cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
3268
Views
65
Helpful
30
Replies

Multilink PPP, IP CEF or Both

nxm
Level 1
Level 1

I will be installing two 2811 routers with four dedicated (point-to-point) T1 connections.

Is it better (faster) to use Multilink PPP, IP CEF, or can I use a combination of both?

30 Replies 30

Peter,

     please keep in mind that this discussion is not about the virtues of CEF. I do see however that you really enjoy discussing the virtues of CEF, so I suggest that you start a new discussion based on the virtues of CEF. If you will you refer to the original post, which I have been trying to guide you to do for some time now, you will see that the original poster did ask IF HE SHOULD USE CEF OR MLPPP TO AGGREGATE 4 T1 LINES!

     So what the heck do you mean when you say,

In doing this, you are trying to compare and contrast two totally different things.

????????

     You should try to realize that in the context of this discussion, YOU CAN USE EITHER CEF OR MLPPP to accomplish what the original poster is trying to accomplish. If you are unaware of this fact, then why are you muddling this discussion with your ramblings? Please keep your postings in the context of the discussion, and quit being argumentative and try to understand that other people may have the same question, and they will refer to this thread for an answer.

Best regards

Hello Chad,

please keep in mind that this discussion is not about the virtues of CEF

True, this discussion is not about the virtues of CEF, but it touches CEF. And once the CEF is mentioned, it has to be discussed in exact terms, and if dubious information about it is posted in the process, it has to be corrected. This goes for every topic.

If you will you refer to the original post...

If you refer to the original post you will see that it has been posted on February 23rd 2006. Your answer comes 7 years late. I am surprised you are defending the relevancy and "on-topic-ness" of your answers so ardently considering they will most probably pass unnoticed after such a time, as the OP must have had his issue resolved sooner. Your posts not only came late, they also contained or suggested incorrect understanding of some aspects of CEF. If I, in my answers, provided an additional context to how CEF works (which you called an off-topic content), it was intentional to properly explain why I considered some of the facts in your answers as being questionable. I could as well just say "this is wrong, that is wrong" without explaining why it is wrong but that would help nobody.

... which I have been trying to  guide you to do for some time now, you will see that the original poster  did ask IF HE SHOULD USE CEF OR MLPPP TO AGGREGATE 4 T1 LINES! 

I am not arguing that. But you posted answers that contained questionable or misleading statements about a particular technology and I reacted to those. If such imprecisions occur in answers, even though they tackle an off-topic theme, I am and will be pointing them out and setting them straight. An imprecision, or an outright incorrect statement in a post can not be just summarily dismissed with a convenient comment that "it is off topic, anyway". Participants reading these answers absorb them in their entirety so our mutual striving should be to post entirely correct answers. Many imprecisions or misunderstandings occur from saying tiny incorrect things here and there, and they tend to accumulate. Trust me on this, I work in education and I am acutely aware of the damage caused by these "slight, off-topic imperfections".

Chad, and one more thing - in my over 5,500 posts on CSC, I have never allowed myself to call anyone argumentative, trite or rambling (you may check using the Search function). Having such strong judgements about the people you debate with, without knowing anything about them and the knowledge/experience they speak from, is utterly presumptuous and undignified. Considering you are a freshman here on CSC and have yet to build up your reputation, you behave in an overbearing, imposing way, not really trying to seek a mutual understanding and a common ground with other people but at most, trying to force your own views upon them, and when it fails, dismiss them offendingly. The only reputation you are building so far is a negative one. Please, try to change your attitude.

Best regards,

Peter

Disclaimer

The  Author of this posting offers the information contained within this  posting without consideration and with the reader's understanding that  there's no implied or expressed suitability or fitness for any purpose.  Information provided is for informational purposes only and should not  be construed as rendering professional advice of any kind. Usage of this  posting's information is solely at reader's own risk.

Liability Disclaimer

In  no event shall Author be liable for any damages whatsoever (including,  without limitation, damages for loss of use, data or profit) arising out  of the use or inability to use the posting's information even if Author  has been advised of the possibility of such damage.

