Recently a post from @Gigabarb popped up on the ole’ Twitter that started a micro-storm of twitter responses.
Ok tech users. Who out there is using a PaaS? For realz. Serious question.
— Barb Darrow (@gigabarb) November 21, 2013
This got me thinking about a number of things and I started to write her an email specifically, but realized I should really just blog it. After all, the topic is actually part of what should be the public conversation. It’s about the changing world of technology, which we’re all part of…
First Topic: Usage of PaaS
Barb, just shortly after the tweet above was posted, this other tweet altered what information I might provide her. @TheSteve0 had responded with some items, which @GigaBarb then responded with
@TheSteve0 those names are the same … looking for more…thanks tho
— Barb Darrow (@gigabarb) November 21, 2013
Now, not to pick on OpenShift & Red Hat (the effort @TheSteve0 is working with), because they have a great open source effort going on around this PaaS Technology, but Barb had a point. If Cloud Foundry responded with something like this, she’d still have a point. The only companies that continually sign up new companies is AWS & Beanstalk (ok, so they don’t call it PaaS, it gets you to the same place – arguably better than most of the others), a little bit at Windows Azure and a few companies pop up every once in a long while that might take Cloud Foundry or OpenShift and run with it. Most of the early adopters are already on board and most that might get on board are still mostly just waiting in the sidelines.
This fact is frustrating for those
in the space that want to see more penetration, but for those that arent’ technically in the space, it seems kind of like ASP. Oh wait, I should add context now, ASPs as in Application Service Providers. The technology from the beginning of the 21st century similar in many ways to what is dubbed SaaS now. At the time it could have been revolutionary. However at the time nobody picked it up either. This is similar to what PaaS is seeing. However…
A Hypothesis of What Will Happen to PaaS Tech
I have a theory of what will happen to PaaS Tech, it is similar to ASP Tech. PaaS will keep plundering along in odd ways, and eventually one day, it will become a mainstream tech. Right now however it will remain limited. In that same turn, by the time it becomes a common tech, it’ll be called something else.
Here’s a few reasons. One, is that many developers see PaaS and their response, especially if they’re seasoned developers with more than a few years under their belt, is to respond will immediate apprehension to the tech. It removes key elements of what they want to control. It hides things they can’t actually get to and it abstracts in ways that don’t always make sense. The result is that many senior devs stay away from pure PaaS offerings and instead use it only for prototyping, but production gets something totally different. I’ve been there more than a few times myself.
However, the result of what most senior devs end up with, when they get their continuous integration and development environments running at full tilt, is exactly what PaaS is attempting to promise. There are some companies, with senior devs, and extremely intelligent members that have taken PaaS and effectively implemented it into their continuous integration and delivery environment giving them strengths that most companies can only imagine to have.
One of those companies is lucky and smart enough to have Jonathan Murray @adamalthus heading up efforts. On his team he also has Dave McCrory @mccrory and Brian McClain @brianmmclain. To boot, they are close to the Cloud Foundry team (and @wattersjames, who cuts a path when there are issues) and keep a solid effort going working with key partners such as @Tier3 (now part of CenturyLink) and other companies that help bring together one of the most strategically and tactically relevant PaaS deployments to date.
Other PaaS deployments are questionable for various reasons, they’re trying, but they aren’t there. At least not the types of companies and efforts that Barb was looking for. So really, if there is another out there that’s hiding, but wants serious street cred. A boost to hiring serious A grade talent, and to push forward past competitors, please let us know. Let me know, let Barb know and let’s hear about what you’re doing. If a company is hiding their implementation and doesn’t want to be part of the community, then fine, they can stay hidden and not gain the benefit of the community that presses forward beyond them. But I would love to hear from those that I might have missed, that want to push forward, so ping me. Ping Barb, we’ll get word out there and get developers checking out and making sure your company is getting it done! 😉
Second Topic: PaaS on PaaS and Start Docker
PaaS is nice. If your company can get it deployed and use it effectively, the you’re going to push forward fast in many regards. Deployments, savings, code cleanliness, effective separation of concerns and abstraction at a systems level are some of the things you can expect from a good PaaS implementation. Sometimes however, as the senior devs I mentioned pointed out, you give up control and certain levels of abstraction. However almost all senior devs understand that they want the ability to abstract at the levels that PaaS enables. They want to break apart the app cleanly at the system level from the software level. No reason for an app to know where or what a hard drive is doing right? That’s a rhetorical question, onward with the topic…
Docker has entered the market with a BOOM, part of the abstraction level that enables PaaS tooling in the first place. This tool enables a team to jump into the code or to just deploy the tool to abstract at a PaaS level, but to build the elements that they need specifically. The components are able to be brought together in a composite way that provides all the advantages of PaaS, while put together specifically for the problem space that the team is attacking. For environments that don’t make cookie cutter apps that fit perfectly to PaaS tooling as it is, that needs that little bit extra control of the environment, Docker is the perfect tool to bring those pieces together.
So really, is Docker and containerization that new word (from a technically old tech! lolz), that new tech, that’s going to bring PaaS into the mainstream as the standard implementation? Is it going to make PaaS become containerization when we developers talk about it? It could very well be the next big step. It could be that last mile coverage that devs want to push environments into a PaaS Tech ecosystem and make full use of hardware, software and move to the next stage of application development. Could it? Will it?
Personally I’m ready for the next stage of the whole PaaS thing, are you?
Next up on other thought patterns, WTF are people using Oracle for still when mariadb and postgres mean their freedom to innovate, move forward and surpass their competition.
14 thoughts on “Is PaaS Tech Still Around? Maybe Containers Will Kill it or Bring it?”
Pretty much all the PaaS players implement a solution for vertically integrated.business, not much different than MicroOracleIBMsoft enables. So devs end up fighting to abstract the Borg. Posit the real problem to solve is horizontal businesses. Over in the consumer internet space they are thrashing enterprise space with VC capital and IPOs. I think this is why Docker looks so attractive.
Another sign is the mind share Akka is attracting. Many sites are being back-fitted for horizontal business.
PS. Ward Cunningham told me to think differently 😮
Hahaaaa, I’ve had very similar discussions lately with Ward. It always helps me to straighten out some of the fogginess of the clouds to talk with Ward.
As you point out, it’s all about horizontal business implementation. The MicroOracleIBMsoft all suck up inordinate amount of cash promising this and often don’t deliver very well. Leaving a gaping hole that Red Hat, Pivotal (or EMC, VMware or whatever the parent company that gobles them up when they make money) and even some others could leap into and make a killing all while significantly cutting the costs and time to market that those big horizontal disasters that MicroOracleIBMsoft all promise.
btw – I’m totally using your MicroOracleIBMsoft everywhere. That’s a great way to refer to the giants that are really starting to feel the disruptions of the PaaS/Container etc space of devops and related. 😉
Also – Akka – Hell yes. 😀
Alas, free vs paid? “Azure’s PaaS services seem to be doing the job.”
Two new-to-the-public PaaS stories defining how traditional businesses are using PaaS to reinvent their business model and service deliver:
I think that users and organisation are continuously looking for an easy way to deploy apps on a cloud.
The current PaaS approach of abstracting the entire infrastructure is only one option to deliver that, but apparently one that is too opinionated and therefore too restrictive.
Most users are looking for the simplicity of PaaS but are not willing to give up the control over the language of choice, the stack that often comes with many of the PaaS offering.
I think that what were seeing is not the end of PaaS but rather a shift toward more open application deployment frameworks. Amazon OpsWorks being one of them which basically put DevOps and Application deployment together. Docker is another interesting example on that regard.
The two represent a more broader trend in which people and organisations are looking for a framework to build their own PaaS rather than pre-packaged solution.
I wrote few posts recently on this subject that you may find relevant:
An Application Centric Approach to Devops – in which Chef and Cloudify is used as a mean to abstract the infrastructure to simplify DevOps continues delivery processes.
My specific note about CloudFoundry and OpenShift in the context of OpenStack
re: Cloud Foundry & OpenStack breaking up, and the OpenShift and OpenStack relationship…
Overall it seems there is a need for OpenStack to have a type of PaaS style tooling. In many ways, why not a docker container type model of self-paas build outs? It could very well work much better than trying to put something on top like Cloud Foundry or OpenShift. Especially being that both are heavy versus what Docker (or any container) is.
