| Yeah, if the workloads are fairly self-contained, unikernels are great. It gets more interesting when they form a distributed system, since a group of unikernels may need to be migrated as a group.
I'd suggest starting by outlining the scenario you want to build, and then we can help you work through the unikernel components to build. Are you looking at ARM or x86 for your POC?
For instance, beginning with just a static website workload, and building the HTTP proxies to do workload balancing would be a small but significant start (since it would require building out the proxy and migration infrastructure).
-anil
Hi,
Which kind of workload are you trying to run? If it's an OCaml program, yes sure, MirageOS might be a good option.
If not, there are other unikernels: HalVM in Haskell, Erlang-on-Xen, OSv for the JVM, etc... They are basically all tied to a specific language runtime environment and they can interoperate between each other and with legacy OS at the API level.
Thomas
I have been trying to come up with an easy way to move workloads around. This is for a home lab POC, not for a commercial application.
I think an OpenMirage unikernel might work well as a workload container (container as in 'a package of stuff', not as in 'Docker'). However, before I burn evenings and weekends going off in the wrong direction (done that plenty), I'd like to get some expert criticism first.
How would I present the high level idea, and how would I get advice on how to use work done so far on OpenMirage?
MirageOS-devel mailing list
MirageOS-devel@xxxxxxxxxxxxxxxxxxxx
http://lists.xenproject.org/cgi-bin/mailman/listinfo/mirageos-devel
MirageOS-devel mailing list
MirageOS-devel@xxxxxxxxxxxxxxxxxxxx
http://lists.xenproject.org/cgi-bin/mailman/listinfo/mirageos-devel
|
