A New DB for 80% of Facebook, YouTube-scale Sites

roadtofailure.com

16 points·gsteph22··13 comments

Open articleView on HN

This is just blog spam!

"I update this blog a few times a month. Don't worry -- I'm worth waiting for."

User gsteph22 has only submitted two articles, both from this crappy blog.

Oh, how horrible -- self promotion :)

I think it's generally considered okay to submit your own blog here.

Anyone tried Cassandra? I thought that was what Facebook was using for this type of thing?

Yes. Actually we [cassandra] are seeing a fair number of people switching to Cassandra from hbase because of speed and reliability (no single points of failure).

That's interesting -- we've seen a lot of the opposite, especially from the .20 release. I guess it all depends on what you need to do with it :)

MongoDB seems to be heading that way, slowly but surely.

I am using couchdb (with couchdb-lounge) to store distribute a large dataset for a computationally intensive scientific application. So far so good.

MongoDB is heading that way, but its support for partitioning, which couchdb-lounge is performing adequately already, is in alpha.

I just looked at couchdb-lounge because I'm very interested in partitioning and hadn't heard about it.

The wiki unfortunately left the important questions unanswered. Does lounge handle the rebalancing when new shards are added/removed? Do I have to take the cluster offline to add new nodes? What are the failure modes (e.g. a node failing and coming back)? How is a replacement node bootstrapped?

Since none of that is mentioned I have my doubts that lounge is handling it adequately. But I'd be glad to hear otherwise!

Berkeley DB?

Drizzle?

Does anyone know where the "Live Nude Girls" are on his site? The link seems to be broken.

It's a Secret!

Reply on news.ycombinator.com