Google is definitely an advertising company, but they do it well.
186 karma · joined April 7, 2009
Google is definitely an advertising company, but they do it well.
We're willing to relocate, and the internship length is flexible. The internships are paid.
We wrote an ad server and have our own ad exchange. We also wrote our own time-series datastore (Erlang and C). Most of our code is Erlang, Java, and C. We're using Cassandra, Riak, CouchDB, Hadoop, and others.
We have datacenters in the US (east/west), EU, and Japan. We handle billions of requests a day, and get a couple hundred milliseconds on each one to determine which ad is the best one to serve. A typical request will hit eight backends in that time period. If that sounds fun, you can reach me at joel at openx . com.
In my limited experience, it seems like finding a good startup job in LA depends a lot on your network - e.g. you worked with so-and-so here and they're doing something new and think of you, where my perception of SF/SV is that awesome startup jobs are so plentiful you can't help but have one.
=)
Surprisingly effective. Also encourages development of other positive character traits.
riak_kv uses gen_fsm, so people looking for a real world use case could check out their usage: https://github.com/basho/riak_kv/blob/master/src/riak_kv_get... https://github.com/basho/riak_kv/blob/master/src/riak_kv_put...
I'm a relatively new to Erlang, but being familiar with gen_server made it pretty easy to understand the gen_fsm docs and figure out what was going on. I don't see the exclusion as a big deal.
I was pissed until I read the FAQ and discovered when they're using data and what they're using it for. Basically: If you've registered with Crystal Something-or-Other they send your data to them. The WSJ article, at least in the Angry Birds case, seems to have sensationalized things.
A colleague just recommended this book last week.
Personally, I'm moving away from Java and languages that target the JVM. I find that Erlang lets me express solutions more concisely and the Erlang emulator/vm has many of the tools needed for massively scalable, distributed, fault-tolerant systems built-in.
The hosted version (called OnRamp) would be the place to start - it's free, performs better (it's not all PHP), and gives you easy market integration (which the original comment was referring to). OpenX may or may not be what you're looking for, but at least you can test the waters without any initial investment other than creating an account and setting up a few tags.
(Disclaimer: I work for OpenX.)
I was defining worth as "the value that something confers" which may (or may not) be tied to the difficulty with which it was obtained, but requires you to define value, which I have a hard time doing for Karma.
I just checked the profiles of the top ten people on the leader-board and while I'm sure they are great people who have really contributed a lot to HN, I would give more weight to a comment/posting from the afore-mentioned HN user with karma 31 than any of them. And, to be honest, I don't really look at people's karma so I'm not sure how you would define the worth of a karma point, let alone determine change in worth over time.
(which I'd certainly agree with)
Edit: [1]What languages/frameworks you know.
But like I said, it's unlikely.
http://www.amazon.com/ERLANG-Programming-Francesco-Cesarini/...
http://www.amazon.com/Programming-Erlang-Software-Concurrent...
Okay, just kidding. =)
On a serious note, Effective Java is one of the best out there.
Anyway, nice job. When I implemented ROAM and then Lindstrom/Pascucci's algo 8 years ago it ran like a dog on an antiquated SGI - I never would have dreamed something like this would run on the browser some day :)
Edit: Ah - never mind, I see you reference ROAM another comment. If you haven't read the paper referenced above you may enjoy it: http://www.pascucci.org/pdf-papers/vis2001.pdf