johngibbon

3 ms·

1. 1.▲
   by johngibbon 14y ago

   Don't give us data-parallelism. Give us task-parallelism! We can build data-parallel libraries on top, but still have the power of task-parallelism for those tricker parallel problems.
2. 2.▲
   by johngibbon 14y ago

   What kind of database are you going to insert into if you have 1000 TB of data? Let alone an open source one. What kind of database is going to allow you to set everything up and strip it all down just for one of a million passes of your da
3. 3.▲
   by johngibbon 14y ago

   Only if your data is stored in such a way that you can monitor adds and modifications. MapReduce is often used on huge bags of data pulled from random places or as one pass in a massive sequence of passes (think pagerank), where any index w
4. 4.▲
   by johngibbon 14y ago

   Where does the index come from?