30 hop limit - TCP-IP

This is a discussion on 30 hop limit - TCP-IP ; I recently heard someone state that the Internet has a 30 hop limit. The only hop limitation I am aware of is TTL, which allows for a maximum of 256 hops. Anyone aware of something I am not?...

+ Reply to Thread
Results 1 to 16 of 16

Thread: 30 hop limit

  1. 30 hop limit

    I recently heard someone state that the Internet has a 30 hop limit. The
    only hop limitation I am aware of is TTL, which allows for a maximum of
    256 hops. Anyone aware of something I am not?

  2. Re: 30 hop limit

    On 2006-05-13, Dom wrote:

    > I recently heard someone state that the Internet has a 30 hop limit. The
    > only hop limitation I am aware of is TTL, which allows for a maximum of
    > 256 hops. Anyone aware of something I am not?


    It's not the Internet that has a 30-hop "limit". Most traceroute
    programs, which use a special technique (thank you, Van Jacobsen!) to
    elicit packets back from each router so that an Internet-attached
    host (yes, I am simplifying this for clarity) can look at the path
    it's taking through the network, have a limit of 30 hops they display
    before stopping.

    You are correct about TTL. TTL was originally thought up of as a
    way to limit packets from looping through the Internet over and over
    again in error due to faulty routing or another anomaly. My
    experience personally is that Internet routes that are greater in
    length than 20 or 30 hops are quite rare except for private
    enterprises using private internets.

    Hope this helps.

    /dmfh

    ----
    __| |_ __ / _| |_ ____ __
    dmfh @ / _` | ' \| _| ' \ _ / _\ \ /
    \__,_|_|_|_|_| |_||_| (_) \__/_\_\
    ----

  3. Re: 30 hop limit

    Dom wrote:

    > I recently heard someone state that the Internet has a 30 hop limit. The
    > only hop limitation I am aware of is TTL, which allows for a maximum of
    > 256 hops. Anyone aware of something I am not?


    Routing protocols would also impose a limit. For example, Routing
    Information Protocol (RIP) has a 15-hop limit. But larger scale routing
    protocols, like Open Shortest Path First (OSPF) are less restricted.
    OSPF provides a 24-bit field for route cost, and some links can have 0
    cost (but some can have cost greater than 1).

    So 30 hops may be a reasonably long path in practice, but I don't think
    there's any reason to think of it as a limit.

    Bert


  4. Re: 30 hop limit

    >> The only hop limitation I am aware of is TTL

    > Routing protocols would also impose a limit.


    In that instance, I think BGP would be the only consideration regarding
    the Internet at large. I am not aware of any hop limitations associated
    with BGP.

  5. Re: 30 hop limit

    "Dom" wrote:

    >> Routing protocols would also impose a limit.

    >
    > In that instance, I think BGP would be the only consideration
    > regarding the Internet at large. I am not aware of any hop limitations
    > associated with BGP.


    I don't agree that "BGP would be the only consideration." That's just
    one part of "the Internet."

    The Internet is composed of multiple autonomous systems, within which
    IGPs (interior garteway protocols) are used. If the IGPs in these ASs
    are unable to support enough hops to reach the other host on the other
    end of the route you're trying to establish, which may be on another AS,
    then clearly that would have a big impact on the usefulness of the
    Internet.

    Bert


  6. Re: 30 hop limit

    > I don't agree that "BGP would be the only consideration."

    Allow me to reiterate that I am not referring to connectivity within an
    AS. I am referring to end-to-end connectivity. Let us assume that
    transit ASes use static routing internally. It was stated that this 30
    hop limit was "by design". I fail to perceive any limitation that would
    impose a 30 hop limit.

  7. Re: 30 hop limit

    Vernon Schryver wrote:

    > What I understand the intent of that paragraph is wrong. Interior
    > gateway protocols generally know nothing about the path outside their
    > own AS. Whether
    > the other host on the other
    > end of the route you're trying to establish, which may be on another AS
    > is 1 or 1000 hops away is irrelevant to the IGP of the local AS.


    So, I presume by this you are saying that the question "is there a
    30-hop limit in The Internet" doesn't include the fact that "The
    Internet" is composed in large measure of ASs. And that hops between
    end hosts for some reason don't count as being part of any purported
    "hop limit" of The Internet?

    > And what's that about "establishing" a route? IP routes are not
    > "established," unless you're talking about a circuit kludge such as
    > MPLS or a VPN, which are not part of the IP universe.


    MPLS and VPNs are another matter. I can see why you dislike them, being
    as they are the illegitimate child of ATM.

    But how does OSPF work? Do OSPF routers not create minimal spanning
    tress with themselves as the root? I don't disagree that nothing
    prevents these trees from changing at the drop of a hat, but it's not
    like the trees need to change constantly either. In fact, they had
    better not. The path to all other routers in the AS and beyond the AS
    are established, *as far as OSPF routers are concerned*. If nothing
    changes in the network topology between two hosts (and there are no
    equal-cost paths), packets between these two hosts will take the same
    route. (Traffic can be divided among equal-path routes.)

