Ping slow on serial interface

Discussion in 'Cisco' started by ns, Dec 14, 2004.

  1. ns

    ns Guest

    Hi all,

    i have two cisco router serie 1700 connected by serial link (193 KBytes/s or
    1544 Kbits).

    The speed access to the remote router seem to be slow. For exemple, when i
    lunch ping, the response is slow...
    =================================================================
    MyRouterA#ping ip
    Target IP address: 192.168.1.2
    Repeat count [5]: 30
    Datagram size [100]: 1500
    Timeout in seconds [2]:
    Extended commands [n]:
    Sweep range of sizes [n]:
    Type escape sequence to abort.
    Sending 30, 1500-byte ICMP Echos to 192.168.1.2, timeout is 2 seconds:
    !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
    Success rate is 100 percent (30/30), round-trip min/avg/max = 180/182/184 ms
    MyRouterA#
    =================================================================
    The serial configuration of my router (MyRouterA) is :

    interface Serial0
    ip address 192.168.1.1 255.255.255.252
    ==================================
    i added commande "bandwidth=1544" but i have the same problem.
    i confirm that's the bandwidth is not saturated, i juste have 3% utilization
    of 193 Kbytes

    Can you tell me what's i need to do please ?

    ThankYou very much for your help

    Best Regadrs
    NS
     
    ns, Dec 14, 2004
    #1
    1. Advertisements

  2. On 14.12.2004 19:03 ns wrote

    Tell us how far away the other end is (YourRouterB). While 180ms is
    pretty ok when you are in NYC and YourRouterB is in Hongkong, this does
    not longer hold if you have connected both routers back to back.
     
    Arnold Nipper, Dec 14, 2004
    #2
    1. Advertisements

  3. ns

    Erik Freitag Guest

    How far apart are these routers? Ping times reflect latency, not bandwidth
    (usually). If you do a show interface serial 0 (or whatever the serial is
    called), what is the delay shown? Look for DLY 20000 usec, or something
    like that. What kind of circuit connects them? PRI? Frame Relay?
    Point-to-Point? The bandwidth setting is used by routing protocols and
    some network management tools, but does not affect your line speed or
    latency.
     
    Erik Freitag, Dec 14, 2004
    #3
  4. At size 1500, shances are you are fragmenting packets, andc fragmenting
    kills performance.

    Try 1400, or whatever the max is for your medium.

    In the extended commands, ask to set the DF (Don't Fragment) bit to probe
    for the max value.
     
    Phillip Remaker, Dec 14, 2004
    #4
  5. ns

    Toby Guest

    Hi

    From your config (what you have provided) I can see that you are just
    pinging the remote end of your point to point connection so have no other
    layer3 (IP) nodes in between.

    The speed the signal travels depends on what it is travelling through but
    usually can be thought of as 70% of the speed of light through a vacuum
    (dont quote me on this figure as it is your problem so I have not googled
    it, thats your job). The speed of light in a vacuum though is approx
    300,000km/sec at 70% speed it would equate to 210,000km. So in 180ms the
    distance travelled would be 5.xx times round the earth.

    I dont think your connection is this long so if you are 100% sure you have
    no queueing due to congestion (you state 3% utilistation) then there must be
    other devices in the path between your 2 routers at layer2, (frame relay/ATM
    etc) the bandwidth on these devices will be shared with other users and you
    will get switching/queing delays adding to the overall delay.

    N.B. most Cisco routers default to an average of traffic over 5 minutes (300
    seconds) and can be configured to read an average over 30 seconds, so please
    ensure you are reading your utilisation correctly.

    If you are 100% sure then

    Contact your service provider to establish what technology is being offered
    and what service level agreement in regards to delay when your bandwidth is
    in contract is being offered for the destination you are connected to.

    Just to clarify. If you had a dedicated point-point link between site A and
    site B your data would travel at about 70% speed of light once allowed onto
    the link. (bandwidth is a mesurement of what is allowed/able to enter the
    link per second and not how quick it travels in the link). So 184ms would be
    5.xx times the circumrence of the world. On a dedicated point-point link the
    data you sent would not have a coffee break so would not be slowed down.
    This indicates you have NOT got a dedicated point to point link. If you did
    have a p-p link 5 times the circumfrence of the earth you would be paying
    approx $5,000,000 per month for it (figure made up for explanation purposes
    only).

    As it is clear you have not got a dedicated p-p link here you will be going
    through a shared medium such as frame relay/ATM, the problem could be either
    of the following.

    1) queing delays in your providers network (as well as
    switching/serialisation and the propogation (distance) delays already spoken
    about), although queuing/propogation delays are the main factor here. See my
    note above about contacting your service provider for your expectations.

    2) the remote site Router B may have a problem with utilisation on it's link
    from the service provider. i.e. your company may have many other sites
    connecting with site B and although site A only has 3% util site B might be
    nearing capacity. (I include this as you have no mention of this sites
    config/util)

    So if site A's link is only 3% util and you can clarify that site B is also
    only 3% util with no other traffic then you will need to clarify your
    service contract with your provider as there are delays in the service
    providers network.

    Toby
     
    Toby, Dec 14, 2004
    #5
  6. On 15.12.2004 00:36 Toby wrote

    If the signal travels with 210,000 km/s and if you see a RTT of 180ms,
    the distance to B-Router is 210,000 * 0.18 / 2 which gives 18,900km.
    This is less than half time round the earth on a great circle.

    When I do a traceroute from here (Heidelberg, Germany) 180ms RTT
    typically means something on the US westcoast, 100ms - 120ms is US
    eastcoast, APA is something like 300ms.

    Please also bear in mind that it takes some time to serialize the
    packet. On slow links and with large packetsize this takes a non
    negligable amount of time.




    Arnold
     
    Arnold Nipper, Dec 14, 2004
    #6
  7. ns

    Erik Freitag Guest

    Just to clarify further, the speed of light in fiber is around 2/3 the
    speed of light in a vacuum, or around 200000 km/s. Ping time is 184 ms so
    a one-way trip is around 92 ms. 200000 km/s = 124300 mi/s = 124.3 mi/ms or
    10192 miles. This is possible on my planet, just a little more than the
    great circle distance from London to Tokyo.
     
    Erik Freitag, Dec 15, 2004
    #7
  8. ns

    Jonathan Guest

    What's wrong with the ping time?

    This is pretty decent for a 1500 byte ping.


    =================================================================
    Because that command is the default (well, 'bandwidth 1536' is), and it is
    only used so EIGRP can accurately calculate its metric.
    Nothing.





    Jonathan
     
    Jonathan, Dec 15, 2004
    #8
  9. ns

    Jonathan Guest

    Which doesn't account for encapsulation, serialization delays, frame-relay
    switch latency, and FRATM latency.

    There is nothing wrong with his ping times.
    The SLA for telcos is 120ms for a 100-byte packet, end to end from our
    ingress switches.

    He is getting about 45% higher ping times for 1400% more data being sent. He
    needs to calm down.



    Jonathan
     
    Jonathan, Dec 15, 2004
    #9
  10. ns

    ns Guest

    MyRouterA is located at Paris
    MyRouterB is located at San Francisco
     
    ns, Dec 15, 2004
    #10
  11. ns

    ns Guest

    DLY = 20000 usec
    Circuit : PPP

    more informations a bout serial interface
    ======================================================
    MyRouterA#sh int s0
    Serial0 is up, line protocol is up
    Hardware is PowerQUICC Serial
    Internet address is 192.168.25.10/30
    MTU 1500 bytes, BW 1544 Kbit, DLY 20000 usec,
    reliability 255/255, txload 1/255, rxload 1/255
    Encapsulation HDLC, loopback not set
    Keepalive set (10 sec)
    Last input 00:00:03, output 00:00:00, output hang never
    Last clearing of "show interface" counters 4d19h
    Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 16
    Queueing strategy: weighted fair
    Output queue: 0/1000/64/16 (size/max total/threshold/drops)
    Conversations 0/13/256 (active/max active/max total)
    Reserved Conversations 0/0 (allocated/max allocated)
    Available Bandwidth 1158 kilobits/sec
    ======================================================

    Thanks a lot
    NS
     
    ns, Dec 15, 2004
    #11
  12. ns

    ns Guest

    I tested with default datagram size (100), 1400 and 1500 and i have the same
    result.
    I tested with option "Set DF bit in IP header? [no]: yes" and i have the
    same result

    Thank You very much
    NS
     
    ns, Dec 15, 2004
    #12
  13. ns

    ns Guest

    Yes, exactly
    ThankYou very much for your answer
    NS
     
    ns, Dec 15, 2004
    #13
  14. ns

    ns Guest


    i have the same result when i test with small packet

    For exemple :
    =============================================================
    MyRouterA#ping ip
    Target IP address: 192.168.1.2
    Repeat count [5]: 30
    Datagram size [100]:
    Timeout in seconds [2]:
    Extended commands [n]:
    Sweep range of sizes [n]:
    Type escape sequence to abort.
    Sending 30, 100-byte ICMP Echos to 192.168.1.2, timeout is 2 seconds:
    !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
    Success rate is 100 percent (30/30), round-trip min/avg/max = 164/166/168 ms
    =============================================================

    it's too late !!
     
    ns, Dec 15, 2004
    #14
  15. Your packets have to travel 6,000 miles and you think 180 ms is slow?

    You'll get a speeding ticket if you try to exceed the speed of light.
     
    Barry Margolin, Dec 15, 2004
    #15
  16. ns

    ns Guest

    Your packets have to travel 6,000 miles and you think 180 ms is slow?

    ....
     
    ns, Dec 15, 2004
    #16
  17. ns

    Ivan Ostreš Guest

    ....and for some QoS mechanisms, and for interface load counter, and
    bunch of other not well-known things.
     
    Ivan Ostreš, Dec 15, 2004
    #17
  18. ns

    anybody43 Guest

    Hi,

    "= 164/166/168 ms it's too late !!"

    Thing is, if it really is too late you will *have* to move one end
    nearer to the other to help.

    I am in London UK and I am happy with the numbers below. The ONLY
    way to change them is to move one of - me, San Jose, a university
    or similar in Australia, Imperial College in London, or Syracuse
    university.

    None of these is going to happen anytime soon, so I have to live with
    the
    result:)

    It is just possible that your provider is using some weird routing and
    that
    your sites are not really 6000 miles apart. If this is the case you
    may be able to use a different one. Upgrading to a higher speed line
    will
    reduce the tranmission delay, but with a 2M line that is only 8ms at
    each end for a 1500 byte packet anyway so it will only bring it down to
    150ms if you eliminate the transmission delay altogether.

    H:\>ping www.cisco.com
    Pinging www.cisco.com [198.133.219.25] with 32 bytes of data:
    Reply from 198.133.219.25: bytes=32 time=155ms TTL=116
    Reply from 198.133.219.25: bytes=32 time=153ms TTL=116
    Reply from 198.133.219.25: bytes=32 time=153ms TTL=116
    Reply from 198.133.219.25: bytes=32 time=152ms TTL=116
    Ping statistics for 198.133.219.25:
    Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
    Approximate round trip times in milli-seconds:
    Minimum = 152ms, Maximum = 155ms, Average = 153ms

    H:\>ping flinders.edu.au
    Pinging flinders.edu.au [129.96.252.36] with 32 bytes of data:
    Reply from 129.96.252.36: bytes=32 time=335ms TTL=47
    Reply from 129.96.252.36: bytes=32 time=333ms TTL=47
    Reply from 129.96.252.36: bytes=32 time=333ms TTL=47
    Reply from 129.96.252.36: bytes=32 time=333ms TTL=47
    Ping statistics for 129.96.252.36:
    Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
    Approximate round trip times in milli-seconds:
    Minimum = 333ms, Maximum = 335ms, Average = 333ms

    H:\>ping www.doc.ic.ac.uk
    Pinging linnet.doc.ic.ac.uk [146.169.1.10] with 32 bytes of data:
    Reply from 146.169.1.10: bytes=32 time=12ms TTL=48
    Reply from 146.169.1.10: bytes=32 time=12ms TTL=48
    Reply from 146.169.1.10: bytes=32 time=12ms TTL=48
    Reply from 146.169.1.10: bytes=32 time=11ms TTL=48
    Ping statistics for 146.169.1.10:
    Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
    Approximate round trip times in milli-seconds:
    Minimum = 11ms, Maximum = 12ms, Average = 11ms

    H:\>ping www.syracuse.edu
    Pinging syracuse.edu [128.230.18.35] with 32 bytes of data:
    Reply from 128.230.18.35: bytes=32 time=99ms TTL=239
    Reply from 128.230.18.35: bytes=32 time=96ms TTL=239
    Reply from 128.230.18.35: bytes=32 time=95ms TTL=239
    Reply from 128.230.18.35: bytes=32 time=97ms TTL=239
    Ping statistics for 128.230.18.35:
    Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
    Approximate round trip times in milli-seconds:
    Minimum = 95ms, Maximum = 99ms, Average = 96ms
     
    anybody43, Dec 15, 2004
    #18
  19. ns

    Toby Guest

    Arnold I did my maths wrong I forgot the round trip part and multiplied in
    error instead of divided the .18 (oops It had been a long day)

    It doesn't alter most of my other facts though. I was trying to point out
    that delays inside a service providers network at layer2 have to be taken
    into account as very few companies will be able to justify the cost of a
    pure point-point link over very large distances and the SP contract has to
    be taken into account over a shared medium such as a Frame-relay/ATM etc
    network.

    My comment on serialisation was based on the service providers trunks being
    usually fast and I ignored this on the local routers bandwidth as this would
    be the same delay for 1mile or 20000miles of link. But you are quite right
    they should be accounted for.

    If I had done my maths correctly to start with I would have noticed the
    distance was infact in the terrestial range and I would have added in a
    serialisation factor for a slow link in my example, I didn't because no
    single hop router delay with a 3% util of a link could cause the kind of
    delay to make it look like 5 times round the world.

    At least I can show my kids with this example why they must show there
    working in there upcomming maths exam.

    Thanks for the correction

    Toby
     
    Toby, Dec 15, 2004
    #19
  20. ns

    Erik Freitag Guest

    Except that he said it is a point-to-point T1, which should mean there's
    no switching. The only encap/decapsulation time should be at his two
    routers, serialization might be an issue but I don't think the 1700s have
    big enough buffers to worry about this.
    Except that they are a lot longer than you would expect for a short
    point-to-point T1.
    Which telcos? My carrier has an 80ms domestic commitment within the
    continental US and Europe.
    It is not the data size causing the long round-trip times, packet size has
    more to do with throughput than latency - I get better ping times going
    through the Internet - he's on a point-to-point. I agree there's no reason
    to panic, but maybe he has an application that depends on better times, or
    maybe he's just trying to improve his latency.

    router#ping
    Protocol [ip]:
    Target IP address: www.yahoo.com
    Translating "www.yahoo.com"...domain server (24.143.128.68) [OK]

    Repeat count [5]: 30
    Datagram size [100]: 1500
    Timeout in seconds [2]:
    Extended commands [n]:
    Sweep range of sizes [n]:
    Type escape sequence to abort.
    Sending 30, 1500-byte ICMP Echos to 66.94.230.34, timeout is 2 seconds:
    !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
    Success rate is 100 percent (30/30), round-trip min/avg/max = 16/38/144 ms
     
    Erik Freitag, Dec 15, 2004
    #20
    1. Advertisements

Ask a Question

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

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.