Scalable memory allocation is critical to extracting maximum value from each machine in Facebook's server farms. A typical Facebook server application concurrently utilizes 8 or more CPU cores, and 5 to 70 gigabytes of RAM. At this scale, few memory allocators are able to provide satisfactory throughput, fragmentation avoidance, and introspection capabilities. We have successfully enhanced... [read more]

Scalable memory allocation is critical to extracting maximum value from each machine in Facebook's server farms. A typical Facebook server application concurrently utilizes 8 or more CPU cores, and 5 to 70 gigabytes of RAM. At this scale, few memory allocators are able to provide satisfactory throughput, fragmentation avoidance, and introspection capabilities. We have successfully enhanced the open source jemalloc memory allocator (http://www.canonware.com/jemalloc/) to meet all of these needs for a broad range of applications.

Join us on Tuesday, January 11 at 7pm PST to hear Jason Evans talk about jemalloc's core algorithms, innovations motivated by Facebook's computing environment, real-world benchmark results, and directions for future development.

Can't make this one?

We’ll be streaming the Tech Talk live from the ‘Facebook Live’ tab at 7pm PST! (http://www.facebook.com/FacebookLive) We invite you to ask questions via the ‘Ask A Question’ button (you can do so both in advance and during the Tech Talk).

NOTE: Due to limited space all +1s must have their name on the list. Please either comment on the wall with their name or have them request an invite on the event.