Posting

chad patterson wrote:

     Actually CEF has absolutely nothing to do with MLPPP multilink. CEF is enabled by default and does 'process switching', and should only be turned off if the the interface has a feature enabled that cannot support CEF, which be unlikely in most cases.

I'm sorry, where did I write CEF does have something to do with MLPPP?  I contrasted how the two can load share across multiple links, but didn't directly relate them.  Although I didn't touch upon their relationship, there is one, as CEF can use a multilink interface as its egress interface.

It's true CEF is generally the default in current IOS versions, but I believe I recall when the feature was first introduced (12.0?), it had to be explicitly enabled.

Cisco doesn't consider CEF "process switching", they consider it "fast path" or "interrupt switching".

     CEF is just a routing trouble, and stores information for hops, i.e.., the best next hop in a route to reach a destination on a given network. If you are connected to the ISP, then the ISP gateway IS YOUR ONLY NEXT HOP, and makes the routing table stored in CEF useless because every singly entry in the table will be "ISP gateway". This assumes you have a simple private local subnet consisting of a mere /24 size. If your private has more than subnet, then CEF will assist in routing to destinations on the local subnet.

As Peter has already touched upon CEF, I refer readers to those posts.

     Lastly a link of  MLPPP link of 4 T1's does make that link behave EXACTLY like a single 6Mbps link. It routes packet fragments all up and down those 4 T1 lines as if it were a single line.

If you mean MLPPP will logically behave like a single link, you're correct.  If you also mean link will perform exactly as a physical link, no it will not.  Simple example, send one packet, w/o MLPPP fragmentation and it will take over 4x longer to transmit.  This because the single packet is serialized across a 1.5 Mbps physical link rather than a 6 Mbps physical link.  If if there are exactly 4 equal size packets (or 4 MLPPP fragments), MLPPP introduces, I believe, a little more protocol overhead on the links and adds some additional processing latency.  The 6 Mbps link will deliver packets just a little sooner and, if there's less protocol overhead, at a very slightly higher transfer rate.

If you think such minor performance differences are insignificant, often they are but sometimes they are not.

     OK. My bad. My apologies to all. I have reread this conversation and have come to a new understanding of our miscommunications. I had assumed we were all on the same page, which I am prone to do sometimes. But it is my fault that we were not, because I did not articulate clearly.

     Whenever I mentioned CEF, in the context of using it with 4 T1 lines, I assumed it was understood that I was talking about 'CEF load sharing'. That is the only option involving CEF in this scenario. Yes I do agree that CEF is a good thing and it is better than 'naive proccess switching' (good term by the way). But I assumed that we all understood that 'naive process switching' is not even an option when it comes to load balancing 4 T1 lines.

     So I tried to guide the conversation towards CEF vs MLPP because all of my CEF statements were made in that context, and not in the context of CEF vs NAIVE PROCESS SWITCHING. So I issue a friendly challenge to all who were arguing with me: I will give you all the kudos if you can post a configuration, using only CEF and not MLPPP, that load balances 4 T1's. Never mind whether or not it ends up being inferior to MLPPP, I just want to see if any of you can even do it successfully.

Disclaimer

The  Author of this posting offers the information contained within this  posting without consideration and with the reader's understanding that  there's no implied or expressed suitability or fitness for any purpose.  Information provided is for informational purposes only and should not  be construed as rendering professional advice of any kind. Usage of this  posting's information is solely at reader's own risk.

Liability Disclaimer

In  no event shall Author be liable for any damages whatsoever (including,  without limitation, damages for loss of use, data or profit) arising out  of the use or inability to use the posting's information even if Author  has been advised of the possibility of such damage.

Posting

!router 1

conf t

in s0/0/0

ip add 192.168.0.1 255.255.255.252

no shut

exit

in s0/1/0

ip add 192.168.1.1 255.255.255.252

no shut

exit

in s0/2/0

ip add 192.168.2.1 255.255.255.252

no shut

exit

in s0/3/0

ip add 192.168.3.1 255.255.255.252

no shut

exit

router os 1

network 192.168.0.0 0.0.255.255 area 0

end

!router 2

conf t

in s0/0/0

ip add 192.168.0.2 255.255.255.252

no shut

exit

in s0/1/0

ip add 192.168.1.2 255.255.255.252

no shut

exit

in s0/2/0

ip add 192.168.2.2 255.255.255.252

no shut

exit

in s0/3/0

ip add 192.168.3.2 255.255.255.252

no shut

exit

router os 1

network 192.168.0.0 0.0.255.255 area 0

end

If you enable "regular" CEF it will hash flows across the serial links.  If you enable the packet-by-packet option, it will round robin a flow's packets across the links.  (Note the above assumes ECMP, allowing 4 routes.)

Again it comes down to IP CEF vs. MLPPP, and NOT  OSPF vs. MLPPP.

You're assuming that the routers will be using the T1's to talk to each other. I am assuming he will be putting 4 T1's on each router, that each router will be internet facing, the routers will not be talking to each other, and therefore would render OSPF completley useless. Now show me the CEF scenario.

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

You're assuming that the routers will be using the T1's to talk to each other.

No, I'm not assuming that, as the OP states:

I will be installing two 2811 routers with four dedicated (point-to-point) T1 connections.

Is it better (faster) to use Multilink PPP, IP CEF, or can I use a combination of both?

On the other hand,

I am assuming he will be putting 4 T1's on each router, that each router will be internet facing, the routers will not be talking to each other, and therefore would render OSPF completley useless.

OP doesn't mention Internet, so yes, I agree, you're assuming.  Also "dedicated (point-to-point T1 connections", at least to me, implies leased lines between two of the OP's routers, i.e. those routers will be talking to each other and have nothing, whatsoever, to do with Internet facing.  So, OSPF might not be completely useless as you also assume.

But, even using your assumption, that we're working with Internet facing links, IP CEF works with more than OSPF.  For example, you could use it with BGP or with, I believe, static routes too.

Now show me the CEF scenario.

Perhaps for Internet statics:

!router 1 (client side)

conf t

in s0/0/0

ip add 192.168.0.1 255.255.255.252

no shut

exit

in s0/1/0

ip add 192.168.1.1 255.255.255.252

no shut

exit

in s0/2/0

ip add 192.168.2.1 255.255.255.252

no shut

exit

in s0/3/0

ip add 192.168.3.1 255.255.255.252

no shut

exit

ip route 0.0.0.0 0.0.0.0 192.168.0.2

ip route 0.0.0.0 0.0.0.0 192.168.1.2

ip route 0.0.0.0 0.0.0.0 192.168.2.2

ip route 0.0.0.0 0.0.0.0 192.168.3.2

end

!router 2 (ISP side)

conf t

in s0/0/0

ip add 192.168.0.2 255.255.255.252

no shut

exit

in s0/1/0

ip add 192.168.1.2 255.255.255.252

no shut

exit

in s0/2/0

ip add 192.168.2.2 255.255.255.252

no shut

exit

in s0/3/0

ip add 192.168.3.2 255.255.255.252

no shut

exit

!where a.b.c.d is some public /24

ip route a.b.c.d 0.0.0.255 192.168.0.1

ip route a.b.c.d 0.0.0.255 192.168.1.1

ip route a.b.c.d 0.0.0.255 192.168.2.1

ip route a.b.c.d 0.0.0.255 192.168.3.1

end

Josep,

Do you get this a lot on ypur routers (just wondering)?

#ping 8.8.8.8

Type escape sequence to abort.

Sending 5, 100-byte ICMP Echos to 8.8.8.8, timeout is 2 seconds:

.!.!.

Disclaimer

The  Author of this posting offers the information contained within this  posting without consideration and with the reader's understanding that  there's no implied or expressed suitability or fitness for any purpose.  Information provided is for informational purposes only and should not  be construed as rendering professional advice of any kind. Usage of this  posting's information is solely at reader's own risk.

Liability Disclaimer

In  no event shall Author be liable for any damages whatsoever (including,  without limitation, damages for loss of use, data or profit) arising out  of the use or inability to use the posting's information even if Author  has been advised of the possibility of such damage.

Posting

I don't routinely have a need to ping external IPs, but just tried 8.8.8.8:

#ping 8.8.8.8 re 1000

Type escape sequence to abort.

Sending 1000, 100-byte ICMP Echos to 8.8.8.8, timeout is 2 seconds:

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!

