News for hackers. Stuff that matters.
Re:Parallel Universe (YC S12) Developing Spatial Databases F... by kelleyk on Monday July 09, 2012 @11:13PM (#4221892)
I'd be curious to hear more about what they're actually doing (technically speaking). Is this a wrapper around kd-trees, R-trees, and their friends and relatives? Is it something fancier?
Re:Parallel Universe (YC S12) Developing Spatial Databases F... by pron on Monday July 09, 2012 @11:25PM (#4221934)
The spatial index is an R-tree variant that doesn't degrade and allows concurrent writes (multiples writers at a time, let alone readers). Readers don't block writers and don't use locks, while writers lock a small subset of the database for atomic transactions (one of the most common operations we need to support is a "move" of an object from one location to another - that has to be atomic).

For parallelization of queries and transactions we use fork-join.

Re:Parallel Universe (YC S12) Developing Spatial Databases F... by tintor on Tuesday July 10, 2012 @01:23AM (#4222248)
How do you do lock-free updates to the R-tree structure?
Re:Parallel Universe (YC S12) Developing Spatial Databases F... by pron on Tuesday July 10, 2012 @01:35AM (#4222282)
Most updates aren't lock-free.
Re:Parallel Universe (YC S12) Developing Spatial Databases F... by rprasad on Tuesday July 10, 2012 @01:38PM (#4225046)
I think what pron is saying is that initial writes are lock-free but updates to existing data will require locks in most cases?