The Tides of Compute Are Changing

Since it's December I'd like to throw on my thought leadership cap for a second and take you dear reader into the year 2017 and the years beyond.

First off, Amazon is still 'killing' it, however...

PRIVATE CLOUD STILL CONTROLS 95% OF WORKLOADS.

That's out of Diane Greene's mouth - you know - ex CEO/founder of VMware now head of Google Cloud. One can look at it one way and say - well that's 95% of compute we can plunder or you can take the more realistic viewpoint and say that not only is public cloud 5% but it's going to be less - it's just that we have engrained technology into everything we do now that the volume is so amazingly huge that 5% is still a incredibly large chunk.

This past week has seen quite a few articles such as this one proclaiming Amazon is killing open source.

On the other hand, you also have teams like the one over at Gravitational whose whole business model states there is this return to private on-prem software which we are particularly bullish on as well.

Also, before we get into definition clashes for the purposes of this article we'll conflate 'private cloud' with private infrastructure or if you want to sound official you can check out NIST's definition of cloud :)

Just to set the stage for everyone reading this post - I was on the ec2 bandwagon back in 06/07 when we were building out bluff.com on it so it's not like I'm new to the "cloud" or anything. I also was an early adopter of Google Cloud and vastly prefer their systems to AWS.

If you have ever had the pleasure of talking to a VC or similar in the year 2016 about "cloud" there is still this perverted notion that AWS is going to rule the earth and private infrastructure is a thing of the past. These might be the same people that mistakenly believe open source is some sort of business model.

They obviously haven't heard this from Michael Dell: "They won’t  take the whole market away ... cloud is not just where you do computing, it’s how you do computing."

The fact remains though that private infrastructure/cloud not only still owns the vast majority of compute but will continue to long after Amazon gets it's anti-trust day in court (remember Microsoft's wasn't that long ago). Of course you don't have to tell this to all the hyper-converged and storage vendors and their extensive channel partners.

I feel that when I talk to people about private cloud and the other things that go with it - those who are against the notion are taking a very top-down approach to viewing the situation and completely ignoring the bottom-up analysis. If they chose to do the latter they might find what they are missing out on cognitively.

Did you know a good 10% of the unicorn club is controlled by enterprise on prem software?

When small 20 person software companies are blowing $20-30k/month on AWS you can rest assured the cost equation doesn't add up for many of these larger companies either.

I often get asked why I believe so strongly and instead of pointing out fact after fact - some people don't like basic rules of logic - we quell arguments by pointing at raw physics. There are many companies out there that spew 100s of Gbs/day from their Oracle databases that they simply can not move to the cloud. Marketing gimmicks like Amazon's snowball truck aside this is just not going to happen anytime soon for many companies in this boat.

Engineers build entire teams - scratch that - engineers start entire companies to write software to get compute closest to the data and this problem is getting worse by the day.

This I believe, is one of the stronger reasons why the tide is starting to turn away from public cloud.

Worst still -- device proliferation is happening at a frightening pace. Telcos have actually gotten the foresight on this one and are already trying to be prepared. Long considered 'dumb pipes' they are tired of missing out on the higher layer software revenue that enjoys greater margins and now find their chance in the coming disaggregation - that is the total annihilation of value in their old hardware plays which is quickly becoming 100% software.

You see software not only ate the world but it ate hardware too! Now tack in the coming thundering storm of 5G and you can realize why they are freaked out.

What's really interesting is that more and more compute is going to be happening closer to the data creating a weird new compute reality. The Next Platform people have dubbed it the 'datanet'. That's a little bit cyberpunk, but they are 100% on the money on this one and we are 100% believers in this reality and see this as a huge driver of new business models around compute workloads in the future.

I can't even tell you what this is going to look like but if you think the future is composed of 2-3 big 'public cloud' infrastructure vendors I'd like to make a side bet with you.

Everyone is out to Kill Devops and this is a Good Thing

Companies are still hiring like mad for devs with ops skill and paying for it. However, between the models of serverless && containers we'll see the CFM companies start to get eaten alive or at least be relegated to maintaining LTS. Don't get me wrong - there's going to be some poor lost soul screwing with puppet/chef recipes till they die. After all we still have Cobol programmers and AS400s.

On that note - there's been a lot of chatter about serverless.

Dystopian nightmares of NPMGate aside there are some interesting models out there like PicoCenter that definitely deserve exploration, but let's be honest - that's a research paper - not a product offering and 200ms is pretty rough.

One thing that took me a while to get ahold of with the serverless talk was the economics didn't make a ton of sense to me - every project out there talks about how they don't have vendor lock-in and you can run it on your own hardware, but then isn't that the whole point? Not to have to deal with "the servers" yourself?

Well - turns out that wasn't the point - it was the fact that a lot of companies have a ton of background jobs, events, and ETL and all sorts of crap like that that is completely ad-hoc and dis-organized. Having a common system to contain all of that improves their workflows and that's where some people are finding value.

I'd say the big contender in this space right now is the folks over at Iron. I will state - that between containers, platforms, serverless and unikernels there is an overriding theme being played out here - companies want to kill devops.

This also says something about the sorry state of our infrastructure today. The politicians talk about roads and bridges but no one talks about how crappy most internet infrastructure truly is. You think the Dyn dns attack was special? That attack can be bought/made cheaper/easier than you think you can. With nailpolish. You know how many millions of windows 98/xp boxes are still online in the world and are completely owned!? The correct answer to that question is to put your fingers in your ears and scream - "I don't want to know."

'Tech' Continues to Consolidate

2016 was the tipping point and had some very large mergers. Some had been coming for quite a while other companies started getting acquired left and right. The fundraising climate was probably only felt by those in the moment but I reckon 2017 is when the rest of the people in the ecosystem will start to feel it as well and we'll pick up even more M&A activity. I don't know if we'll see people throwing pink slip parties and abandoning their cars on the highway at the airport but who knows.

Intel was one of the big names to start it's restructuring earlier this year with 12,000 employees laid off. Then Cisco responded earlier this fall with 14,000.

Broadcom snapped up Brocade.

Dell finally completed the largest merger in IT history and proclaimed it was to "Go Big, Win Big".

I don't know what the implications of this are but I'd be surprised if at least one company the size of AirBnB or Twitter didn't implode next year and I think that's fine. Either implode or be bought and then dramatically downsized and prepare for it's long sunsetting - Twitter was already turned down by numerous suitors earlier this year. If it's something like Twitter I think we might be in for a really interesting time period coming up - that is - the return to true technology innovation and less companies merely utilizing technology for scale/distribution.

This will impose interesting demographic changes across the spectrum.

This will obviously also affect the labor market if two or three of these big ones drops. These engineers will have to go somewhere which will probably exert downward pressure on the smaller companies. Now, I'm not an economist so I don't really know what to expect. Does this mean that hiring for companies becomes easier? Maybe jobs for less talented engineers becomes more difficult to find? Who knows, but that many engineers looking for new jobs will do something.

Let's switch gears and let's move our prediction resolution window out a few years.

The Compute Revolution

Uber recently shelled out $680M for Otto earlier this fall.

A lot of people seem to think/agree that autonomous driving is going to need crazy amounts of compute.

Driverless is driving artificial intelligence which is driving very strong demands for compute in the future.

The question here though is - where all does this run? One could weasel out of this and point out everything coming out of Google today is artificial intelligence and machine learning but one must realize that's their data.

One has to go back to our first point that compute needs to live closer to the data.

There are a few other trends I see affecting the big clouds too. First, environmentalists have been strangely quiet about datacenters. Do they not know how much power goes into these things or how incredibly wasteful they are? There is definitely awareness of this issue but it's still a problem. Uncle G is already the world's largest buyer of renewables. Whose second? Amazon.

In 2014 datacenters represented 2% of all energy consumption in america. Chew on that for a while.

When we've got pipelines being protested in North Dakota or the city of Oakland blocking coal shipments what do you think is going to happen here? I think it's only a matter of time before we start seeing some backlash from this.

The other thing is interest rates. They've been so low for such a long time but they won't stay that way forever. This will affect debt service and the associated cap rates for these buildings and for newer investments.

Then we have things like 5G starting to prop out in the coming years. This is going to be pushing tremendous amounts of data to the so-called edge - and we aren't just talking about a glorified CDN folks. This is where I think a lot of compute is moving towards and where a lot of opportunity lies.

Then we have peak x86 starting to really hit home. We see the triumvirate rise of on-demand virtualized GPUs, FPGAs, and the rise of ARM in the datacenter. The Packet.net crew currently is offering 96cores of ARMv8 @ $0.50/hour. ARMs race for real. This might've not been doable without mobile's success but you can be guaranteed SoftBank has some plans now.

This all leads us to wonder - in what alien form does compute look like in the future? Where does it live? Who operates it?

Throw out the old models.

So, these are some of my short term thoughts and predictions on what next year holds and directions we are going to be walking down in the next 5-6 years.

I haven't even touched on software memories, cognitive computing, VISC architecture or anything like that has the potential to inflict it's own mayhem on the markets of the future, likewise I've tried to stay within the internal confines of software as well and haven't approached the external concerns that exist.

Having said that, I do think we are at the start of a new beginning for tech and it's going to be different.