[j-nsp] juniper router reccomendations

Fri Jul 29 08:43:59 EDT 2016

"Turbo FIB" is an internal Juniper name for some speed optimizations in a
version of the 15.1F6 code base, that only work on 64-bit REs.    The test
results I saw (a powerpoint presentation) showed a greater than 10% speed
increase from the previous release for installing prefixes into the FIB
from RIB.  I don't know if this is a public release yet, I will ask our SE
team.  I didn't pay close attention because we don't have any 64-bit REs
yet.

On Fri, Jul 29, 2016 at 12:36 AM, Giuliano Medalha <giuliano at wztech.com.br>
wrote:

> Hello James
>
> What do you mean with turbo fib ?
>
> Is a new config from 15.1 ?
>
> Thanks a lot
>
> Giuliano
>
> >
> >
> > I saw test results from the latest 15.1 with "turbo fib" on RE-1800 that
> can do convergence of multiple feed full table in about 55 seconds. And
> that is still a single core RPD process.
> >
> >> On Jul 28, 2016, at 2:34 PM, Matthew Crocker <matthew at corp.crocker.com>
> wrote:
> >>
> >>
> >> Mike,
> >>
> >> Here is the view of my MX80.   This router has a couple full tables and
> a bunch of peers through various IXes.   I have an MX480 on order to
> replace this MX80.   I’ll use this a dedicated IX peering router so I won’t
> have full tables on my IX border later this year.
> >>
> >> The MX80 has horrific full table convergence (8 minutes +/-).  The
> MX104 is a bit better.  You would need to go to a MX240 with a real RE to
> get decent convergence times.
> >>
> >> matthew at MX80> show bgp summary
> >> Groups: 10 Peers: 15 Down peers: 0
> >>
> >> matthew at MX80> show route summary
> >> Autonomous system number: XXXX
> >> Router ID: A.B.C.D
> >>
> >> inet.0: 614169 destinations, 1807913 routes (614160 active, 10
> holddown, 0 hidden)
> >> Restart Complete
> >>             Direct:      7 routes,      7 active
> >>              Local:      6 routes,      6 active
> >>               OSPF:    511 routes,    508 active
> >>                BGP: 1807386 routes, 613636 active
> >>             Static:      1 routes,      1 active
> >>                LDP:      2 routes,      2 active
> >>
> >> inet6.0: 14443 destinations, 28877 routes (14443 active, 0 holddown, 0
> hidden)
> >> Restart Complete
> >>             Direct:      6 routes,      4 active
> >>              Local:      6 routes,      6 active
> >>                BGP:  28865 routes,  14433 active
> >>
> >>
> >> matthew at MX80> show system memory
> >> System memory usage distribution:
> >>      Total memory: 2072576 Kbytes (100%)
> >>   Reserved memory:   36896 Kbytes (  1%)
> >>      Wired memory:  302092 Kbytes ( 14%)
> >>     Active memory: 1399432 Kbytes ( 67%)
> >>   Inactive memory:  120000 Kbytes (  5%)
> >>      Cache memory:   69720 Kbytes (  3%)
> >>       Free memory:  143680 Kbytes (  6%)
> >> Memory disk resident memory:  349640 Kbytes
> >> VM-Kbytes(  %  ) Resident(  %  ) Map-name
> >>  913972(87.16)   343424(16.56) kernel
> >>
> >> matthew at MX80> show system processes summary
> >> last pid: 34226;  load averages:  0.24,  0.31,  0.23  up 477+00:51:09
>   18:31:50
> >> 142 processes: 4 running, 110 sleeping, 28 waiting
> >>
> >> Mem: 1367M Active, 117M Inact, 295M Wired, 68M Cache, 112M Buf, 140M
> Free
> >> Swap: 2915M Total, 2915M Free
> >>
> >>
> >> —
> >>
> >> Matthew Crocker
> >> President - Crocker Communications, Inc.
> >> Managing Partner - Crocker Telecommunications, LLC
> >> E: matthew at corp.crocker.com
> >> E: matthew at crocker.com
> >>
> >>
> >>> On Jul 28, 2016, at 12:09 PM, Mike <mike+jnsp at willitsonline.com>
> wrote:
> >>>
> >>> On 07/28/2016 12:50 AM, Adam Vitkovsky wrote:
> >>>>
> >>>> And on how effective is the NPU's lookup process, that is how
> effective is the actual lookup algorithm with CPU cycles and memory
> accesses, some NPUs can even offload complex lookup tasks to a specialized
> chip.
> >>>
> >>> I appreciate your presence on other forums, but I'm pretty sure nobody
> here needs a basic explanation of how modern router platforms work. If you
> missed it, the question was specifically about juniper and bang for the
> buck and routing bgp on 10g and filtering.
> >>>
> >>> Some folks helpfully suggested using strategies to to decrease the
> required size of the FIB, potentially meaning a lower box could do that
> job. That has some merit, as the OP was right in that for this job I don't
> really care about timbuktu more as whats 'close' to my two ip transit
> providers. I know nothing of juniper and I'm just wondering if MX80 is
> enough box for this or if I need to go higher up in the food chain. The one
> iptransit provider at my 'A' location appears to originate about 20
> networks from various netblocks and this would be easy to statically enter
> into config while accepting defaults from both, achieving the same net
> result.
> >>>
> >>> Mike-
> >>> _______________________________________________
> >>> juniper-nsp mailing list juniper-nsp at puck.nether.net
> >>> https://puck.nether.net/mailman/listinfo/juniper-nsp
> >>
> >>
> >> _______________________________________________
> >> juniper-nsp mailing list juniper-nsp at puck.nether.net
> >> https://puck.nether.net/mailman/listinfo/juniper-nsp
> > _______________________________________________
> > juniper-nsp mailing list juniper-nsp at puck.nether.net
> > https://puck.nether.net/mailman/listinfo/juniper-nsp
>

-- 
Jim Troutman,
jamesltroutman at gmail.com
800-605-0192 (main)