Multilink PPP

Discussion in 'Cisco' started by Smash, Jan 11, 2004.

  1. Smash

    Smash Guest

    Can I bind together 4 T1s successfully on a
    2611 series router using Multilink PPP?

    Do you think a 2611 is powerful enough to handle this?


    w0rd,


    -Tg
     
    Smash, Jan 11, 2004
    #1
    1. Advertising

  2. (Smash) writes:
    >Can I bind together 4 T1s successfully on a
    >2611 series router using Multilink PPP?


    >Do you think a 2611 is powerful enough to handle this?


    Yes I think a 2611 is powerful enough to do MLPPP on 4 DS1's if you
    aren't doing NAT, CBAC or VPN encryption (without the AIM) on the
    resulting link. The realworld packet throughput I typically see on my
    busiest DS1 customers going full out times 4 is well within the
    packet-per-sec rating of this platform.

    It might get a bit dicey to do 4 DS1 MLPPP and NAT or CBAC all at the
    same time on that platform. Nor would I do more than 4 on this box.
    --
    Doug McIntyre
    Network Engineer/Jack of All Trades
    Vector Internet Services, Inc.
     
    Doug McIntyre, Jan 11, 2004
    #2
    1. Advertising

  3. On 11 Jan 2004 15:49:42 GMT, Doug McIntyre <> wrote:

    > (Smash) writes:
    >>Can I bind together 4 T1s successfully on a
    >>2611 series router using Multilink PPP?

    >
    >>Do you think a 2611 is powerful enough to handle this?

    >
    >Yes I think a 2611 is powerful enough to do MLPPP on 4 DS1's if you
    >aren't doing NAT, CBAC or VPN encryption (without the AIM) on the
    >resulting link. The realworld packet throughput I typically see on my
    >busiest DS1 customers going full out times 4 is well within the
    >packet-per-sec rating of this platform.


    Cisco's pps ratings are with 64-byte packets, which makes it difficult
    to correlate them with maximum throughput since larger packets are
    typically used in the real world.

    That being said, a 2611 should be able to saturate 6Mbits worth of T1s
    regardless of what the average packet size is. It's rated at 15k pps,
    which is around 7.5Mbps with 64-byte packets. Note however that
    figuring out the throughput potential with full-sized packets (or any
    size other than 64 bytes for that matter) isn't as easy as multiplying
    1500 X 15k, since pps decreases as packet size increases. (A 2621,
    rated at 25k pps, can do about 40Mbps with full-sized packets per some
    testing that I did last year. So a 2611 should max out at around
    25Mbps or so if the pps/throughput ratio stays the same.)

    MLPPP will likely lower the throughput potential, however, as extra
    CPU cycles are required. But just how many CPU cycles seems to depend
    on what hardware and IOS version is involved, so it's hard to take a
    guess without running some tests.

    -Terry
     
    Terry Baranski, Jan 11, 2004
    #3
  4. On Sun, 11 Jan 2004 18:05:08 -0500, Terry Baranski wrote:
    > On 11 Jan 2004 15:49:42 GMT, Doug McIntyre <> wrote:
    >
    >> (Smash) writes:
    >>>Can I bind together 4 T1s successfully on a
    >>>2611 series router using Multilink PPP?

    >>
    >>>Do you think a 2611 is powerful enough to handle this?

    >>
    >>Yes I think a 2611 is powerful enough to do MLPPP on 4 DS1's if you
    >>aren't doing NAT, CBAC or VPN encryption (without the AIM) on the
    >>resulting link. The realworld packet throughput I typically see on my
    >>busiest DS1 customers going full out times 4 is well within the
    >>packet-per-sec rating of this platform.

    >
    > Cisco's pps ratings are with 64-byte packets, which makes it difficult
    > to correlate them with maximum throughput since larger packets are
    > typically used in the real world.


    The reason the testing is done with 64 byte packets, is that it is
    what stresses the router the most.

    And in general it's the number of packets per second that limits a
    router, not the bits per second, until you hit the interface speed.

    --
    Jesper Skriver, CCIE #5456, FreeBSD committer
     
    Jesper Skriver, Jan 11, 2004
    #4
  5. On 11 Jan 2004 23:39:55 GMT, Jesper Skriver <> wrote:
    >>
    >> Cisco's pps ratings are with 64-byte packets, which makes it difficult
    >> to correlate them with maximum throughput since larger packets are
    >> typically used in the real world.

    >
    >The reason the testing is done with 64 byte packets, is that it is
    >what stresses the router the most.


    I've always assumed that 64 byte packets are used because (in general)
    the smaller the packet size, the higher the packets-per-second.

