Unikernels have been shown to provide incredible performance benefits, including boot times of a few milliseconds, low CPU utilization, throughput in the range of 10-40 Gb/s and memory footprints of a few MBs or even KBs, to name but a few metrics. The clear downside is development time: it can take several months or longer to produce a working, optimized unikernel, and up until now, the process needed to be repeated for each target application. There was no shared research and development on the building and creation of unikernels.
In this presentation we will introduce Unikraft, a new, open source Xen Project sub-project under the auspices of the Linux Foundation that seeks to automate the development process for unikernels. Unikraft decomposes operating systems and libraries into elementary modules called libraries (e.g., schedulers, memory allocators, drivers, filesystems, network stacks, etc) and then allows users to quickly pick and choose, through a menu, which components to use when building a unikernel. Once done, Unikraft builds one image per target “platform” (e.g., Xen, KVM or Linux user-space), severely reducing unikernel development time. We will show a brief demo on how to use Unikraft, and cover the basics of the system including how to contribute to it.
We have spent quite a bit of our time over the last few years developing unikernels – specialized virtual machine images targeting specific applications. We have been particularly interested in them since because of their fantastic performance benefits: tiny memory footprints (hundreds of KBs or a few MBs), boot times comparable to those of processes, and small migration times, to name a few metrics.
Despite the fact that this work and work from several others is proof of their potential, unikernels have yet to see massive adoption. One of the main showstoppers is development time: for instance, developing Minipython [4], a MicroPython unikernel, took the better part of 3 months to put together and test. ClickOS [5], a unikernel for NFV, was the result of a couple of years of work. What’s particularly bad about this development model besides the considerable time spent is that each unikernel was basically a “throwaway”: every time we wanted to create a new unikernel targeting a different application, we would start more or less from scratch.
One thing we did do to try to re-use the work a bit more was to create a separate repo consisting of a “toolstack” that would contain functionality useful to multiple unikernels, mostly platform-independent versions of newlib and lwip (a C library and network stack intended for embedded systems).
That got us thinking that we should take that a (much bigger) step further: wouldn’t it be great to be able to very quickly choose, perhaps from a menu, the bits of functionality that we want for a unikernel, and to have a system automatically build all of these pieces together into a working image? It would also be great if we could choose multiple platforms (e.g., Xen, KVM, bare metal) without having to do additional work for each of them.
The result of that thought process is Unikraft. Unikraft decomposes operating systems into elementary pieces called libraries (e.g., schedulers, memory allocators, drivers, filesystems, network stacks, etc.) that users can then pick and choose from, using a menu, to quickly build images tailored to the needs of specific applications.
Speakers: Simon Kuenzer