[j-nsp] RE-S-X6-64G-BB

Saku Ytti saku at ytti.fi
Mon May 23 12:57:15 EDT 2016


On 23 May 2016 at 19:39, raf <raph at futomaki.net> wrote:

> The first good effect is all platform will converge and use x86 processors.

I think this is driven by not having options mostly, freescale isn't
there for today's control-plane scale.

> So we will converge on one processor target which it a good news for the
> stability of the software, test and so on. We can hope for a real one OS for
> all platform. :)
> Other thing I saw; is that we it will allow separating components of the NOS
> in multiple VMs with real QOS between VMs.
> We could imagine running a vm dedicated to RR for example.

Sure, it's better than logical-systems but still, it's added
complexity, which reduces availability. BOM of HW is small. I'd rather
pay premium to have more robust network.

> I think the contrary. As I say it should allow running much better isolated
> process. And perhaps snmpd will not kill the enitre control plane...
> OK this also possible inside one OS with a good software architecture but it
> is more complicated with the actual code.

I don't understand your arguments. You're saying it's more complicated
to do IPC inside single machine than between two different machines?
The problem today for JunOS is, that largely everything is inside
single OS process. So it does not matter how good FreeBSD or Linux
kernel is doing this stuff, it matters how good Juniper is doing this
stuff. And Juniper isn't even planning on moving stuff away from RPD,
but rather introduce OS threads.

If we could move significant portions of the control-plane to another
VM, then we need to do IPC over network. If you can produce
control-plane which allows for this, then why distribute the work to
another VM, why not distribute it dedicated compute over network?
Having said that, my concern on reliability wasn't on vendor itself
distributing work on VMs, but we, the customers, running our own VMs
in the router hardware, and I don't see any wins there, just reduced
availability of whole system.


-- 
  ++ytti


More information about the juniper-nsp mailing list