    >And in general it's the number of packets per second that limits a
    >router, not the bits per second, until you hit the interface speed.


    In my experience PPS tends to drop dramatically as packet size
    increases, so from my standpoint both packets and bits play a role.

    -Terry
     
    Terry Baranski, Jan 12, 2004
    #5
  6. On Sun, 11 Jan 2004 20:53:20 -0600, Terry Baranski wrote:
    > On 11 Jan 2004 23:39:55 GMT, Jesper Skriver <> wrote:
    >>>
    >>> Cisco's pps ratings are with 64-byte packets, which makes it difficult
    >>> to correlate them with maximum throughput since larger packets are
    >>> typically used in the real world.

    >>
    >>The reason the testing is done with 64 byte packets, is that it is
    >>what stresses the router the most.

    >
    > I've always assumed that 64 byte packets are used because (in general)
    > the smaller the packet size, the higher the packets-per-second.
    >
    >>And in general it's the number of packets per second that limits a
    >>router, not the bits per second, until you hit the interface speed.

    >
    > In my experience PPS tends to drop dramatically as packet size
    > increases, so from my standpoint both packets and bits play a role.


    I do not agree, below are some actual measurements from a random Cisco
    router with software based forwarding, between 2 GigE interfaces.

    no drop forwarding rate, packets per second
    64 byte packets 748502
    128 byte packets 744048
    256 byte packets 753012
    512 byte packets 468164 this is 99.63% of the theoretical limit

    The packets per second forwarding rate is basicly unchanged within the
    accuracy of the measurement, untill we hit the interface speed limit.

    /Jesper

    --
    Jesper Skriver, CCIE #5456, FreeBSD committer
     
    Jesper Skriver, Jan 12, 2004
    #6
  7. Smash

    AnyBody43 Guest

    Doug McIntyre <> wrote in message news:<40017096$0$546$>...
    > (Smash) writes:
    > >Can I bind together 4 T1s successfully on a
    > >2611 series router using Multilink PPP?

    >
    > >Do you think a 2611 is powerful enough to handle this?

    >
    > Yes I think a 2611 is powerful enough to do MLPPP on 4 DS1's if you
    > aren't doing NAT, CBAC or VPN encryption (without the AIM) on the
    > resulting link. The realworld packet throughput I typically see on my
    > busiest DS1 customers going full out times 4 is well within the
    > packet-per-sec rating of this platform.
    >
    > It might get a bit dicey to do 4 DS1 MLPPP and NAT or CBAC all at the
    > same time on that platform. Nor would I do more than 4 on this box.


    Hi,

    I think that there is a pretty large processing overhead with
    MLPPP. Each packet gets split up and a piece is sent down
    each link in the bundle.

    I am not too close to this nowadays however some tests that I did
    a few years ago (details now lost) indicated that MLPPP sucked up
    pretty much a whole 4700 in the case of 4xE1 (2Mbps) and small
    packets. (I think that a 4700 was quite like a 3640, or better,
    in terms of CPU)

    My feeling is that a 2611 would be marginal in this application.
    I would certainly not want to sell one for it in case it bounced
    right back:(

    As was discussed by other respondents the support for MLPPP
    does vary by platform. I seem to recall that it was claimed that
    "particle buffers" helped a lot with MPPP, and there is the distinct
    possibility that other hardware features could make a substantial
    difference to the performance. The 2611 will probably not have
    these features.
     
    AnyBody43, Jan 12, 2004
    #7
  8. On 12 Jan 2004 12:06:46 GMT, Jesper Skriver <> wrote:
    >
    >I do not agree, below are some actual measurements from a random Cisco
    >router with software based forwarding, between 2 GigE interfaces.
    >
    > no drop forwarding rate, packets per second
    >64 byte packets 748502
    >128 byte packets 744048
    >256 byte packets 753012
    >512 byte packets 468164 this is 99.63% of the theoretical limit
    >
    >The packets per second forwarding rate is basicly unchanged within the
    >accuracy of the measurement, untill we hit the interface speed limit.


    My experience has been the opposite on the few devices that I've
    tested, as well as sporadic results that I've seen posted by others.
    The 2621 tests I mentioned earlier are an example: from a
    packets-per-second standpoint, a max of 40Mbps with 1500 byte packets
    is nowhere near the router's rating of 25k (64 byte) pps. Another
    example: ~75Mbps max on a 3620 per a fellow tester's results, despite
    it's rating of 20-40k (64 byte) packets-per-second.

    So it's likely platform-dependant, though from what I've seen so far
    pps is likely to drop off as packet size increases regardless of
    interface saturation.

