08-07-2023 01:34 AM
Hi guys,
When would it be ideal to use ip unnumbered ?
1. Within DC environment ?
2. Within SP cloud ?
3. Within Enterprise Core ?
Regards
Solved! Go to Solution.
08-11-2023
08:52 AM
- last edited on
09-04-2023
05:49 AM
by
Translator
Jim, a very nice, if arcane (laugh) use case.
Also laugh, I didn't think my earlier replies said, or implied, that there's no use cases for unnumbered, anywhere, even within SPs. I do think some of the reasons for the feature, especially within the Enterprise, just aren't as common, or as "helpful", as they once were. I also didn't state, nor hopefully imply, that unnumbered might not even still be an ideal solution in some cases, again even for SPs.
In other words, I have no disagreements with your reply, or with what @Flavio Miranda or @marce1000 wrote.
Going back to OP, of the three environments, the most likely environment to use unnumbered, today, probably is within a SP. Would that be the "ideal" environment? Cannot really say, as "ideal", to me, implies the best possible. As even same "kind" of environments range over so many different issues, "ideal" to me, would depend on actual network goal, independent of the environment.
I'm unsure I'm making these distinctions clearly.
For example, I've worked in various Enterprises, some exclusively using Cisco equipment, and the question arises, which is "better" (or "ideal"?), EIGRP or OSPF?
Guess which Cisco will recommend. ; )
Other that Cisco having, perhaps, a bit of a bias in such a recommendation, technically, there is much to recommend EIGRP, especially depending on particular use situations within an Enterprise network.
Yet, everyone of those Enterprise environments chose to use OSPF (well except for one that decided to use IS-IS for IPv6, rather than OSPFv3 [which I was was very much against - a minority of 1 - but a year after I left, they moved from
IS-IS to OSPFv3
] - but that's another story). Why? Principally, to avoid specific network vendor lock in! To those "paying" for equipment, this is what made OSPF "ideal", yet often (Cisco only equipment) network engineers considered EIGRP "ideal".
So, because "ideal", like beauty, is in the eye of the beholder, and, to me, solutions/technologies are targeted to "issues", answering a question like OP's with just "#2", down the road, might lead to something like unnumbered not being considered within DCs or Enterprises, or conversely, we have to use unnumbered because we're an SP.
Now, I think it very unlikely, you Jim, or @Flavio Miranda or @marce1000 think like that, but inexperienced engineers, reading a topic's posting, in my experience, might not make such distinctions. I.e., well I "read" unnumbered are "ideal" for SP clouds.
I'm I making any sense?
BTW, this is why so many of my replies tend to run on, and on (and on - like this reply), because although likely boring to an experienced engineer (yea, we know that already), lots of inexperienced engineers, I suspect, read what we write too. I don't want them to infer something that sends them down the wrong path.
08-07-2023 01:41 AM
- FYI : https://networklessons.com/cisco/ccie-routing-switching/ip-unnumbered-explained
M.
08-07-2023
01:41 AM
- last edited on
09-04-2023
05:44 AM
by
Translator
Service Provider clould.
"To avoid problems with large numbers of
/30s
floating around the ISP's internal routing protocol, and to avoid the problems of keeping internal documentation consistent with network deployment (especially true in larger ISPs), many are now using unnumbered point-to-point links."
https://www.ccexpert.us/isp-essentials/ip-unnumbered.html
08-07-2023
04:42 AM
- last edited on
09-04-2023
05:45 AM
by
Translator
In current networks, using CIDR, where you can use
/30s
or possibly
/31s for p2p
links, likely the ideal use case for unnumbered is NOT to use it, especially for DC or Enterprise networks.
SP networks, due to size, look as being possible better use cases, but a well designed network, generally, isn't "flat", and uses equipment much more capable then when unnumbered was released.
Also, unsure whether this feature is compatible across vendors.
08-10-2023 11:15 PM
So basically for an SP it is not ideal either ? So why do Cisco IOS XR's code still support them if they were meant to be and older technology
08-11-2023 04:07 AM
Just because a feature isn't as useful on newer equipment means there's no new use cases, but perhaps more importantly, for backward compatibility.
