Userspace RCU Will Be Much Faster For Its Next Release Paired With Linux 4.14+

[phoronix]

Phoronix: Userspace RCU Will Be Much Faster For Its Next Release Paired With Linux 4.14+

The liburcu Userspace RCU data synchronization library should be significantly faster when built with a modern Linux kernel release...

http://www.phoronix.com/scan.php?page=news_item&px=URCU-Membarrier-Performance

[FireBurn]

What uses this library? It doesn't not even appear in Gentoo's portage

Edit

OK so it's called userspace-rcu in Gentoo and it's used by:

• dev-util/lttng-tools - LTTng, the Linux Trace Toolkit Next Generation, is a project that aims at producing a highly efficient full system tracing solution
• dev-util/lttng-ust - The userspace tracer is designed to provide detailed information about userspace activity. UST is a port of the LTTng kernel tracer to userspace. Like the LTTng kernel tracer, performance is the main goal. Tracing does not require system calls or traps. UST instrumentation points may be added in any userspace code including signal handlers and libraries.
• net-analyzer/netsniff-ng - is a free Linux networking toolkit, a Swiss army knife for your daily Linux network plumbing if you will. Its gain of performance is reached by zero-copy mechanisms, so that on packet reception and transmission the kernel does not need to copy packets from kernel space to user space and vice versa.
• net-dns/knot
• sys-cluster/glusterfs- GlusterFS is a cluster file-system capable of scaling to several peta-bytes. It aggregates various storage bricks over Infiniband RDMA or TCP/IP interconnect into one large parallel network file system. GlusterFS is based on a stackable user space design without compromising performance.
• sys-fs/multipath-tools

[Chewi]

Faster GlusterFS would be nice but I suspect this won't make that much difference on the lowly ARM system I tried it on.

[rene]

well, given that the Intel security vulnerability workarounds will make everything dramatically slower, this few percent user-space speedup will certainly not be notable at all :-/

[dwagner]

As much as RCU can help to make multi-threaded code more scalable, even this library cannot make it "easy to use" for programmers. Much unlike conventional container classes, significant efforts are required from the application's side to go through all the RCU phase transition hoops without producing horrible, hard to find bugs.

Also, many real world applications also need great scalability on the write side, and that is especially hard to accomplish using RCU.