To Virtualize Or Not To Virtualize. Part 4 - Conclusions.

While the state of using legacy virtualization and its bundled IO for running a network element in software at present is somewhat dire it does not really need to be that way. There is a LOTS of room for improvement of the actual virtualization IO. Let's be clear, the results of doing all the work needed for this will never ever beat something written with let's say DPDK in terms of raw performance. They, however, will be cloud ready and will be able to coexist with regular cloud loads - something DPDK and similar frameworks will never be.

The work required is not that much, the pity is that there is nobody doing it. I, nowdays work on something else. I have managed to push some fixes here and there continuing my old work. However, all in all, I do not have the cycles to work on this - unless someone sponsors putting proper effort into it.

If you are interested at what is required for the particular hypervisor - have a look at the actual details and the backlog for QEMU and UML in their respective articles.

-- AntonIvanov - 28 Jan 2017
Topic revision: r1 - 31 Jan 2017, UnknownUser

This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Foswiki? Send feedback