(Sorry if this is a dup. Last attempt to send this got hung up in the mail relay and it's fate is unknown. Sending from my personal gmail now.) It's great to see all the work being done to improve swap performance. I'd like to contribute to this discussion by focusing a topic specifically at the value of having a compressed swap cache in the kernel. I am currently working on the zswap code [1] to do just this. Compressed swap cache can mitigate the latency and nondeterminism associated with swapping by drastically reducing I/O to the swap device. The I/O reduction with zswap is very impressive [2]. I think a compressed swap cache would be a very attractive mm feature for the kernel. Zswap contains solutions for avoiding inverse LRU and writeback for pages contained in the compressed pool. I would like to talk about the approach zswap uses and also discuss the policies involved like the sizing of the compressed pool and writeback policy. [1] https://lkml.org/lkml/2013/1/29/543 [2] http://ibm.co/VCgHvN -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>