KernelDedup

Feed Rss

Q&A

Q: What platform can benefit from the technologies provided by this website?

A: Currently we focus mainly on Linux & Xen x86 platforms. Some people have ported UKSM to android. Those ports are not officially supported by us, but we welcome any kind of feedbacks.

 

Q: My machine has abundant memory, do I still need memory deduplication?

A: That depends on your workloads。If your workloads are deduplication friendly(i.e. have many duplicated memory pages), then deduplication will be helpful no matter how much memory you have. Modern OS will always use your free memory as file system cache. The more free memory you have the less I/O need to do. That will improve your computer’s overall performance.

 

Q: What kind of workloads are deduplication friendly?

A: The typical cases are virtualized environments like KVM, Qemu, LXC, OpenVZ, Xen where many identical/similar workloads are running in parallel. For daily desktop users, there are applications who fill their memory with identical contents(e.g. same color, empty data, etc). We have noticed that popular pdf readers like Okular is a very good case. You can have a try of UKSM and see how much memory is saved in you daily work.

 

Q: What if my workload is not deduplication friendly?Will UKSM/Xen-dedup eat up much CPU?

A: One of the key features of our technology is the very low CPU consumption when there is not so much identical pages in system. The performance overhead of UKSM/Xen-dedup are negligible. Please go the benchmark pages of your projects for details.

 

Q: Is your code stable enough to be used in production systems?

A: For every release, we do internal testing which we consider “enough”. And to our knowledge, our patches have been already used world wide by many people including in production systems. However, we are not responsible for any consequences caused by using our patches.