2 ms·
Complex in memory data structures are the main use case that is not well supported by multi process architectures. There's a trend towards in memory databases a
by fauigerzigerk 17y ago
Complex in memory data structures are the main use case that is not well supported by multi process architectures. There's a trend towards in memory databases and using disk for archival only, so this issue is becoming critical.
Processes don't share pointers, they only share BLOBs. Threads do share pointers and that's why they are more suitable to in memory data analysis/manipulation.
- krakensden 17y agoDon't the in-memory databases generally run in another process (memcached) anyway? (Not that your general point is wrong)
- old-gregg 17y agoBut don't you want your in-memory database to be non-locking, i.e. serving many clients simultaneously and fast?
- krakensden 17y agoThe point was your in-memory database is an external, non-python program, and you run lots of python clients. The way most web applications are structured, for instance.
- fauigerzigerk 17y agoYes but if I'm the one writing in memory database/analytics stuff I might want to use a Python/C combination to do that. Right now that's not an option because of the GIL issue.