I'm just curious - what is the largest OSPF core (in terms of number of routers) out there?
Subject: Re: largest OSPF core
On 02/09/2010 13:20, lorddoskias wrote:
I'm just curious - what is the largest OSPF core (in terms of number of routers) out there?
You don't expect anyone to actually admit to something like this? :-)
For giggles: http://books.google.com/books?id=uBwEAAAAMBAJ&pg=PA59&dq=practical+limits+of+OSPF&hl=en&ei=qud_TNTAFYL68AautJXoAQ&sa=X&oi=book_result&ct=result&resnum=2&ved=0CCwQ6AEwAQ#v=onepage&q=practical%20limit&f=false Network World April 9, 1990 (page 59): "There is no practical limit to the number of interconnected networks OSPF and Dual Intermediate System-to-Intermediate System can support"... "From the onset, OSPF was intended to be short-term, for IP-only".. "Dual routing is intended to be more of a long-term solution because there will be very few pure OSI or TCP/IP routing environments in the future." --- Technology prognosticators shouldn't try their hands in Vegas. Just saying. With respect to these OSPF questions, how many people are running two OSPF processes on each router (v4 and v6) to support dual stack rather than migrating (or just enjoying their existing) ISIS (OSI) implementations? Deepak
Sent from my iPad On Sep 3, 2010, at 3:42 AM, Deepak Jain <deepak@ai.net> wrote:
Subject: Re: largest OSPF core
On 02/09/2010 13:20, lorddoskias wrote:
I'm just curious - what is the largest OSPF core (in terms of number of routers) out there?
You don't expect anyone to actually admit to something like this? :-)
For giggles:
Network World April 9, 1990 (page 59):
"There is no practical limit to the number of interconnected networks OSPF and Dual Intermediate System-to-Intermediate System can support"...
"From the onset, OSPF was intended to be short-term, for IP-only"..
"Dual routing is intended to be more of a long-term solution because there will be very few pure OSI or TCP/IP routing environments in the future."
---
Technology prognosticators shouldn't try their hands in Vegas. Just saying.
With respect to these OSPF questions, how many people are running two OSPF processes on each router (v4 and v6) to support dual stack rather than migrating (or just enjoying their existing) ISIS (OSI) implementations?
You left out the option of using ospf3 to do both v4 and v6. Works on juniper and foundry at least. Owen
Deepak
.
With respect to these OSPF questions, how many people are running two
OSPF processes on each router (v4 and v6) to support dual stack rather than migrating (or just enjoying their existing) ISIS (OSI) implementations?
You left out the option of using ospf3 to do both v4 and v6. Works on juniper and foundry at least.
Owen
http://www.cisco.com/web/strategy/docs/gov/OSPFv3_aag.pdf Thank you. Apparently Cisco supports it (or something like it) too. Deepak
On Thu, Sep 02, 2010 at 05:32:30PM -0400, Deepak Jain wrote:
With respect to these OSPF questions, how many people are running two OSPF processes on each router (v4 and v6) to support dual stack rather than migrating (or just enjoying their existing) ISIS (OSI) implementations?
You left out the option of using ospf3 to do both v4 and v6. Works on juniper and foundry at least.
Owen
http://www.cisco.com/web/strategy/docs/gov/OSPFv3_aag.pdf
Thank you. Apparently Cisco supports it (or something like it) too.
Seems silly to migrate your existing OSPFv2 to an extra instance of OSPFv3, leaving 2 separate OSPFv3 instances. Why not just stick with your existing OSPFv2 and add OSPFv3 for IPv6? Or if you want to migrate your IPv4 IGP, go directly to IS-IS so you can have a single link-state database, single process, etc. for both IPv4 and IPv6.
On Sep 2, 2010, at 11:11 AM, Nick Hilliard wrote:
On 02/09/2010 13:20, lorddoskias wrote:
I'm just curious - what is the largest OSPF core (in terms of number of routers) out there?
You don't expect anyone to actually admit to something like this? :-)
Of course I do -- 'tis much for your reputation to have wrangled a poorly designed, ugly network under control than to have only worked at places with smooth sailing.... I *don't* expect the owner / designers of these to come forward, rather those who inherited a pile of choss to share war stories... :-P I worked on a network that had >350 routers in an (non-zero) area. Now, ~350 routers in an area doesn't sound *that* impressive, but on average these devices had 6 interfaces in OSPF, and many of these links were of the form: [router A]-- {GRE} --- [firewall]-- {GRE in IPSEC} ------- [Internet]------- {GRE in IPSEC} ---[firewall]---{GRE} --- [router B] Routers A and B would form an OSPF adjacency. Much of this was an overlay network (over the Internet) and so the firewalls would build IPSec tunnels. Of course, said firewalls would not pass OSPF, so we had to build GRE tunnels between routers A and D and run OSPF over those -- traffic would enter the router, get encapsulated in GRE and then the GRE would be encapsulated in IPSec and tossed into the void.... In other places (in the same OSPF area) we would purchase parallel T1 / E1s that we would run MLPPP over, and / or plain DS3s. Oh, did I mention that network was primarily to support international call centers that had been outsourced to wherever was *really* cheap, and that many places with very cheap labor have very poor infrastructure? It was not uncommon to have interfaces that would bounce 5 or 10 times a day*.... W *: And yes, we did 'ave to get up out of shoebox at twelve o'clock at night and lick road clean wit' tongue. We had two bits of cold gravel, worked twenty-four hours a day at mill for sixpence every four years, and when we got home our Dad would slice us in two wit' bread knife.
