Mary-Jo Foley: MS to drop Dryad, focus on Hadoop

http://www.zdnet.com/blog/microsoft/microsoft-drops-dryad-puts-its-big-data-bets-on-hadoop/11226

As far as I understand, both are Map-Reduce implementations, one written in Java, and is open source, and the other is Microsoft proprietary.

Not sure if this applies, but to me it seems that community (not to mention openness) always wins over great tech. ie Java might be an old version of dotnet, but has much more community support behind it. I'm not sure Dryad is a dotnet product.

Not sure if this applies, but to me it seems that community (not to mention openness) always wins over great tech. ie Java might be an old version of dotnet, but has much more community support behind it. I'm not sure Dryad is a dotnet product.

So basically you wrote that whole paragraph only to make it irrelevant by that last sentence?

http://www.zdnet.com/blog/microsoft/microsoft-drops-dryad-puts-its-big-data-bets-on-hadoop/11226

As far as I understand, both are Map-Reduce implementations, one written in Java, and is open source, and the other is Microsoft proprietary.

No, Dryad is not a MR implementation at all. It is a far more general graph-based distributed computation platform which happens to also be able to do MR, but it's much more flexible than that.

Dryad is written in C++, but one of the (several) ways to write programs for it was DryadLINQ, a LINQ provider that converts LINQ expressions into distributed programs and executes them on Dryad. It was an incredibly powerful approach that puts similar initiatives like Pig and Hive for Hadoop to shame.

Dryad itself used Microsoft Cosmos as the underlying execution engine, which I believe is used extensively inside MS. I severely doubt they are abandoning that in favour of Hadoop (which is often very inefficient and currently doesn't run very well on Windows). They may be abandoning Dryad as an offering for third parties to use, I could definitely believe that.

Thanks Sven! So would Microsoft be working on a Linq provider for Hadoop?

Funny comment in that article: 

"The fact that it took almost a week before anyone commented on the drop has to be considered a pretty good indicator that there wasn't much interest in the product.

curph"

Dryad never managed to build up any momentum outside of MS. This is because there was a very long delay between the publication of the paper and the first public release, which was not only basically a crippled version but also had an "Academic use only" license. They just kept it out of the public eye for too long, in which time Hadoop built up a very strong following.

@Sven Groot:Plus, even though they obviously tried, they just didn't come up with a name as utterly rediculous as 'hadoop'.

@Sven Groot:

You (and fanbaby) nailed it. Dryad was dead before it even started because of its incredibly stupid license. It doesn't matter if Dryad is better or not, it doesn't mean crap unless people can use it.

@Sven Groot, et all:  Yes, this was the problem in trying to consider it's use for us as well.  Now I'm glad I didn't spend time learning much about it.

1 day ago, Sven Groot wrote

Dryad never managed to build up any momentum outside of MS. This is because there was a very long delay between the publication of the paper and the first public release, which was not only basically a crippled version but also had an "Academic use only" license. They just kept it out of the public eye for too long, in which time Hadoop built up a very strong following.

Yes.  When they finally started to talk about it (and I have only ever seen it mentioned on C9), it wasn't immediately available, and only then with a license that makes it irrelevant for most of it's potential customers.  It was as if they didn't really want to release it in the first place.  Or maybe they were trying to gauge reaction and make improvements before some eventual commercial license release -- which never came.  I wouldn't think many people would be willing to invest the time and effort to set up something like that if they didn't have the license (and pricing) info up front.  Nobody ever knew about it.  Another good idea that died on the table.

36 minutes ago, ryanb wrote

*snip*

 Another good idea that died on the table vine.