As you point out in the article about Cloud Foundry & OpenStack and the related break ups and Alex Freedland points out in his somewhat hyperbolic statement that all PaaS vendors are doomed (re: http://bit.ly/paasisdoomed)…
PaaS tooling is getting pushed against heavily with IaaS moving up stack, or one could maybe just say that pure infrastructure is becoming less relevant as developers and operational teams are getting more and more used to dealing purely with compute, storage and related networking capacity. Many teams that do start with IaaS almost immediately push for the commoditization of the next immediate layer of the stack, which puts them squarely in PaaS tooling space – irrelevant of actually deploying something like Cloud Foundry and OpenShift.
That brings me back around to finding it rather intelligent that AWS decided to NOT call anything they build on top of the infrastructure of EC2, S3 and such PaaS. It really doesn’t make sense, it’s all services and all kind of converges into a single tooling stack that is there to enable application development and customers to get things as fast as possible.
Which also is kind of the summary statement, “it’s all about getting developers and customers something to work as soon as possible”. That loops me back around to the main topic of the article, what drives that faster for more customers – container tech or paas tech? Is PaaS really doomed as Freedland points out or is it going to continue in parallel with containers? Is Red Hat’s approach of bridging those two tech pieces into a single tool stack offering the best of both worlds? (I’m sure TheSteve0 would say yes, right Steve?) 😉
Just want to say that those are the only companies I can talk about publicly. In terms of POC and actual deployments I would love to talk about others.
We are really about 2 years in for PaaS. Think about that for the internet or even social networking. Would you have said 2 years after twitter or facebook launched that people wanted something more or just that people had to get used to new ways of thinking.
Finally, I love docker and I love that they are basically merging their code base with OpenShift. So if you love docker, you will love OpenShift even more as we bring our code bases together.
Actually one last point. If you know the sales and deployment life cycle for enterprises, the actual uptake for PaaS is astonishing. Even if you say PaaS is 3 years old – look at the people I listed and see what other tech they adopted that quickly. Instead, I see these large enterprise early adopters as being indicators that the next 3 years should be very interesting in the PaaS space.
I agree in large part, and have seen similar trends myself, which is awesome. Enterprises are prime territory for PaaS tooling, especially since they’ve been trying to reach that level of deployment acapability for years but they generally can’t accomplish things like that themselves. It usually takes an outlier entity to disrupt the business as usual approach in an enterprise. PaaS is definitily setup for that, however I still wonder if container tooling doesn’t eat into that for some companies since it’s so much more lightweight and allows great focus for some enterprises – especially if they’re sold on a non-polyglot type environment. For those companies things liek a container tech and rolling their own PaaS tooling or going with somebody like Apprenda might even be advantageous.
However I’d admit I personally am the open source PaaS camp. I’d always suggest a company maintain a polyglot environment over trying to stick to one framework or stack. I’ve seen too many companies already burned by sticking to just .NET or just the Java ecosystem without branching out. These companies might get some benefits, but have never seen that be better than using the right tools for the job. In this case, using a polyglot PaaS like Red Hat’s OpenShift that you guys are building or even Cloud Foundry and that ecosystem is a superior option. Companies that are polyglot, and have a good culture, are almost always doing better in their Enterprise IT than companies that have good or bad culture and are pushing a single stack.
Good points all, I’m very curious to see how Docker & OpenShift travel in their respective paths going forward. I’m sure, as you point out, there will be some solid convergence and also some divergence in how people use the respective tools to manage their environments. I think a lot of it will continue to disrupt too, just as the standard disruption is detailed in Innovator’s Dilemma.
I have a problem with your description of PaaS and how the role of containers.
1) First, let us discuss the description of PaaS. Your definition is based on PaaS as defined by early market pioneers like Google App Engine and Heroku. They argued that the restrictions put in place by them will help achieve operational efficiencies. However, modern enterprise PaaS offerings like OpenShift and others has shows that you can offer the same kind of abstraction offered by GAE and Heroku without the restrictions they have in place. In doing so, these PaaS offerings leverage the operation efficiencies at the IaaS layer to offer abstraction without unwanted restrictions.