Nick
-- It's a mistake trying to cheer up camels. You might as well drop meringues into a black hole. -- Terry Prachett
In a message written on Thu, Sep 02, 2010 at 03:20:05PM +0300, lorddoskias wrote:
I'm just curious - what is the largest OSPF core (in terms of number of routers) out there?
I'll admit to having seen a network with over 400 devices in an OSPF area 0, didn't design it, and in the end didn't get to work on it. Far as I know worked just fine though, no issues reported. How well your IGP scales depends a lot more on what you put in it, and how dynamic your network situation is than the protocol or number of devices. -- Leo Bicknell - bicknell@ufp.org - CCIE 3440 PGP keys at http://www.ufp.org/~bicknell/
I think it is really depending on how your network topology looks like. If you have top-down design with star topology to limit the network connections to individual routers, it may scale well. But if you connect every routers to each other such as full-mesh, it will be a problem during interface flapping or something like that. Alex On Thu, Sep 2, 2010 at 1:37 PM, Leo Bicknell <bicknell@ufp.org> wrote:
In a message written on Thu, Sep 02, 2010 at 03:20:05PM +0300, lorddoskias wrote:
I'm just curious - what is the largest OSPF core (in terms of number of routers) out there?
I'll admit to having seen a network with over 400 devices in an OSPF area 0, didn't design it, and in the end didn't get to work on it.
Far as I know worked just fine though, no issues reported. How well your IGP scales depends a lot more on what you put in it, and how dynamic your network situation is than the protocol or number of devices.
-- Leo Bicknell - bicknell@ufp.org - CCIE 3440 PGP keys at http://www.ufp.org/~bicknell/
In a message written on Thu, Sep 02, 2010 at 03:20:05PM +0300, lorddoskias wrote:
I'm just curious - what is the largest OSPF core (in terms of number of routers) out there?
The stability of the topology plays a most prominent role, but it wouldn't surprise me if a OSPF network largely comprised of router LSAs (no redistribution), using today's hardware, could easily scale to 1000 nodes in an area. Some newer techniques may affect this scale in either direction (ie, subsecond hellos and fast convergence would negatively impact scale on some platforms, while using demand circuit emulation on p2p links would impact scale positively). YMMV... C
The stability of the topology plays a most prominent role, but it wouldn't surprise me if a OSPF network largely comprised of router LSAs (no redistribution), using today's hardware, could easily scale to 1000 nodes in an area.
i believe the original poster asked about actual operating deployment, not theory. and, i suspect one wants to know about full mesh under real load, i.e. topology change, which can be exciting when one gets to a network of significant size. randy
On Sep 2, 2010, at 7:35 PM, Randy Bush <randy@psg.com> wrote:
The stability of the topology plays a most prominent role, but it wouldn't surprise me if a OSPF network largely comprised of router LSAs (no redistribution), using today's hardware, could easily scale to 1000 nodes in an area.
i believe the original poster asked about actual operating deployment, not theory.
and, i suspect one wants to know about full mesh under real load, i.e. topology change, which can be exciting when one gets to a network of significant size.
Randy, Fair enough. 7 years ago, I was privy to an OSPF BB of 300 or so routers supporting a BGP overlay. No NBMA, passive broadcast subnetworks, all running on systems without the capacity to offload adjacency maintenance into linecards. I'd argue that this type of network is also uninteresting from a NANOG viewership POV. I also operated a network that supported over 70 OSPF VRF instances on a single PE. CPU loads were higher, but we didn't observe intractable workloads. And this was with a 500 route limit per VRF, with who knows what kinds of messiness running in those VRFs. (and yea there were sham links and router LSAs flying around!!) :) There are many variables, and several studies have tried to capture, algorithmically and in terms of computational complexity, a formulaic approach to determining the boundaries of OSPF network scalability. Admittedly, these approaches can be very approximate in nature. But the point stands. Stable topologies absent large, frequent, compulsively updated data can scale extremely well. Unstable topologies with lots of leaf data (20,000 type 5 LSAs, for example), don't. The most interesting point to make, however, is how much legacy thinking in this area continues to be stranded in a rut that emerged 15 years ago. It is not uncommon to hear network folks cringe at the thought of an OSPF area exceeding 100 routers. Really? When simulations using testing tools show that properly tuned OSPF implementations (with ISPF, PRC, etc) comprised of 1000 can run full SPFs in 500 ms? That said, my experience, as stated above, is that 300 routers is completely workable. Cheers Chris
randy
In a message written on Thu, Sep 02, 2010 at 09:40:39PM -0400, Christian Martin wrote:
The most interesting point to make, however, is how much legacy thinking in this area continues to be stranded in a rut that emerged 15 years ago. It is not uncommon to hear network folks cringe at the thought of an OSPF area exceeding 100 routers. Really? When simulations using testing tools show that properly tuned OSPF implementations (with ISPF, PRC, etc) comprised of 1000 can run full SPFs in 500 ms?