Success rate is 100 percent (1000/1000), round-trip min/avg/max = 8/15/42 ms

Of course, I'm not using MLPPP

Of course that's not tested on a router that has more than one WAN facing interface, using only CEF. I'm trying to point out that you need to use route-maps to acheive that result. With MLPPP, of course you need not configure route maps.

Disclaimer

The   Author of this posting offers the information contained within this   posting without consideration and with the reader's understanding that   there's no implied or expressed suitability or fitness for any purpose.   Information provided is for informational purposes only and should not   be construed as rendering professional advice of any kind. Usage of  this  posting's information is solely at reader's own risk.

Liability Disclaimer

In   no event shall Author be liable for any damages whatsoever (including,   without limitation, damages for loss of use, data or profit) arising  out  of the use or inability to use the posting's information even if  Author  has been advised of the possibility of such damage.

Posting

Oh, sorry, I didn't realize you were looking for a WAN facing IP CEF test.

Unfortunately, in my current position, I don't support any multi-p2p to the Internet.  My "WAN" is generally Ethernet, either dedicated/private fiber or on our OTN, usually gig, some 10g.  We may have a few multi-p2p between internal pairs, but it's rather rare as either we're using bundled links (Etherchannel) or the multi-p2p goes to different devices on other side (which may or may not be ECMP).

In any case, since CEF came out, don't recall ever seeing a ping issue like you describe on any logical ECMP.

Unsure why you belive you need to use route-maps with CEF.

Perhaps you should post your (complete) config that encounters the ping issue you describe.

  We discussed this in another thread, and when I said something like, "CEF is not best for WAN", well this scenario is what I was refering to. OK, it's fine you have multilple parrallel links, but still not as good as MLPPP.

However in the scenario we are discussing now, Yiou simply cannot do it with CEF alone.How could you? CEF uses Layer2 info to decide what the best next hop is, and with a routing protocol like OSPF to advertise all the hops in a route, CEF will not get you end to end. In a WAN facing situation, CEF only knows that the next best hopis the ISPs router, and nothing more.  In short, CEF will get you to next best-hop. That means all of the routers in the chain have to share route info, which is what OSPF does, and why CEF cannot route over a WAN facing link.

Now I realize that MLPPP also resides at layer 2, but you have to coordinate it with your ISP and so it works perfectly. Thats why you get 6 mbps with 4 T1's using MLPPP. You simply cannot do this with CEF.

Disclaimer

The  Author of this posting offers the information contained within this  posting without consideration and with the reader's understanding that  there's no implied or expressed suitability or fitness for any purpose.  Information provided is for informational purposes only and should not  be construed as rendering professional advice of any kind. Usage of this  posting's information is solely at reader's own risk.

Liability Disclaimer

In  no event shall Author be liable for any damages whatsoever (including,  without limitation, damages for loss of use, data or profit) arising out  of the use or inability to use the posting's information even if Author  has been advised of the possibility of such damage.

Posting

Chad as Peter would say "I beg to differ" with much of your last post's contents as it appears to have many misconceptions about CEF.

Can you please provide a working config consisting of multiple (2 or more) WAN facing interfaces, each with their own public IP addresses, and routing out of those interfaces with CEF? I'd like to see it work in the way that you insist it does. (obviously you shan't be able to use OSPF or BGP or such)

Disclaimer

The    Author of this posting offers the information contained within this    posting without consideration and with the reader's understanding that    there's no implied or expressed suitability or fitness for any  purpose.   Information provided is for informational purposes only and  should not   be construed as rendering professional advice of any kind.  Usage of  this  posting's information is solely at reader's own risk.

Liability Disclaimer

In    no event shall Author be liable for any damages whatsoever  (including,   without limitation, damages for loss of use, data or  profit) arising  out  of the use or inability to use the posting's  information even if  Author  has been advised of the possibility of such  damage.

Posting

As I've already written, I don't currently support any multi-link, on the Internet, with public IPs.

As I've also already written, I might find a current example with private IPs, but as I also noted, these in my current environment would be rare somewhat excepting ECMP.

Do you have such a device I could configure for you?  Or review its config?

Review Cisco Networking products for a $25 gift card