    > When we talk about IP, we're talking about packet switching instead of
    > circuit switching. In IP there is no notion of establishing an end-to-end
    > path or anything like a circuit between hosts. There are only individual
    > hosts tossing packets on the ocean of routers and hoping that the
    > currents will be favorable. And that's a Good Thing(tm).


    That's from the hosts' point of view. From the routers' point of view,
    paths are calculated and retained unless something changes. Each router
    only sends packets to its neighbor, true, but the OSPF router does have
    a complete path calculated to the other host.

    Here's the quote from RFC 2328:

    2.2 ... The tree gives the entire path to any destination network or
    host. However, only the next hop to the destination is used in
    the forwarding process. Note also that the best route to any
    router has also been calculated. For the processing of
    external
    data, we note the next hop and distance to any router
    advertising external routes.
    ....

    2.3. Use of external routing information

    After the tree is created the external routing information is
    examined. This external routing information may originate from
    another routing protocol such as BGP, or be statically
    configured (static routes). Default routes can also be
    included
    as part of the Autonomous System's external routing
    information.
    ...

    The AS
    boundary router advertising the smallest external metric is
    chosen, regardless of the internal distance to the AS boundary
    router. Suppose in our example both Router RT5 and Router RT7
    were advertising Type 2 external routes. Then all traffic
    destined for Network N12 would be forwarded to Router RT7,
    since
    2 < 8. When several equal-cost Type 2 routes exist, the
    internal distance to the advertising routers is used to break
    the tie.

    As I said, the metric used in OSPF is assigned 24 bits, so this is
    enough for quite a number of "hops," even though OSPF doesn't use
    "hops" per se, but "cost."

    Bert


  8. Re: 30 hop limit

    In article ,
    "Albert Manfredi" wrote:

    > "Dom" wrote:
    >
    > >> Routing protocols would also impose a limit.

    > >
    > > In that instance, I think BGP would be the only consideration
    > > regarding the Internet at large. I am not aware of any hop limitations
    > > associated with BGP.

    >
    > I don't agree that "BGP would be the only consideration." That's just
    > one part of "the Internet."
    >
    > The Internet is composed of multiple autonomous systems, within which
    > IGPs (interior garteway protocols) are used. If the IGPs in these ASs
    > are unable to support enough hops to reach the other host on the other
    > end of the route you're trying to establish, which may be on another AS,
    > then clearly that would have a big impact on the usefulness of the
    > Internet.


    The IGP only has to be able to handle the number of hops within that AS.

    So the design of your AS should inform your choice of IGP, so that it
    can handle its diameter. But you don't have to worry about the total
    number of hops to the destination.

    --
    Barry Margolin, barmar@alum.mit.edu
    Arlington, MA
    *** PLEASE post questions in newsgroups, not directly to me ***
    *** PLEASE don't copy me on replies, I'll read them in the group ***

  9. Re: 30 hop limit

    In article <1147661215.294322.160550@j73g2000cwa.googlegroups. com>,
    Albert Manfredi wrote:

    >So, I presume by this you are saying that the question "is there a
    >30-hop limit in The Internet" doesn't include the fact that "The
    >Internet" is composed in large measure of ASs. And that hops between
    >end hosts for some reason don't count as being part of any purported
    >"hop limit" of The Internet?


    No, I'm not saying that.

    I suspect the other person had the right answer, that the supposed
    30-hop limit is the default probe limit in `traceroute`.
    Failing that, it's IP TTL limit. I seem to recall something about
    increasing the default TTL in BSD derived IP code from something like
    30 a bunch of years ago.


    >MPLS and VPNs are another matter. I can see why you dislike them, being
    >as they are the illegitimate child of ATM.


    I don't dislike them. They're just not IP. Packet switching has major
    advantages in many situations, but it's not a religion. Circuits also
    have their places.

    Never mind that VPNs are not children of ATM, since they were around
    before anyone talked about using ATM for data and because they included
    nothing that looks remotely like an ATM cell.


    >But how does OSPF work? Do OSPF routers not create minimal spanning
    >tress with themselves as the root? I don't disagree that nothing
    >prevents these trees from changing at the drop of a hat, but it's not
    >like the trees need to change constantly either. In fact, they had
    >better not.


    Only because the computation of the tree takes a lot of cycles.


    >> When we talk about IP, we're talking about packet switching instead of
    >> circuit switching. In IP there is no notion of establishing an end-to-end
    >> path or anything like a circuit between hosts. There are only individual
    >> hosts tossing packets on the ocean of routers and hoping that the
    >> currents will be favorable. And that's a Good Thing(tm).

    >
    >That's from the hosts' point of view. From the routers' point of view,
    >paths are calculated and retained unless something changes. Each router
    >only sends packets to its neighbor, true, but the OSPF router does have
    >a complete path calculated to the other host.


    The OSPF router has a model of the network, but it's only a model
    subject to change. It exists only to make computing the next hop faster.


    >Here's the quote from RFC 2328:
    >
    >2.2 ... The tree gives the entire path to any destination network or
    > host. However, only the next hop to the destination is used in


    Are you are misinterpreting that RFC? RFCs do not specify how things
    actually work inside routers, hosts, or anything else. They only specify
    how things must appear to work from outside the black box. A router
    using OSPF must act as if it has previously computed the tree when it
    forwards a packet, but it could recompute the tree based on saved link
    state advertisements every time it receives a packet to forward. In
    fact, sometimes it must recompute the tree.

    That fact breaks your circuit switching model of IP routing. Between
    the time a packet leaves the previous hop and arrives at the next
    router, your "established path" can change completely. A link state
    advertisement can arrive at the next router just before the packet,
    causing the router to decide to send the packet on an entirely new
    path. Routing protocols and implementations have "hold downs,"
    "anti-flap," and other things to minimize the chances of a packet being
    forwarded in circles, but they are not and cannot be perfect. As I
    said before, the IP header has a TTL field while ATM doesn't because
    IP involves packet instead of circuit switching.


    Vernon Schryver vjs@rhyolite.com

  10. Re: 30 hop limit

    On 2006-05-15, Vernon Schryver wrote:

    > In article <1147661215.294322.160550@j73g2000cwa.googlegroups. com>,


    > Albert Manfredi wrote:


    >>But how does OSPF work? Do OSPF routers not create minimal spanning
    >>tress with themselves as the root? I don't disagree that nothing
    >>prevents these trees from changing at the drop of a hat, but it's not
    >>like the trees need to change constantly either. In fact, they had
    >>better not.

    >
    > Only because the computation of the tree takes a lot of cycles.


    This dialog has brought up a point in my mind I wanted to post about.
    Besides RFC's being a reference and not a "standard", there is a lot
    that goes on in the "black box" that's magic and from my experience
    with hardware / software / routing design over the years, the
    customer experience seems to speak to the vendor finding a lot of
    tricks and shortcuts in either hardware or software that lead to a
    race by the vendor to "Conform to the RFC", or exert enough influence
    on a standards committee to get The Next Great Thing into the
    standards cycle.

    OSPF calculation *does* take a lot of cycles for what could loosely
    be called a "full re-convergence" - I don't use the OSPF terminology
    for it because I feel it fails to take into account the trigger
    conditions that can cause this outside of the OSPF standard. Many
    vendors minimize this by either trying to eliminate the usual causes
    of a "full re-convergence" (other than uncontrollable-s like link
    failures) and cache structures that optimize hardware flow
    programming in an attempt to determine "does this change really
    affect me?".

    Personally, I like designs with the fewest "corner case" conditions
    that can be found and just overall simplicity. Opaque LSA's, and
    Virtual Links in some OSPF implementations, IMHO, are examples of how
    expanding out from RFC tree can create problems.

    /dmfh

    ----
    __| |_ __ / _| |_ ____ __
    dmfh @ / _` | ' \| _| ' \ _ / _\ \ /
    \__,_|_|_|_|_| |_||_| (_) \__/_\_\
    ----

  11. Re: 30 hop limit

    "Vernon Schryver" wrote:

    > A router
    > using OSPF must act as if it has previously computed the tree when it
    > forwards a packet, but it could recompute the tree based on saved link
    > state advertisements every time it receives a packet to forward. In
    > fact, sometimes it must recompute the tree.
    >
    > That fact breaks your circuit switching model of IP routing.


    No, it doesn't, because I have no circuit switching model.

    (I'm sure you already know all of this ...)

    IP routing with OSPF separates the sending of packets by hosts from the
    setting up of routes through an internet. The two are simply not done
    together. The hosts toss out packets whenever they feel like it, and the
    routing function will be ready to do something as these packets arrive,
    even if this new packet is sent on a different route compared with the
    previous one.

    Instead in ATM (I'm speaking native ATM, not IP over ATM schemes like
    LANE), the data cells aren't transmitted until the route has been set up
    end to end. Connection oriented at the link layer, from IP's
    perspective.

    That's why the cells don;t need a TTL. Because the route they take has
    already been calculated and is expected to remain stable for the
    duration of that session.

    However, leaving that aside, and leaving aside the "constraint-based
    routing" that PNNI is supposed to offer, how OSPF sets up routes is
    almost identical to how PNNI does.

    > Between
    > the time a packet leaves the previous hop and arrives at the next
    > router, your "established path" can change completely. A link state
    > advertisement can arrive at the next router just before the packet,
    > causing the router to decide to send the packet on an entirely new
    > path. Routing protocols and implementations have "hold downs,"
    > "anti-flap," and other things to minimize the chances of a packet
    > being
    > forwarded in circles, but they are not and cannot be perfect. As I
    > said before, the IP header has a TTL field while ATM doesn't because
    > IP involves packet instead of circuit switching.


    So we are saying the same thing. Routing in IP can change at the drop of
    a hat, but that's a far cry from any implication that the routes are
    random. They are calculated and they are stable, unless something in the
    network causes them to change.

    Also strayed off topic, but at least it's an interesting topic.

    Bert


  12. Re: 30 hop limit

    "Barry Margolin" wrote:

    > The IGP only has to be able to handle the number of hops within that
    > AS.
    >
    > So the design of your AS should inform your choice of IGP, so that it
    > can handle its diameter. But you don't have to worry about the total
    > number of hops to the destination.


    Well, that's not entirely true, right? OSPF does in fact worry about the
    total *cost* to the destination (not hops), even though the cost beyond
    the local AS of the sending host might be somewhat made up. Make-believe
    or no, it's still a consideration of OSPF.

    But again, with 24 bits assigned to the total end to end cost, that
    should allow for well over any 30 hops, under normal circumstances.

    Bert


  13. Re: 30 hop limit

    "Vernon Schryver" wrote:

    > Never mind that VPNs are not children of ATM, since they were around
    > before anyone talked about using ATM for data and because they
    > included
    > nothing that looks remotely like an ATM cell.


    VPNs implemented over MPLS is what I was referring to. MPLS's labels and
    stable routes are certainly something that grew out of ATM cell VPI/VCI.
    The concept is the same, even if MPLS takes it further by allowing the
    labels to be stacked.

    The fundamental idea is, first you set up the end to end path, then you
    route packets through this path using much less header information that
    it takes to route the packet the old-fashioned, path-independent way.

    ATM was devceloped by Bellcore ca. 1989. I think that predates MPLS by
    quite a few years, though less than 10.

    Bert


  14. Re: 30 hop limit

    "Albert Manfredi" writes:
    > So we are saying the same thing. Routing in IP can change at the drop
    > of a hat, but that's a far cry from any implication that the routes
    > are random. They are calculated and they are stable, unless something
    > in the network causes them to change.
    >
    > Also strayed off topic, but at least it's an interesting topic.


    when we were working on the original payment gateway
    http://www.garlic.com/~lynn/aadsm5.htm#asrn2
    http://www.garlic.com/~lynn/aadsm5.htm#asrn3

    we were looking at doing high availability configuration,
    in part because we had earlier done the ha/cmp product
    http://www.garlic.com/~lynn/subtopic.html#hacmp

    for some trivia drift ... some of the people in this
    ha/cmp meeting
    http://www.garlic.com/~lynn/95.html#13

    later showed up at a small client/server startup responsible for
    turning out a payment gateway. this small client/server startup also
    had this technology called SSL.

    in any case during the process of initial deployment ... we were
    looking at advertising routes for ip-addresses as countermeasure to
    various kinds of route outages. however, in that period, the internet
    backbone decided to move to hierachical routing. we had assumed that
    we could get diverse physical routing to some number of different,
    carefully selected ISPs. If any of the ISPs were having connectivity
    problems (like taking down critical components on sunday afternoon
    prime-time for maintenance) we could advertise routes via other paths.

    the transition to hierarchical routing eliminated attention being paid
    to alternate advertise routes. that primarily left multiple a-records
    as an alternate path mechanism (i.e. same domain/host name mapping to
    a list of different ip-addresses).

    --
    Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

  15. Re: 30 hop limit

    Please, tell us more about your renowned work with IP over Carrier Pigeon.

  16. Re: 30 hop limit

    In article ,
    "Albert Manfredi" wrote:

    > "Barry Margolin" wrote:
    >
    > > The IGP only has to be able to handle the number of hops within that
    > > AS.
    > >
    > > So the design of your AS should inform your choice of IGP, so that it
    > > can handle its diameter. But you don't have to worry about the total
    > > number of hops to the destination.

    >
    > Well, that's not entirely true, right? OSPF does in fact worry about the
    > total *cost* to the destination (not hops), even though the cost beyond
    > the local AS of the sending host might be somewhat made up. Make-believe
    > or no, it's still a consideration of OSPF.


    But since that cost outside the AS isn't generally a number of hops
    (it's more likely to be something just based on the speed of the
    interconnection link to the peer AS), my comment stands.

    --
    Barry Margolin, barmar@alum.mit.edu
    Arlington, MA
    *** PLEASE post questions in newsgroups, not directly to me ***
    *** PLEASE don't copy me on replies, I'll read them in the group ***

+ Reply to Thread