I do think a lot of the thinking is out of date. I strongly agree that all the references I know of about scaling are based on the CPU and RAM limitations of devices made in the 1990's. Heck, a "branch" router today probably has more CPU than a backbone device of that era. The larger issue though is that as an industry we are imprecise. If you talk about when a routing protocol /fails/, that is can't process the updates with the available CPU before the session times out, you're probably talking a network of 250,000 routers on a low end device. Seriously, how large does a network need to be to keep OSPF or ISIS CPU-busy for 10-20 seconds? Huge! Rather, we have scaling based on vauge, often unstated rules. One vendor publishes a white paper based on devices running only the IGP and a stated convergence time of 500ms. Another will assume the IGP gets no more than 50% of the CPU, and must converge in 100ms. Also, how many people have millisecond converging IGP's, but routers with old CPU's so BGP takes 3-5 MINUTES to converge? Yes, for some people that's good, if you have lots of internal VOIP or other things; but if 99% of your traffic is off net it really doesn't matter, you're waiting on BGP. Lastly, the largest myth I see in IGP design is that you can't redistribute connected or statics into your IGP, those go into BGP so the IGP only has to deal with loopbacks. As far as I understand the computational complexity of OSPF and IS-IS depends solely on the number of links running the protocol, so having these things in or out makes no difference in that sense. It does increase the amount of data that needs to be transferred by the IGP's which does slow them a bit, but with modern link speeds and CPU's it's really a non-issue. I'm not saying it's "smart" to redistribute connected and static, it really does depend on your environment. However there seems to be a lot of folks who automatically assume the network is broken if it has such things in the IGP, and that's just silly. Plenty of networks have that data in the IGP and deliver excellent routing performance. Fortunately we've gotten to the point where 95% of the networks don't have to worry about these things, it works however you want to do it. However for the 5% who need to care, almost none of them have engineers who actually understand the programming behind the protocols. How many network architects could write an OSPF implementation or understand your boxes architecture? -- Leo Bicknell - bicknell@ufp.org - CCIE 3440 PGP keys at http://www.ufp.org/~bicknell/
On Thu, Sep 2, 2010 at 2:37 PM, Leo Bicknell <bicknell@ufp.org> wrote:
In a message written on Thu, Sep 02, 2010 at 03:20:05PM +0300, lorddoskias wrote:
I'm just curious - what is the largest OSPF core (in terms of number of routers) out there?
I'll admit to having seen a network with over 400 devices in an OSPF area 0, didn't design it, and in the end didn't get to work on it.
I know of a large enterprise with ~4k devices in area-0, according to their vendor^H^H^H^H^Hdesigner that was all perfectly fine.
Far as I know worked just fine though, no issues reported. How well your IGP scales depends a lot more on what you put in it, and how dynamic your network situation is than the protocol or number of devices.
I think the only reason the one I saw worked at all was it was relatively stable. If things happened though (like say the code-red incident in ... whenever that was) the network turned into a steaming pile of fail. really, not a good plan, of course as Leo says ISIS probably gets super unhappy if a large percent of interfaces start to go bouncey-bouncey. -Chris
On Thu, 02 Sep 2010 15:20:05 +0300 lorddoskias <lorddoskias@gmail.com> wrote:
I'm just curious - what is the largest OSPF core (in terms of number of routers) out there?
Presuming OSPF and IS-IS SPF costs are fairly similar, the following page from "The complete IS-IS routing protocol" (really quite a good book, a bit of a shame that there are occasional minor errors that better technical editing would have picked up) shows that relatively modern (although a number of years old now) routers can perform SPF calcs on SPF databases with 10000 routers and 25000 links in less than a second. From that, it would seem that areas / levels are obsolete for most networks for the purposes of reducing SPF calculation time. Still possibly useful for route aggregation, although if BGP is carrying nearly all your routes, that may not be that useful either. http://books.google.com.au/books?id=NxIadsCKZxMC&lpg=PP1&dq="IS-IS"&pg=PA481#v=onepage&q&f=false
On 2 Sep 2010, at 8:20 PM, lorddoskias wrote:
I'm just curious - what is the largest OSPF core (in terms of number of routers) out there?
I have seen (as a consultant, not operator) a production SP network that had over 800 routers in the backbone area. The LSDB was rather small as the network only carried links and loopbacks for P/PE routers. All other prefixes were in MP-BGP. Truman
participants (14)
-
Alex Ryu
-
Christian Martin
-
Christopher Morrow
-
Chuck Anderson
-
Deepak Jain
-
Leo Bicknell
-
lorddoskias
-
Mark Smith
-
Nick Hilliard
-
Owen DeLong
-
Randy Bush
-
Truman Boyes
-
Valdis.Kletnieks@vt.edu
-
Warren Kumari