"Here is something you've may have heard but never quite believed before: Google's vaunted scalable software infrastructure is obsolete. Don't get me wrong, their hardware and datacenters are the best in the world, and as far as I know, nobody is close to matching it. But the software stack on top of it is 10 years old, aging and designed for building search engines and crawlers. And it is well and truly obsolete.
Protocol Buffers, BigTable and MapReduce are ancient, creaking dinosaurs compared to MessagePack, JSON, and Hadoop. And new projects like GWT, Closure and MegaStore are sluggish, overengineered Leviathans compared to fast, elegant tools like jQuery and mongoDB. Designed by engineers in a vacuum, rather than by developers who have need of tools."
if true, this is a strong indicator that Google is well and truly a BigCo now. Everyone expects a company of Google's size to have its share of politics and crappy middle managers and so on, but this is the first time that I've heard a Google engineer (ok an ex Google engineer) say that its software is bloated and ugly (and more importantly, not getting fixed - see the bits on the rewards structure encouraging territoriality leading to rejection of patches.)
Yea so the infrastructure comments really shocked me but it makes complete sense. They had to do massive scaling early on and a lot of what they use now is from that era. I can imagine even at google they have some sort of "if its not broken dont fix it" mentality. There is a reason why megastore is sluggish right? Its tries to uphold CAP theorem across datacenters and because of that write performance suffers.
Not only that, but the Second System Effect is alive and well. Many times, it's happening because nobody has ever heard of The Mythical Man Month.
If you ever wanted to go out and do something but were worried about Google entering that space, go do it anyway. Odds are good they aren't going to catch you due to the 500 pound weight any project will have to lug around.
Is it really true that Hadoop > MapReduce? I was under the impression (as a limited outsider) that Hadoop is still trying to catch up to MapReduce.
Also, I've heard from a Google friend that MapReduce is no longer the production-level thing they use for large-scale calculations, but rather that Pregel has replaced it. (He said MapReduce is more analogous to what you use when you want a quick and dirty one-off command-line job.) Again, secondhand knowledge, though.
4 comments
[ 2.7 ms ] story [ 18.1 ms ] threade.g:
"Here is something you've may have heard but never quite believed before: Google's vaunted scalable software infrastructure is obsolete. Don't get me wrong, their hardware and datacenters are the best in the world, and as far as I know, nobody is close to matching it. But the software stack on top of it is 10 years old, aging and designed for building search engines and crawlers. And it is well and truly obsolete.
Protocol Buffers, BigTable and MapReduce are ancient, creaking dinosaurs compared to MessagePack, JSON, and Hadoop. And new projects like GWT, Closure and MegaStore are sluggish, overengineered Leviathans compared to fast, elegant tools like jQuery and mongoDB. Designed by engineers in a vacuum, rather than by developers who have need of tools."
if true, this is a strong indicator that Google is well and truly a BigCo now. Everyone expects a company of Google's size to have its share of politics and crappy middle managers and so on, but this is the first time that I've heard a Google engineer (ok an ex Google engineer) say that its software is bloated and ugly (and more importantly, not getting fixed - see the bits on the rewards structure encouraging territoriality leading to rejection of patches.)
Mind blowing.
If you ever wanted to go out and do something but were worried about Google entering that space, go do it anyway. Odds are good they aren't going to catch you due to the 500 pound weight any project will have to lug around.
Also, I've heard from a Google friend that MapReduce is no longer the production-level thing they use for large-scale calculations, but rather that Pregel has replaced it. (He said MapReduce is more analogous to what you use when you want a quick and dirty one-off command-line job.) Again, secondhand knowledge, though.