08-11-2023
07:12 AM
- last edited on
09-04-2023
05:46 AM
by
Translator
Hi Joe,
I believe use of unnumbered is still the recommended practice for dynamic templates in an
ASR9000
BNG implementation for IPoE subscribers (admittedly, a rather arcane use-case). A 9K BNG might have many thousands of IPoE subscribers and using unnumbered saves an IP address on the BNG end of each dynamically created session (not to mention the multiple re-use of the same template, simplifying the config).
The 9K falls in line with your observation about "newer equipment" as I do not believe BNG has been ported to
NCS5500/5700
(NPU scaling issues), nor the Cisco 8000 (yet; but would not be surprised to see TR-459 disaggregated BNG appear on 8K in the future).
I do not believe BNG is going away, but may very well merge with 5G/6G fixed wireless access subscriber management in the future. Meaning that the use of IP unnumbered for this use-case could continue indefinitely.
08-11-2023
08:52 AM
- last edited on
09-04-2023
05:49 AM
by
Translator
Jim, a very nice, if arcane (laugh) use case.
Also laugh, I didn't think my earlier replies said, or implied, that there's no use cases for unnumbered, anywhere, even within SPs. I do think some of the reasons for the feature, especially within the Enterprise, just aren't as common, or as "helpful", as they once were. I also didn't state, nor hopefully imply, that unnumbered might not even still be an ideal solution in some cases, again even for SPs.
In other words, I have no disagreements with your reply, or with what @Flavio Miranda or @marce1000 wrote.
Going back to OP, of the three environments, the most likely environment to use unnumbered, today, probably is within a SP. Would that be the "ideal" environment? Cannot really say, as "ideal", to me, implies the best possible. As even same "kind" of environments range over so many different issues, "ideal" to me, would depend on actual network goal, independent of the environment.
I'm unsure I'm making these distinctions clearly.
For example, I've worked in various Enterprises, some exclusively using Cisco equipment, and the question arises, which is "better" (or "ideal"?), EIGRP or OSPF?
Guess which Cisco will recommend. ; )
Other that Cisco having, perhaps, a bit of a bias in such a recommendation, technically, there is much to recommend EIGRP, especially depending on particular use situations within an Enterprise network.
Yet, everyone of those Enterprise environments chose to use OSPF (well except for one that decided to use IS-IS for IPv6, rather than OSPFv3 [which I was was very much against - a minority of 1 - but a year after I left, they moved from
IS-IS to OSPFv3
] - but that's another story). Why? Principally, to avoid specific network vendor lock in! To those "paying" for equipment, this is what made OSPF "ideal", yet often (Cisco only equipment) network engineers considered EIGRP "ideal".
So, because "ideal", like beauty, is in the eye of the beholder, and, to me, solutions/technologies are targeted to "issues", answering a question like OP's with just "#2", down the road, might lead to something like unnumbered not being considered within DCs or Enterprises, or conversely, we have to use unnumbered because we're an SP.
Now, I think it very unlikely, you Jim, or @Flavio Miranda or @marce1000 think like that, but inexperienced engineers, reading a topic's posting, in my experience, might not make such distinctions. I.e., well I "read" unnumbered are "ideal" for SP clouds.
I'm I making any sense?
BTW, this is why so many of my replies tend to run on, and on (and on - like this reply), because although likely boring to an experienced engineer (yea, we know that already), lots of inexperienced engineers, I suspect, read what we write too. I don't want them to infer something that sends them down the wrong path.
08-17-2023 05:54 AM
Thanks so much @Joseph W. Doherty for taking the time to flesh this out especially with regards to the "ideal" mindset.
I very much believe that a solution/tech should be fixing issues not just slapped on for the sake of using the tech.
So basically, one of the "ideal (or let me say preferable ) " use-cases from what I gather, would be for an SP that a lot of transits in its core? as @Flavio Miranda mentioned.
Another option might be to just suppress the transits? However not including them in the first place alleviates the stress of having to design and manage transit IP's.
Please correct me if I am off.
08-17-2023 07:42 AM
"I very much believe that a solution/tech should be fixing issues not just slapped on for the sake of using the tech."
Laugh - another ideal. Many of us, including myself, do like to work with the latest and best technologies, but as professionals, we need to remember these technologies are not just our personal play things.
As to avoiding design and management stress, it might help, although anytime you start to do things, that are "unusual", creates its own stress because it is unusual.
I would say a technology like unnumbered is more oriented for usage stress, versus designing and/or managing stress.