2) Second, I completely disagree that containers alone will replace PaaS. It is like saying VMs maketh the cloud. You need orchestration and management pieces for containers and additional developer tools. The whole thing makes up for PaaS. Role of containers in PaaS is akin to role of VMs in IaaS.
My 2 cents.
I’ll hit on these backwards.
#2 -> My actual notion isn’t that they’ll replace PaaS, but that they could dramatically decrease the need for cookie cutter PaaS style offerings for large enterprises that need to structure their PaaS as they see fit. Many enterprises, especially Bank of America, Citigroup and other companies (and I say this after building and running environments in these institutions) have ecosystems that are already PaaS in almost every way. For some, the next logical migration point for them is to move to containers. They don’t need other’s PaaS tooling like Cloud Foundry or OpenShift, but just better, clearer abstractions that they can implement in a more reliable manner. This doesn’t preclude that some of the departments couldn’t use OpenShift or Cloud Foundry to catch up to the more developed departments. Because it would be a perfect use case. The real question is, where is the real traction – at offering the whole PaaS offering or providing containers that actually provide the ability for large enterprises to more accurately, efficiently and reliably use their existing IaaS investments?
To summarize. I don’t disagree with what you’re saying, I only posed questions and intent to question in regards to which way the market is going. Also what the market is calling or considering PaaS or not PaaS. It’s becoming more and more of a meaningless marketing term just like “cloud computing”. The later, which effectively means the internet these days.
#1 -> My definition of PaaS… I don’t really have one. I work from one or the other definition. I tend to go with those that paved the way – which happens to be Heroku, AppEngine and those. I understand their arguments, and the later arguments, but it doesn’t really change the questions posed above. What is the direction of the market and what would or what will really be “…the next stage of the whole PaaS thing…” Because as I pointed out a minute ago, it doesn’t matter much what we define PaaS as now, it’s becoming and old marketing term and whoever redefines it as traction is gained, will likely become the Enterprise leader in that space. It could be anybody from Red Hat to Pivotol to Apprenda. Who knows. It matters less – unfortunately – on what the tech actually is and who actually takes a leadership position in the industry. Whoever does that will end up defining what these things are actually called 5-10 years from now when we look back.
At least, that’s what keeps happening. 😉 re: ASP (Application Service Providers) are at this point almost entirely forgotten.
btw – Thanks for commenting Krish! Good thoughts and solid disagreements. I appreciate you calling me to task on the definitions.
With regards to #2, I think we both are in agreement to a considerable extent. If an organization has already invested heavily on infrastructure services, it makes sense to use existing tooling to offer the abstraction but for organizations moving to cloud now, it makes more sense to use tools that offer the level of abstraction out of the box. If they can take the set of tools that increase their operational efficiency by an order of magnitude, what is the point in investing on services that requires additional operational overhead. I would argue that if PaaS providers do a good job of educating the users on the value offered by PaaS, I think the industry will benefit better due to better efficiencies. I have spoken to large organizations (NDA stops me from naming them) who had personally told me that they would have gone the PaaS route if tools like OpenShift and others (those PaaS that offers the flexibility with higher efficiencies) existed when they were plotting their cloud strategy. It is about not knowing that they can do better than anything else. I would say it is upon the PaaS providers to educate on the value PaaS offer than IaaS+.
The term IaaS, PaaS, etc. will eventually go away. I am not talking about the terminologies but about the underlying technologies. Eventually, it is going to be an enterprise platform (in tune with Jonathan’s composable enterprise concept) and such a platform can be built using many ways. Some approaches offer higher efficiencies than others. A customer has to be well educated on various plumbing tools available to build their platform before they venture on their path.
It’s worth noting that there are two sizable and successful places where PaaS is the norm: small-medium website hosting and the *entirity* of shared HPC (like top500).
Shared HPC? Not sure I’d put that in the same category.
However small-medium website hosting, I’ll buy that. However, pending the questions around Cloud Foundry, OpenShift and others I’m not sure if that gives us a direction for the future of those products or the intended space for those products.
Comments are closed.