Hacker News new | ask | show | jobs
Microsoft announces Service Fabric for highly scalable cloud services (techcrunch.com)
26 points by mwadams 4077 days ago
5 comments

Another article by Mary Jo Foley of ZDNet:

http://www.zdnet.com/article/microsoft-readies-first-develop...

SDK is now up (full disclosure - I work on the team). http://azure.microsoft.com/en-us/campaigns/service-fabric/
The article itself contains no real information. Although it mentions Java specifically, it doesn't call out .Net (which I'm assuming this will support), or node.js (which MS has invested a lot of time making work in Azure).

There also doesn't seem to be a link to MS, as to how this is any different from Azure Cloud Services, other than probably only privately available to one's virtual network.

Blog post here,

http://azure.microsoft.com/blog/2015/04/20/announcing-azure-...

More information will be disclosed at BUILD next week.

It is a bit more interesting than that - the diagram in the image gives quite a bit away about what it supports; notice that there is both on prem and azure, windows and linux.

It does talk about MS's stack, BTW (which includes .NET): "this first version of Service Fabric will focus on Microsoft’s own technologies and Java applications"

But, that doesn't quite mean much to me... to me it sounds a lot like the "*.Net/.Net FOO" marketing from Microsoft around 2000-2002. The message is really muddled.
Ah, OK.

You know that they have existing PaaS (windows only "cloud apps" that run your code in a scalable way without you having to worry about managing the Windows OS it is running on) and IaaS (Windows/Linux VMs where you provision the OS, and manage it yourself).

This is kind of the next generation of that where you get your code run in a managed environment (so kind of like PaaS) but you get to choose the platform (kind of like IaaS) - and it has the same model for running your code on prem, in the cloud, or a mix of the two.

With a whole bunch of other services to help you do that more efficiently/scalably/reliably.

At least, that's what the diagrams and commentary seem to imply!

I understand that... and to be honest, it looks cool... It would be nice if they simply implemented a tiered "todomvc" style application + backend in each of the supported languages/backends ... I know this would be a bit of an effort from them, but would probably better demonstrate what they are trying to accomplish here.

I know that they have PaaS structures, which I actually did a proof of concept hello world against just yesterday, with node.js which was actually pretty cool. It's much better in terms of developing against than it was in the past when I'd looked at it. Apparently it's a similar experience for PHP (not sure on python or ruby though). The github hook integration was pretty damned cool as well... merge to `{BRANCH_NAME} == push` to production/test/qa etc. And definitely more straight forward than setting up your own dokku-alt or other cluster yourself.

Like I said, it just seems to me the terminology and marketing message is very muddled. I don't mean to say that I don't get it... I just think it's more convoluted that it needs to be. As an aside, the more I play with Azure Tables, the more awesome it becomes...

So is this like the Microsoft equivalent of Zookeeper on steroids? Seems like I can get a naming service, leader election, messaging and potentially use this to build an AP or a CP system.
There's a widely used python library called fabric. Now Apple-owned Crashlytics is calling themselves Fabric. Now we have Service Fabric from Microsoft. Strange.
To be fair, this has been around for many years within Microsoft under the name Windows Fabric. I'm not sure how old the python lib is, but the other one is much newer.
Fabric has been a term used in networking for at least 20 years.