Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
FathomDB (YC W08) introduces PlatformLayer: XaaS (justinsb.com)
70 points by tjpd on April 6, 2012 | hide | past | favorite | 12 comments


That hit HN fast! I'm the founder of FathomDB (YC W08) & the primary author of PlatformLayer.

PlatformLayer lets you run anything as a service: MySQLaaS, RedisAAS, NginxAAS, DnsAAS etc. It's based on the code & know-how from running MySQL-as-a-service for FathomDB.

I'm planning a series of explanatory blog posts & better documentation. In the meantime, I'll hang out here to answer any questions.


How does this compare to dotcloud?


Dotcloud is closed source, but much more mature - they have customers using it in production. PlatformLayer is open source, but I wouldn't run a bank on it - yet!

But the big difference is one of scope: PlatformLayer makes it easy/easier to build a service like DotCloud. You could choose to run it internally (e.g. if you want to run on a private cloud). You could choose to make it publicly available as a DotCloud competitor (maybe not today, but soon!)

The end-goal is that it will be much easier to build these services, so there will be more of them, covering any software you want to run. So if you just want to consume services your options should be better, cheaper and more numerous; even if you never run PlatformLayer directly.

But this is the beginning of a project, where DotCloud is a useful hosting platform today.


Ironically dotCloud started out in 2008 as a very similar project, with the same lofty goals. We had VM templating and versioning (https://github.com/dotcloud/cloudlets), cross-infrastructure deployment (https://bitbucket.org/dotcloud/vm2vm), various goodies based on container virtualization, copy-on-write filesystems (https://github.com/dotcloud/aufs.py), and an orchestration layer to tie it all together (I found an old fork at https://bitbucket.org/Foi3GraS/dotcloud). It powered various 3d-party "XaaS" throughout 2009 - here's a French provider that still seems to be alive: http://www.diva-cloud.com/softwares

Then in 2010 we joined YC, and actually tested our claim that "you can make it publicly available as a service!". Within 6 months we had rewritten 100% of the codebase around the real constraints of deploying and being responsible for tens of thousands of applications. We are only now getting around to open-sourcing some of these new, battle-tested components: a great example is ZeroRPC (http://github.com/dotcloud/zerorpc-python). More to come.

My hard-earned advice to you Justin is: you have to operate PlatformLayer as a service first and foremost, and the code you publish must flow from the real-world experience of operating that service, charging money for it, and being accountable for its reliability - not the other way around. There's a reason VMWare is a flop in the cloud world: they don't know how to operate their stuff.

In any case, I'm super excited to see so much open-source activity around solving these problems. Back in 2008 it was quite lonely :)

We hackers all benefit from this friendly competition in the end - happy hacking!


Very interesting - thanks! I look forward to seeing more open-source contributions from DotCloud.

This does come out of the experience of running FathomDB, so it isn't an ivory-tower project. FathomDB grappled with the fact that traditional relational databases are a terrible match for the cloud, so we've built a new database. When it came to how to run that new database as-a-service, there was so much "undifferentiated heavy lifting" involved that we decided that open source was the way to go here.

So this _is_ the rewrite that comes from the real-world experience of running a production service. And there will be at least one real service, charging real money, and supporting real customers, built on it. But the secret sauce of FathomDB isn't the "aaS" bit of "DBaaS" - it's the "DB". By open sourcing PlatformLayer, we're hoping the community will share some of the work of "the aaS end", but even if not, PlatformLayer will be solving the problem for real.

And if that encourages you to open-source DotCloud, and if that replaces PlatformLayer, we all win. The strange economics of open source!


Interesting -- what were the problems that came up from running real services? Was there a fundamental architectural issue, or just polishing off all the corner cases? Fault tolerance issues? Too complex? Idempotency, state management, etc.?

I tried VMWare's cloud foundry. It's hidden behind a signup form that has to be approved. Other than that it seemed decent. I managed to get through the hello world and have a working rails app. Not as good as Heroku. I'm not really sure if VMWare really wants people to use their service, or they just want people to run their own Cloud Foundry so they can sell VM licenses.


> I'm not really sure if VMWare really wants people to use their service, or they just want people to run their own Cloud Foundry so they can sell VM licenses.

I'm not sure they've figured it out themselves, either :)


(I forgot the traditional disclaimer: I'm the co-founder and CEO of dotCloud :)


Looks very interesting, would love to see a guide or video walkthrough from the start (installation/configuration OpenStack) to building a simple service.


Thanks. That's definitely the next step - just a matter of finding the time.

NodeSocket looks cool - would you mind sharing a bit about how you built it? Would PlatformLayer have helped?


Sure our hypervisor management is OnApp. We have our own hardware in a datacenter here in San Francisco. Are you here in the valley?


I hadn't seen OnApp before - I'll check it out. Glad to see you went with your own hardware rather than locking your architecture in to AWS - I hope that'll be a big advantage for you long term.

DM me on twitter if you're interested in meeting up sometime to talk platforms! @justinsantab




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: