperscale. it’s a byproduct of cloud at a different unit of scale. by being that good at offering images you become able to offer a unit of compute at a much granular level. which means there’s an operational aspect to evaluating services, not just the technology. Can a business ever be efficient enough to make running serverless themselves economic? I’m doubtful, frankly. A model where essentially putting images on machines is something we do very fast and we don’t need to pass on the cost.
John Evdemon says:
July 1, 2016 at 7:36 pm
Great post. Note: Azure Functions runtime, templates, UI and SDK are open source. Azure Function runtime will be portable so you can run Functions anywhere – on Azure, in your datacenter or other clouds.
jgovernor says:
July 1, 2016 at 8:21 pm
thanks @johnevdemon appreciate the comment and clarification
clive boulton says:
July 1, 2016 at 8:41 pm
Serverless coordination overheads could lead me to drink, or lead me to use blockchain to role my transactions commits over distributed systems.
Mary Branscombe says:
July 4, 2016 at 5:17 pm
If I wanted serverless plus distributed systems, I’d stay on azure and use DocumentDB for the distributed bits (an Azure service I think is as under estimated as Functions), or add Flow for workflow between Functions if they needed services as well as storage. Matching the nicely granular unit of compute with integration tools. Also, my last line is a quote from the Functions team to me about their business model;I think it’s the kind increase in scale where it becomes a step change.
Sinclair Schuller says:
July 6, 2016 at 9:14 am
Are many people viewing serverless as an approach to replace existing computing models, or a pattern to supplement them where appropriate? It seems that where a reactive, event driven solution is the answer, it’s 100% the right pattern. The applicability of serverless seems important, but possibly narrow in scope
jgovernor says:
July 6, 2016 at 10:20 am
great point @sinclair. we have a wonderful varied toolbox, and of course no one tool in it meets all needs. just as monoliths have a place, alongside microservices, so serverless is particularly apposite for reactive event driven apps. all that said, until developers really get to it, it’s hard to say how narrow the scope will be. you’re certainly on point that use cases seem a bit limited at this point. It was a running joke at serverlessconf that every demo or hello world was a photo or video processing app. But there are startups going all in on serverless- notably ACloudGuru – but then again, they’re a online training app using… video! 😉
Dmitri K says:
August 12, 2016 at 6:43 am
I thought both apps that Joe Emisson talked about were neither video nor photo..
jgovernor says:
August 15, 2016 at 1:47 pm
@dmitri yes for sure. the point rather was that almost everyone was doing those.
The serverless economy – why you should care about serverless - Enterprise Irregulars says:
July 6, 2016 at 5:21 pm
[…] Tweet […]
This Week in Cloud: May 13, 2016 says:
July 8, 2016 at 2:27 pm
[…] The Serverless Economy – Why You Should Care About Serverless […]
La próxima frontera (otra más): la computación sin servidores | SmallTechTalk says:
July 16, 2016 at 7:48 am
[…] Lo único que tienes que hacer es proporcionar el código. La computación bajo demanda llevada a su conclusión natural, dicen. […]
Conversational IoT–’SUP? - Enterprise Irregulars says:
September 13, 2016 at 4:22 pm
[…] infrastructure requirements? We talk a lot in tech today about event-driven reactive platforms. The serverless economy is going to come into its own here. An architecture based on waiting, where you only pay at the […]
Breaking the glass – a way forward for Google Cloud - Enterprise Irregulars says:
March 20, 2017 at 4:51 pm
[…] With Google Cloud Functions, the issue is still live because the product is still labeled as a beta. Google may have invented the notion of a perpetual beta, but it’s not something enterprises are ever going to love. Note to Google – production mode soon please. Cloud Functions aside, Google has one back end as a service, or perhaps that should be backend as a serverless – that developers have bought into – Firebase. The platform is a really good beachhead for Google, especially now Parse is gone. If Google acquired Auth0 it would be in a position to start building a suite of best of breed serverless platforms, that even Joe Emison might approve of. […]
Lambda kicks in, a serverless world made of messages - Enterprise Irregulars says:
May 9, 2017 at 5:06 pm
[…] is a great model for efficient cloud computing. I have written before about how serverless is as much an economic phenomenon as a technical one. Please check out this great Trek10 post for more on that (I totally stole the […]
So about OpenWhisk progress. What’s up with serverless frameworks - Enterprise Irregulars says:
June 26, 2017 at 5:30 pm
[…] using OpenWhisk for check clearing, doing all the image processing. Image processing is very much a hello world kind of app in the Serverless ecosystem, but check processing is an application with a more immediate revenue […]