Peter Memishian wrote:
> This means that instead of some 250k possible open network streams,
> there will only be some 128k possible open network streams (that is
> streams that go all the way to a GLDv3 device) in a single system.
> Frankly, I doubt this is much of a limitation given current systems.
> (How many plumbed interfaces, and simultaneous copies of snoop/tcpdump,
> do we really think we need to support?)
If you consider that we can scale to thousands of zones and consider that
there may be many non-physical GLDv3 datalinks (tunnels, VNICs, ...), 128K
possible open streams doesn't seem doesn't seem that far out of reach.
I'm uncomfortable with that sort of a limitation.
What about the idea of dealing with this at a later date, when it
actually shows up as a problem? I can certainly address it now, but it
adds a lot of complexity, and it doesn't really *solve* the problem.
(Because you still wind up with a 250K limit.)
Even with a fairly complex configuration, I think it would be hard to
hit this 128K limitation. Remember, to hit this limit, you have to have
all of these links open at the same time. So if you had 8192 zones
(which would be a large humongous configuration, I think -- in fact I
think 8k zones is an actual implementation limit), each of them could
have 16 active mac clients.
While having 16 active mac clients is not hard to imagine a single zone
that is somewhat complex (a couple of vnics, or alternate stack
instances will do it for you), having 16 active mac clients on *all*
8192 zones is -- I think -- rather unlikely. (The system would at that
point probably be so complex that nobody could understand it.) If you
roughtly assume that 80% of the zones are going to have on average 8 or
fewer MAC clients -- a pretty fair assumption I think -- most zones will
only need maybe 4 mac clients during normal operation -- then you wind
up with over 78,000 available minor numbers for the remaining 1638
zones. (Or an average of 48 mac clients per zone.)
All this means I think it is extraordinarily unlikely that we'll see
128K in-use minor numbers against GLDv3 devices without at least first
making other implementation changes.
- Garrett
_______________________________________________
networking-discuss mailing list
[email protected]
