RIP and IGRP load balancing

totts

I'm using the Cisco press and Sybex books and having difficulty in differentiating load balancing between RIP and IGRP. Can anyone help clear this up for me?

This is what I understand/believe...

RIP uses hop count as its metric so if it finds 2 or more routes to a remote network with the same hop count then it will perform equal cost load balancing between them regardless of their bandwith.

RIP has a default of 4 paths and a range of 1-6 paths which you can change issuing the
maximum-paths command. If you issue the command maximum-paths 1, then it'll use the first path it learned and only that path.

...............................................................................

IGRP uses bandwidth and delay as its metric by default, but it also believes that each of the routers serial interfaces are using a T1 link. In common with RIP, it also performs equal cost load balancing over 4 paths by default and again can be changed issuing the maximum-paths command to the same effect as RIP. This is where I'm confused... am I right so far?

OR...

is the maximum-paths command not used with IGRP and the variance command used in its place?

I understand that you can issue the 'traffic-share min' command with IGRP to so that only the lowest cost path is used instead of load balancing.

Netstudent

well the IGRP metric is a long formula which means that 2 routes might not have the same metric. The variance command sets a mulitplier that will set the amount of variance that is considered equal among the routes.

totts

Thanks for that netstudent. Do you know whether the maximum-path command is used with IGRP as it is with RIP?

Netstudent

I think it works for all Ip routing protocols

totts

Cisco Press bunches the commands for RIP and IGRP together so I'm thinking it could be misleading and the explanation frustrates me rather than irons it out. However, I think this may be the case, ie RIP, IGRP and EIGRP uses the maximum-path command, but I haven't seen it featured for OSPF!

dtlokee

totts wrote:

Cisco Press bunches the commands for RIP and IGRP together so I'm thinking it could be misleading and the explanation frustrates me rather than irons it out. However, I think this may be the case, ie RIP, IGRP and EIGRP uses the maximum-path command, but I haven't seen it featured for OSPF!

In router configuration mode you will find the maximum-paths command, and on newer IOS versions it can be increased to 16

Netstudent

What routing protocols will allow for 16 equal path routes to the same destination in the routing table? I'm thinking probably an EGP?

dtlokee

IOS 12.4

P1R1(config)#router ospf 100
P1R1(config-router)#maxim
P1R1(config-router)#maximum-paths ?
  <1-16>  Number of paths

P1R1(config-router)#maximum-paths 4
P1R1(config-router)#exit
P1R1(config)#router eigrp 100
P1R1(config-router)#maxim
P1R1(config-router)#maximum-paths ?
  <1-16>  Number of paths

P1R1(config-router)#maximum-paths 4
P1R1(config-router)#
P1R1(config-router)#exit
P1R1(config)#router rip
P1R1(config-router)#maximum
P1R1(config-router)#maximum-paths ?
  <1-16>  Number of paths

P1R1(config-router)#maximum-paths 4
P1R1(config-router)#
P1R1(config-router)#exit
P1R1(config)#router bgp 100
P1R1(config-router)#ma
P1R1(config-router)#maximum-paths ?
  <1-16>  Number of paths
  ibgp    iBGP-multipath

P1R1(config-router)#
                  

It is available for RIP, OSPF, EIGRP, and BGP (IGRP is removed from IOS 12.4)

Netstudent

oh cool. So the number of equal cost paths that can populate the table isn't necesarily a characteristic of the routing protocol version, but rather a characteristic of the IOS version?

tech-airman

totts wrote:

I'm using the Cisco press and Sybex books and having difficulty in differentiating load balancing between RIP and IGRP. Can anyone help clear this up for me?

This is what I understand/believe...

RIP uses hop count as its metric so if it finds 2 or more routes to a remote network with the same hop count then it will perform equal cost load balancing between them regardless of their bandwith.

RIP has a default of 4 paths and a range of 1-6 paths which you can change issuing the
maximum-paths command. If you issue the command maximum-paths 1, then it'll use the first path it learned and only that path.

...............................................................................

IGRP uses bandwidth and delay as its metric by default, but it also believes that each of the routers serial interfaces are using a T1 link. In common with RIP, it also performs equal cost load balancing over 4 paths by default and again can be changed issuing the maximum-paths command to the same effect as RIP. This is where I'm confused... am I right so far?

OR...

is the maximum-paths command not used with IGRP and the variance command used in its place?

