Hi Anil & Dan,I did some experimentation with the Packet.net over the weekend to try and get myself a repeatable/shareable development environment for solo5. This environment is on the smallest instance type they have (Intel - baremetal_0), but the same structure should be re-usable for the 96 core ARMv8.
I can compile and run kvm+solo5 unikernels using this setup and also make small changes to target the 96 core ARM monster. This avoids the pain of cross-compilation toolchains. Of course the ARM build fails early in the compilation, but that in itself was instructive.
Let me know if this is a mis-directed Yak-shave or could possibly be useful. I’ve only seen Dan’s development, would be interesting to see what Mato & WeiChen use.
I’m assuming that Wei Chen was busy over Chinese New Year - any recent news?
Thanks, Chris
Dear Chris, In addition to Nik's very useful reply, you may also be interested in some UKVM work that will finally connect the dots and provide an end-to-end ARM64 unikernel based on Solo5: https://github.com/Solo5/solo5/issues/151Any contributions there would be most welcome, as once that exists, several of the older projects such as Jitsu, and newer projects such as FLICK, could be brought up to speed with the Mirage 3 module types and APIs. I recommend the ODroid C2 as a pretty portable ARM64 box to take to the hackathon :) regards Anil On 3 Feb 2017, at 03:02, Chris Fegan <cfegan@xxxxxxxxxxxx> wrote:
Hi Folks,
I've been reading a lot of this group's research papers in preparation for Marrakech. I think
I'm getting a sense for how the various pieces fit together.
The reading has lead me to wonder if there is a "master plan" for linking MirageOS, FLICK, many-core ARMv8 &
SR-IOV. It seems as though a stack something like the below would be a great substrate for building high performance and flexible network applications (that look at the packet payloads, not just headers).
Jitsu
FLICK
Mirage
Solo5
Microkernel | Hypervisor [seL4]|[KVM]
Cavium many-core ARMv8 SoC with Ethernet fabric
Besides the question about the existence of a master plan.
* Are the Flick tools available in source code to study?
* Do you think that Flick will need to remain outside of the Mirage & OCaml ecosystem to make use
of multicore & DPDK or can you see a path bring it "into the fold"?
[Flick] https://www.usenix.org/system/files/conference/atc16/atc16_paper_alim.pdf
[seL4] https://ts.data61.csiro.au/projects/TS/realtime.pml
[KVM] http://www.linux-kvm.org/page/Main_Page
[Cavium] http://www.cavium.com/ThunderX_ARM_Processors.html
Thanks,
Chris
_______________________________________________
MirageOS-devel mailing list
MirageOS-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/cgi-bin/mailman/listinfo/mirageos-devel
|
