||Ravikiran G Thirumalai
||Design of Dynamic per-cpu kernel memory allocator
||Frequent updates to global shared data in linux kernel results in expensive cache activities on multi-processor systems. Per-cpu data provides a way for avoiding costly cache activities by reducing the amount of sharing resulting in performance improvements.
In 2.5 kernel a new set of interfaces were introduced to support static per-cpu data area, but, it does not work with dynamically allocated data and kernel modules. This talk presents a dynamic per-cpu data allocator that we have developed. This work has been published in the linux kernel mailing list and is a likely candidate for 2.5. Speaker is the co-author of this feature.
||Linux kernel developer working on scalability aspects of linux kernel. Author of scalable statistics counters feature and some scalability related contributions to the kernel for 2.5, co-author to the kernel dynamic per-cpu mem allocator. Working with the Linux Technology Center, IBM for over 2 1/2 years now.
See Talk Schedule