Here is a long and detailed and recent article on serverless. … As you read through it, much of what I brought up stands true.
No, it doesn’t.
Please show us where Mr. Fowler confirms what you said, which was: Serverless actually adds complexity to the architecture, thus even greater need for management of it. He doesn’t, because your repeated statements to this effect are false.
Matter of fact, Fowler sums up serverless’s advantages thusly: Serverless architectures may benefit from significantly reduced operational cost, complexity, and engineering lead time
To be extremely clear, there is NOTHING in Mr. Fowler’s article about implementing a serverless architecture on a private cloud. It is, as I have been forced by Tinker to repetitively state, a public cloud service. Tinker is simply wrong on that.
The article even discusses, as I brought up the use of hybrid (although hybrid can be in the service cloud itself, so it is still third part like AWS giving you the hybrid, but also, as I brought up, the use of serverless without a third-party provider within the on-premise data center.
No, it doesn’t. Tinker keeps insisting on talking about implemention of serverless on private or hybrid clouds as being complex. But, AGAIN, those don’t exist, as Fowler himself states: Also some people use PaaS platforms like Cloud Foundry to provide a common development experience across a hybrid public and private cloud; at time of writing there isn’t a FaaS equivalent as mature as this.
Remember serverless is one example of FaaS (Function as a Service). Tinker, please put this false and misleading argument to bed.
Fowler’s remaining use of the word “hybrid” is not in the same context as Tinker’s. Tinker refers to a “hybrid cloud” whereas Fowler talks about “hybrid architectures” in which different programming paradigms are used to support a given application.
Again, there is nothing in Fowler’s article to support Tinker’s stubborn statements on serverless being complex. Fowler says what I’ve been saying, that’s its LESS COMPLEX for programmers.
Serverless is a lot like Pivotal Cloud Foundry.
Tinker is a bit mixed up here, too. Pivotal Cloud Foundry includes a product they call Pivotal Function Service (PFS), which is an example of a serverless API. There is also a Pivotal Container Service (PCS), which is an API for use within Containers like Kubernetes. So, enterprises have a choice of how you want to use Pivotal. But, it’s incorrect to characterize serverless as “a lot like” PFS.
Serverless also requires less computing power than typical data center functions (this is by design, as for example AWS limits any such app to run time of no more than 5 minutes, and something like 256 mb of space, along with the app unloading if not constantly in use creating latency issues for infrequently run instances).
This isn’t the problem Tinker apparently thinks it is. Serverless computing is designed to be used with an application architecture based on what are called micro-services. The idea is that instead of having antiquated, giant, monolithic applications, solutions are structured in small pieces, each providing a small service (hence, “micro-services”). This is as big a shift in SaaS programming as Object-Orientation was to desktop apps a few decades ago. We can get into micro-service architecture if anyone wants, but it’s not relevant, Tinker attempting to say that a micro-service architecture doesn’t support monolithic applications is like saying streaming videos don’t support physical movie theaters.
It does not sound like serverless is going to replace the containers of the today
In some cases yes, in some cases no. Here’s one case study: https://serverless.com/blog/why-we-switched-from-docker-to-s…
Essentially, this company switched from containers to serverless, and gained high availability, resiliency, and lower costs. Since integration, we’ve taken a serverless first approach; all new services are built in a serverless fashion unless there is an obvious reason not to go serverless. This has helped us dramatically shorten our release cycles, which, as a startup and a SaaS provider, has been hugely beneficial.
serverless comes no where close to replacing what is done in the data center
This is tiring, Tinker. Stop creating strawmen to tear down. No-one has said serverless is replacing data centers or what is done in data centers. Stop making crap up.
I really do not see where it changes the Nutanix thesis
OK, let’s drop the technology and talk trends. Your Nutanix thesis is that enterprises are moving applications from public clouds to private clouds, and so since there are lots of public cloud applications and Nutanix’s HCI helps enterprises setup and manage private clouds, they’ll do well.
My counter argument is that enterprises are not moving applications from public clouds to private clouds en masse. Sure, there are isolated examples, but the real trend supporting Nutanix is movement from old-fashioned, non-cloud data centers (perhaps old J2EE servers) to private clouds. Of course, the mega-trend over the past half decade has been the movement from those legacy data centers to the public cloud, but by now it’s too late to invest in that trend.
So, the question becomes what the TAM for private clouds really is. If it’s many public cloud applications moving to private, then Tinker is right and Nutanix’s TAM is large since today public cloud usage is large. However, if it’s mostly legacy data centers being replaced by private clouds, then the TAM is much smaller and limited to the number of remaining legacy data centers. This is my concern. As the legacy data centers disappear, so will the demand for new private cloud infrastructures. As the public clouds get better and better, it’ll be harder and harder for enterprises to justify private cloud installations.
Note that the rationale given by Nutanix for moving from public cloud to private cloud is mostly based around cost. Yeah, for some enterprises it’s compliance or security, but those aren’t the majority. And, cost is where public cloud features like serverless can help. As the case study I cited earlier shows, going from containers to serverless can greatly reduce public cloud service costs, and enable rapid scalability. That’s all I was saying, that the public clouds are responding to market needs, yet Tinker somehow blows that up into some serverless taking over the world argument, which it never was. That’s just another false strawman invented by him.
Private clouds do have a tough time with scalability, since scaling up means buying additional hardware. This gets us to what may be considered the holy grail of hybrid cloud computing - where you run something locally most of the time, but if demand increases the additional workload moves seamlessly to a public cloud. As I’ve said earlier, this is a nice vision, but Nutanix isn’t there yet.
Back to today’s reality, I think Nutanix’s management is aware of the limited TAM for new private clouds. That’s why they came out with Beam, a tool that helps enterprises understand and manage how they’re using public cloud services. A side effect is that it’ll show enterprises how expensive the public cloud can be, and maybe Nutanix thinks they can use that to sell the customer on Nutanix private clouds. But, migration from public cloud to private isn’t simple.
Finally, Tinker brought up vendor lock-in as a problem for serverless usage. But, while it’s true that using a serverless architecture can lock you into that vendor, and similarly for most of AWS and Azure as well, the same is just as true for almost everything that Nutanix provides. Once you’ve committed to Nutanix’s HCI for setting up server compute, distributed storage, and networking, you’re locked in. Same for using Nutanix’s Acropolis Hypervisor, which is another form of lock-in. Enterprises may be asking themselves whether they want to be locked into Amazon or Nutanix - which would you choose?
I remain skeptical of a Nutanix investment thesis that involves a shift of usage from public clouds to private clouds. Even if that were to start happening, I think Amazon and Microsoft are too smart to let it continue. They’ll adjust pricing and features as necessary. In the meantime, they both continue to make it easier to deploy applications on their public clouds than on enterprise private clouds. Serverless is but one example of that ease of use, which has a side benefit of reduced costs in many instances. Scalability remains an area where the public cloud shines over the private cloud.
So, I have a position in NTNX, but I’m watching it closely. I suspect that like most Saul positions, it’ll be abandoned in not that long of a time as its high growth won’t prove to be sustainable over many years.