Many of our networking techniques are for scalability Ever been asked the question, from someone learning about networks, why route? Why not have one huge L2 "flat" network? (A "flat" network certainly removes design stress and it's easy to manage, right? Unfortunately, what works well for 10 hosts, might not work so well for 10,000 hosts.)
So, something like unnumbered, can be very useful, but I suspect, most would consider it only when there's a scalability issue and/or, possibly, a security issue involved.
In the reference @Flavio Miranda provides, mentions:
However, this led to problems because IGPs of some of the larger ISPs were starting to carry several thousand networks, affecting convergence time and resulting in an administrative and documentation nightmare.
To avoid problems with large numbers of /30s floating around the ISP's internal routing protocol, and to avoid the problems of keeping internal documentation consistent with network deployment (especially true in larger ISPs), many are now using unnumbered point-to-point links.
(BTW, we might also use route aggregation, to avoid sloshing about thousands of routes, not just /30s, unnecessarily. [My initial reply's allusion to "a well designed network".])
Of the 3 choices, how might you rate them for possible scalability issues?
08-21-2023 03:31 PM
Basically design ur network well.
Do not use ip unnumbered.. Unless u absolutely have to?
08-21-2023 04:16 PM
"Do not use ip unnumbered.. Unless u absolutely have to?"
If implying as a last resort, no, I didn't mean such an implication.
Whatever technology you choose, should be what you believe is best for you. Determining your "best" can be quite difference from someone else's best. For example, we both chose to choose different colors? Is your choice of color better or worse then mine? Subjectively, no, objectively, yes, depending on criteria important to us. For example, some car colors show dust more than others. I chose my color to minimize maintenance. You chose your color because you enjoy washing and washing your car.
So, likewise, I'm not saying "absolutely" you must never have a red colored car. You like the color red. Fine, but objectively, you might want to recognize it "wears" worst subject to sunlight. If you know that, and want red, fine, if you didn't know that and now want a more durable color instead, that's fine too.
The forgoing making any sense?
08-30-2023 04:41 PM
Yes.. kinda still sounds to me like IP unnumbered is not generally a preferred (has some draw backs) solution, and it comes down to choice of engineer(s), unless I missed something.
Another place I have seen unnumbered used is as transit links for SPINE-LEAF EVPN/VXLAN underlays in DC environments. as the transit links do not matter much.. of course they all have to be p2p capable.
08-31-2023 10:28 AM
"kinda still sounds to me like IP unnumbered is not generally a preferred"
In many case, most likely that's true, because if you use a "unusual" approach, that's really unnecessary, it's a tad more of a maintenance headache because of general unfamiliarly. Of course, keep in mind what might be uncommon in one kind, might be the norm in another kind of network.
Simple example. . .
Last company I worked for had a fairly large (5,000 devices) Enterprise network, 100,000 users, and we also had our "Service Provider" network, also having about 5,000 devices (although their "entry" device was about equal to our "top tier" device).
For the IGP in the Enterprise, we ran OSPFv2. For the SP network's IGP, they ran, IS-IS (for both IPv4 and v6).
When we began to plan adding IPv6 to the Enterprise network, the question arose, use OSPFv3 or IS-IS. The decision was made to use IS-IS, in the Enterprise, for IPv6. I (pretty much a minority of one) was very much against using IS-IS, because, all the Enterprise engineers were familiar with OSPF, we still needed to maintain/use OSPFv2 for IPv4, and we ran into the issue that often "Enterprise" equipment didn't support IS-IS. Basically, I asked, what's the benefit of moving to IS-IS? The answer was, well our SP network engineers could help us and/or we could help them. To which I said it takes a quite a while to "learn" either network, as each are rather large.
Well, several years later, just after I retired, I heard they were, for the Enterprise, going to replace IS-IS with OSPFv3.
The moral of this story isn't is OSPF better/worse than IS-IS, but the "usual" IGP for Enterprise engineers is OSPF or possibly EIGRP and for SP engineers, IS-IS and BGP. (Of course BGP is used in Enterprises too, but not day-in, day-out, without all the complexity seen in a SP network.)
Again, nothing "bad" about unnumbered, but possibly part of whether to use it or not, if applicable for purpose, should consider the larger context too.
08-31-2023 07:53 PM
Thanks Joseph
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide