The proper size of ping packet (is fragmentation allowed)

Discussion in 'Cisco' started by Yang Jun, Dec 17, 2003.

  1. Yang Jun

    Yang Jun Guest

    Hi, I am just wondering what is the "legal" size of ping command. I
    remember vaguely (could be wrong) that the payload should be less that
    1500bit so that it can be put in a single ethernet packet. But
    according to Cisco, their datagram size is in the range from 40 to
    18024.

    The question is: if datagram size is large (so fragmentation is
    needed)and mutiple packets are used to carry the ping-data, how to
    calculate the response time?

    Thanks.
    Jun
    Yang Jun, Dec 17, 2003
    #1
    1. Advertising

  2. * Yang Jun wrote:
    > Hi, I am just wondering what is the "legal" size of ping command.


    64 to 65535 byte incl. ICMP/IP header.

    > I remember vaguely (could be wrong) that the payload should be less that
    > 1500bit so that it can be put in a single ethernet packet.


    64 to 1500 byte is the legal payload of an ordinary ethernet frame.
    There are other types of ethernet frames with different legal payload sizes.
    Ethernet I -> 1500 byte always
    Ethernet II -> 64 - 1500 byte determined by the physical layer
    802.3 -> 64 - 1500 byte determined by a length field
    Gigabit Eth -> 64 - 18... byte determined by a length field

    The smallest IP header is 20 byte.
    The ICMP header is 8 byte.

    > But according to Cisco, their datagram size is in the range from 40 to
    > 18024.


    The minimum would be 64 - 20 - 8 = 36 byte ICMP payload.
    The maximum (unfragmented) occours on Gigabit links.

    > The question is: if datagram size is large (so fragmentation is
    > needed)and mutiple packets are used to carry the ping-data, how to
    > calculate the response time?


    Roundtrip = Fragmentation + Transmit + Reassemby + Answer + the way back.
    Lutz Donnerhacke, Dec 17, 2003
    #2
    1. Advertising

  3. Yang Jun

    Yang Jun Guest

    Lutz Donnerhacke <> wrote in message news:<-jena.de>...
    >
    > > The question is: if datagram size is large (so fragmentation is
    > > needed)and mutiple packets are used to carry the ping-data, how to
    > > calculate the response time?

    >
    > Roundtrip = Fragmentation + Transmit + Reassemby + Answer + the way back.


    If frag/reassemble is invovled, there could be outstanding packets
    with long delay due to taking different route. This waiting period
    make it hard to comprehend the real vaule of ping. Is the concept of
    "average" introduced due to variety of connections (e.g., load
    balancing.)

    Best Regards,
    Jun
    Yang Jun, Dec 17, 2003
    #3
  4. Yang Jun

    Ivan Guest

    "Yang Jun" <> wrote in message
    news:...
    > Lutz Donnerhacke <> wrote in message

    news:<-jena.de>...
    > >
    > > > The question is: if datagram size is large (so fragmentation is
    > > > needed)and mutiple packets are used to carry the ping-data, how to
    > > > calculate the response time?

    > >
    > > Roundtrip = Fragmentation + Transmit + Reassemby + Answer + the way

    back.
    >
    > If frag/reassemble is invovled, there could be outstanding packets
    > with long delay due to taking different route. This waiting period
    > make it hard to comprehend the real vaule of ping. Is the concept of
    > "average" introduced due to variety of connections (e.g., load
    > balancing.)
    >


    What is a "real value" of ping?

    Ivan
    Ivan, Dec 18, 2003
    #4
  5. Yang Jun

    Yang Jun Guest

    "Ivan" <> wrote in message news:<brs0q8$6tqdi$-berlin.de>...
    > >
    > > If frag/reassemble is invovled, there could be outstanding packets
    > > with long delay due to taking different route. This waiting period
    > > make it hard to comprehend the real vaule of ping. Is the concept of
    > > "average" introduced due to variety of connections (e.g., load
    > > balancing.)
    > >

    >
    > What is a "real value" of ping?
    >
    > Ivan


    Sorry, I mean since one important usage of ping is to determine
    response time (RTD, round trip delay). If multiple packets are used
    just to carry the jumbo test data, considering delay variation for
    different packets on the trip, the reliability of response time
    obtained this way could be inaccurate.

    Brgds
    Yang Jun, Dec 19, 2003
    #5
  6. * Yang Jun wrote:
    > Sorry, I mean since one important usage of ping is to determine
    > response time (RTD, round trip delay).


    No. ping can't provide this. You need a remote monitor for this purpose in
    order to determine the real transport delay instead of the current load of
    the responder.
    Lutz Donnerhacke, Dec 19, 2003
    #6
  7. Yang Jun

    shope Guest

    "Lutz Donnerhacke" <> wrote in message
    news:-jena.de...
    > * Yang Jun wrote:
    > > Hi, I am just wondering what is the "legal" size of ping command.

    >
    > 64 to 65535 byte incl. ICMP/IP header.
    >
    > > I remember vaguely (could be wrong) that the payload should be less that
    > > 1500bit so that it can be put in a single ethernet packet.


    Sending pings bigger than max frame size is useful - it can tell you if the
    interface is handling fragmentation correctly.
    >
    > 64 to 1500 byte is the legal payload of an ordinary ethernet frame.
    > There are other types of ethernet frames with different legal payload

    sizes.
    > Ethernet I -> 1500 byte always
    > Ethernet II -> 64 - 1500 byte determined by the physical layer
    > 802.3 -> 64 - 1500 byte determined by a length field
    > Gigabit Eth -> 64 - 18... byte determined by a length field
    >
    > The smallest IP header is 20 byte.
    > The ICMP header is 8 byte.
    >
    > > But according to Cisco, their datagram size is in the range from 40 to
    > > 18024.


    approx 18k bytes is the max size supported by Cisco on a 16 Mbps T/R.

    It looks like they have limited you to the largest frame size supported on
    the router.
    >
    > The minimum would be 64 - 20 - 8 = 36 byte ICMP payload.
    > The maximum (unfragmented) occours on Gigabit links.


    Last time i checked the max jumbo frame size supported by cisco was 9.2
    Kbytes, but this varies by equipment type - e.g. you may get slightly
    different answers on a IOS router such as a 3700 and a IOS router using
    hardware forwardin such as an MSFC2 in a Cat 6k
    >
    > > The question is: if datagram size is large (so fragmentation is
    > > needed)and mutiple packets are used to carry the ping-data, how to
    > > calculate the response time?

    >
    > Roundtrip = Fragmentation + Transmit + Reassemby + Answer + the way back.

    --
    Regards

    Stephen Hope - remove xx from email to reply
    shope, Dec 20, 2003
    #7
    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. edward
    Replies:
    0
    Views:
    511
    edward
    Jun 29, 2004
  2. djoe

    Packet fragmentation

    djoe, Apr 25, 2006, in forum: Cisco
    Replies:
    3
    Views:
    520
    Arnold Nipper
    Apr 25, 2006
  3. djoe

    Packet fragmentation

    djoe, Jul 27, 2006, in forum: Cisco
    Replies:
    18
    Views:
    3,182
  4. lfnetworking
    Replies:
    3
    Views:
    4,863
    lfnetworking
    Aug 27, 2006
  5. mediumkuriboh
    Replies:
    0
    Views:
    1,476
    mediumkuriboh
    Feb 9, 2009
Loading...

Share This Page