300,000 requests / second

Re:300,000 requests / second by rbanffy on Wednesday April 07, 2010 @12:04PM (#1247629)

On what structure? A single IP? A single hostname? A hundred servers? Same datacenter? From a single SPARCstation 5 under the floor?

Without context, a number like this is meaningless.

[ Parent | on HN ]

Re:300,000 requests / second by stingraycharles on Wednesday April 07, 2010 @12:07PM (#1247633)

Since they name it "Real time DART data", I suppose it's about DART, which is the name of their adserving platform.

One can only assume it's about their entire adserving platform.

[ Parent | on HN ]

Re:300,000 requests / second by dm_mongodb on Wednesday April 07, 2010 @12:09PM (#1247643)

It's ads served per second by DART. Doesn't count adwords, adsense. It is served from several data centers and this is the total.

[ Parent | on HN ]

Re:300,000 requests / second by encoderer on Wednesday April 07, 2010 @12:26PM (#1247685)

I (think) know that mongo was created by doubleclickers.. am I right? Is doubleclick actually using it?

What I'd love to know is.. what kind of sharding is in place accomplish that?

[ Parent | on HN ]

Re:300,000 requests / second by dm_mongodb on Wednesday April 07, 2010 @12:39PM (#1247718)

Correct - MongoDB project was started by ex-DoubleClickers.

DoubleClick doesn't use MongoDB (afaik). My understanding is that it is switching over gradually to internal google infrastructure services - so presumably bigtable like things, and its internal successors, whatever those are. That said, MongoDB is basically what I wish I had had when we began building DART long ago.

In the early days of DART we build a proprietary distributed key/value store called UIS. It was super simple and super fast and worked well. The request volume, both read and write to it, was huge: it was hit in realtime by the ad servers.

[ Parent | on HN ]