    -Terry
     
    Terry Baranski, Jan 13, 2004
    #8
  9. Smash

    Hansang Bae Guest

    In article <>,
    0VE says...
    > So it's likely platform-dependant, though from what I've seen so far
    > pps is likely to drop off as packet size increases regardless of
    > interface saturation.


    It does in that CPU and memory matters. It just depends on whether the
    box is bound by buffer space or memory.


    --

    hsb

    "Somehow I imagined this experience would be more rewarding" Calvin
    *************** USE ROT13 TO SEE MY EMAIL ADDRESS ****************
    ********************************************************************
    Due to the volume of email that I receive, I may not not be able to
    reply to emails sent to my account. Please post a followup instead.
    ********************************************************************
     
    Hansang Bae, Jan 13, 2004
    #9
  10. Smash

    Hansang Bae Guest

    In article <>,
    says...
    > It does in that CPU and memory matters. It just depends on whether the
    > box is bound by buffer space or memory.


    Of course I meant "bound by buffer or CPU"


    --

    hsb

    "Somehow I imagined this experience would be more rewarding" Calvin
    *************** USE ROT13 TO SEE MY EMAIL ADDRESS ****************
    ********************************************************************
    Due to the volume of email that I receive, I may not not be able to
    reply to emails sent to my account. Please post a followup instead.
    ********************************************************************
     
    Hansang Bae, Jan 13, 2004
    #10
  11. Smash

    Rob Guest

    To the original poster.

    I know for a fact that a 2611 is not enough to run four T1's
    multilinked. Get at least a 2651 or a newer 2621XM at a minimum.

    The older 3640's will work fine too.

    I have several routers in my organization where I do multilink and am
    quite familiar with the limitations of the 2600 routers.

    And off topic slightly, the 2611 is only 10Mb half-duplex Ethernet.
    Fast Ethernet would be better for a combined maximum of 12Mb
    throughput.

    -Robert



    On 12 Jan 2004 10:17:32 -0800, (AnyBody43)
    wrote:

    >Doug McIntyre <> wrote in message news:<40017096$0$546$>...
    >> (Smash) writes:
    >> >Can I bind together 4 T1s successfully on a
    >> >2611 series router using Multilink PPP?

    >>
    >> >Do you think a 2611 is powerful enough to handle this?

    >>
    >> Yes I think a 2611 is powerful enough to do MLPPP on 4 DS1's if you
    >> aren't doing NAT, CBAC or VPN encryption (without the AIM) on the
    >> resulting link. The realworld packet throughput I typically see on my
    >> busiest DS1 customers going full out times 4 is well within the
    >> packet-per-sec rating of this platform.
    >>
    >> It might get a bit dicey to do 4 DS1 MLPPP and NAT or CBAC all at the
    >> same time on that platform. Nor would I do more than 4 on this box.

    >
    >Hi,
    >
    >I think that there is a pretty large processing overhead with
    >MLPPP. Each packet gets split up and a piece is sent down
    >each link in the bundle.
    >
    >I am not too close to this nowadays however some tests that I did
    >a few years ago (details now lost) indicated that MLPPP sucked up
    >pretty much a whole 4700 in the case of 4xE1 (2Mbps) and small
    >packets. (I think that a 4700 was quite like a 3640, or better,
    >in terms of CPU)
    >
    >My feeling is that a 2611 would be marginal in this application.
    >I would certainly not want to sell one for it in case it bounced
    >right back:(
    >
    >As was discussed by other respondents the support for MLPPP
    >does vary by platform. I seem to recall that it was claimed that
    >"particle buffers" helped a lot with MPPP, and there is the distinct
    >possibility that other hardware features could make a substantial
    >difference to the performance. The 2611 will probably not have
    >these features.
     
    Rob, Jan 13, 2004
    #11
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. Gerd Thümmler

    ppp multilink lost many fragments

    Gerd Thümmler, Jul 11, 2003, in forum: Cisco
    Replies:
    0
    Views:
    2,889
    Gerd Thümmler
    Jul 11, 2003
  2. Klaus Kruse

    Re: ppp multilink on 2500/3600

    Klaus Kruse, Oct 23, 2003, in forum: Cisco
    Replies:
    1
    Views:
    672
    Aaron Leonard
    Oct 27, 2003
  3. Gerd Thuemmler
    Replies:
    0
    Views:
    2,344
    Gerd Thuemmler
    Nov 27, 2003
  4. Richard Field

    Multilink PPP for 4 T1's?

    Richard Field, Jan 14, 2004, in forum: Cisco
    Replies:
    6
    Views:
    8,957
    AnyBody43
    Jan 16, 2004
  5. Karnov
    Replies:
    1
    Views:
    8,541
    CCIE8122
    May 24, 2004
Loading...

Share This Page