Hacker News Reader

Comment by exelius | Hacker News Reader

They basically make a Java-based application abstraction layer that sits on top of AWS, Azure, VMWare, etc. Basically a platform for managing storage and compute across multiple on-prem and cloud providers. It’s real enterprisey stuff — as in, it’s probably not even interesting to you until you’re big enough to run in a half dozen data centers and within all the cloud providers.

View on HN

For Cloud Foundry, Java is just one runtime target, provided by build packs.

You can run whatever you like, even Windows stuff. You only need to have (a) build pack(s) (aka runtime layer) and a stemcell (base operating system).

CF is like Heroku on steroids or like Lamda+API Gateway+Services for more complex applications.

And Pivotal provides one product of Cloud Foundry ( Pivotal Cloud Foundry), that can be deployed to various Cloud Providers (such as AWS, Azure or OpenStack). There are other vendors, like IBM (Bluemix).

I guess, you mixed it up with the Spring ecosystem, which surely a big driver in the adoption of Cloud Foundry. Pivotal is the core maintainer of the ecosystem.

And no, you don't need to be a big enterprise to run a CF.

I don't foresee companies continuing to pay for the ability to switch cloud providers. Mostly because I doubt the pricing differences between providers will ever be significant enough to justify moving.

I also suspect that providers will not keep feature-competitive with one another. It seems like every month, AWS is releasing new tools, and there's no way that other providers are maintaining this pace. My suspicion is each provider will specialize in non-overlapping domains. So unless your company will always be running a lowest-common-denominator site, you'll eventually encounter a vendor lock-in feature.

Well, a lot of what PCF does is help you map the services available in AWS / Azure to an internal service catalog as defined by an enterprise architecture group. This is really powerful in the “small enterprise” space ($1 billion - $5 billion) because that’s often where the EA value proposition becomes strong enough to worry about.

This is the level PCF plays at; and the level they’re useful, but in my experience Pivotal doesn’t know how enterprise companies develop software. Their scrappy pair-programming Dojos don’t really build systems the way those kinds of companies need to design system requirements.

So in my mind, Pivotal’s consulting services are a bunch of startup guys coming in to the enterprise world telling enterprise developers how to build java services. They’re smart and know what they’re doing, but they’re totally out of touch with the kinds of internal controls, reporting metrics and architecture management that larger companies often require.

I absolutely disagree. Pivotal understands the "needs" of the enterprise world and they prove them wrong at every level. Its that "enterprise world" that needs to change in order to survive or at least to be able to hire new people and to manage the ever growing complexity of enterprise architecture. That's whats referred to as "Digital Transformation".

Its just extremely hard for the old "enterprise world" to adopt to change that is inevitable.

Funny, I see the opposite trend. Technologies like kubernetes enable immutable infrastructure that’s not locked into a single cloud. There’s no reason you couldn’t build a system that regularly and automatically buys the cheapest capacity, sets up a kubernetes cluster, and deploys the latest build to it.

Do they make it easy to migrate from tools like AWS EMR, SageMaker, or ElasticTranscoder? It's one thing to be able to migrate VM images or microservices, but in my experience, the big draw to AWS is the features they offer beyond that.

It's not always about the price. I've come across government RFP/RFQs that require data to be stored locally (through cloud or self-hosting).

The main benefits of PCF are not cloud vender lock-in. Where I saw it used, they weren't even on AWS, but using PCF to migrate to microservices. Microservices were written with in Java with Spring Framework, another pivotal product. Pivotal consulting also taught them how to do pair programming, agile, and other "modern" software techniques.

one example might be financial services which may get regulated to be multi-cloud.

Also DR.

I worked at a place that had recently switched over to microservices in the Cloud from a monolith on premises. Rumors were we were spending a lot of money on our cloud providers. I heard other rumors that pricing got less competitive after awhile because the size of our bills implied that we were locked in. I'd guess that you could get more competitive pricing if you were able to easily switch between cloud providers.

Platforms like this also allow you to choose between using Amazon RDS or a homerolled Postgres DBaaS in the same way. You can make a strategic/tactical decision to implement certain parts of your infrastructure yourself vs outsource the capability entirely.

But again, this is really only useful if you’re big enough to have many application groups developing for a bunch of different business units. I have yet to see a good use case for PCF outside enterprise IT departments.

On the contrary, many major companies, especially finservs have multicloud strategies to avoid lock-in or concentration risk (a real concern of regulators). Platforms like pcf and K8s allow for (easier) porting between clouds rather than building directly to cloud provider api's.

The wildcard here is acquisitions. Large companies make a lot of acquisitions, it can be one of the only ways to grow some companies. Each acquisitions brings its own potentially locked in cloud ecosystem.

Large competent software companies like Facebook and Google have still not completely integrated Instagram and Waze respectively for example.

Only a couple components are Java-based. CloudFoundry started out almost entirely Ruby, but most of it has since been rewritten in Go.

At least when I was working with them last (6-9 months ago) Java had by far the most platform support as a language. I wasn’t talking about the language the PCF code was written in, but rather the development language you use to write services. I realize they support a half dozen plus languages, but Java was by far the best supported in my recollection.

Reply on news.ycombinator.com