I understand that you can issue the 'traffic-share min' command with IGRP to so that only the lowest cost path is used instead of load balancing.

totts,

Within the scope of the CCNA, it is important to understand the difference between:

1. Equal Cost Load Balancing
2. Unequal Cost Load Balancing

For the case of RIP, here's how Equal Cost Load Balancing works.

|                                                                                   |
|                                                                                   |
|                [S0]-----z-----[S1](RouterC)[S0]-- --z-----[S1]                    |
+---[E0](RouterA)[S1]-----z-----[S0](RouterB)[S1]-----z-----[S0](RouterD)[E0]-------+
|                                                                                   |
|                                                                                   |
|                                                                                   |

192.168.1.0                                                                192.168.2.0

As we know, RIP uses a metric called "hop count." So based on the above network, from the perspective of RouterA, it takes 2 hops to get from network 192.168.1.0 to 192.168.2.0. Since the path from RouterA->RouterB->RouterD is 2 hops and the path from RouterA->RouterC->RouterD is also 2 hops, they are both considered "equal cost." So based on equal cost load balancing, RouterA will send the first packet out of it's Serial0 interface, the second packet out of it's Serial1 interface, send the third packet out of it's Serial0 interface, send the fourth packet out of it's Serial1 interface, and so on and so forth. So therefore the load is "balanced" between the two egress interfaces.

To appreciate the benefits of "load balancing", imagine there was no load balancing. So if there's no load balancing, RouterA will send traffic destined for 192.168.2.0 out of a single and same serial interface all the time. The effect is that the other router will effectively be an expensive doorstop that just happens to be networkingly connected to the network.

Now, to appreciate the benefits of IGRP's "unequal cost load balancing" feature, imagine the links between the following pairs of routers are 56kbps: 1) RouterA and RouterB 2) RouterB and RouterD. Imagine the links between the following pairs of routers are 1.544Mbps: 1) RouterA and RouterC 2) RouterC and RouterD.

Based on RIP's "hop count" metric, it doesn't care about the faster path from RouterA->RouterC->RouterD compared to the path from RouterA->RouterB->RouterD. So just like the above "equal cost load balancing" situation, RouterA is going to route packets equally out of Serial0 and Serial1. However, since IGRP's metric takes "bandwidth" into consideration, under unequal cost load balancing, IGRP is likely to forward more packets out of it's Serial0 before a packet is forwarded out of it's Serial1 interface. As previously mentioned, this is done so that RouterB isn't a networked doorstop.

So in summary:

1. RIP uses equal cost load balancing
2. IGRP uses unequal cost load balancing

I hope this helps.

cristi.grigore

tech-airman wrote:

So in summary:

1. RIP uses equal cost load balancing
2. IGRP uses unequal cost load balancing

I hope this helps.

That is not at all correct.

All routing protocols running in Cisco IOS use equal cost load balancing by deafult. The default number of paths used is 4, for all IGP's (RIP, IGRP, EIGRP, OSPF and IS-IS), and 1 for BGP. The maximum number of paths is 16 (starting from IOS 12.3T - it was 6 before that).

IGRP and EIGRP, beside being able to do equal cost load balancing, can also do unequal cost load balancing, by using the variance command. It works like this: (E)IGRP determines the lowest metric to the destination (let's say 150); if you set the variance to a value of 2, all the paths that don't contain a loop and have the metric greater than the minimum metric and lower than, or equal to, the minimum metric times the variance is selected for load balancing, i.e. all routes that have a metric between 150 and 150x2=300.

By default, the variance is set to 1, which means equal cost load balancing for both IGRP and EIGRP - the interval for the above example would be 150 to 150x1=150.

totts, bear in mind that IGRP is no longer present in IOS 12.4 or in the new CCNA certification exam (640-802).

So in summary:

1. RIP, IGRP and EIGRP use equal cost load balancing by default
2. IGRP and EIGRP can use unequal cost load balancing

Netstudent

Okay so I don't understand how Techairman is "not at all correct" when the only difference in your and his explanation is that he didn't say anything about being default. I believe he was explaining pin-hole congestion from a low bandwidth bottleneck caused by an unintelligent metric. Instead of contradicting a reputable member, maybe you could have added to his explanation instead of going on about how he is "not at all correct". Oh and welcome to the board.

dtlokee

Netstudent wrote:

Okay so I don't understand how Techairman is "not at all correct" when the only difference in your and his explanation is that he didn't say anything about being default. I believe he was explaining pin-hole congestion from a low bandwidth bottleneck caused by an unintelligent metric. Instead of contradicting a reputable member, maybe you could have added to his explanation instead of going on about how he is "not at all correct". Oh and welcome to the board.

The statement made about "IGRP uses unequal cost load balancing" isn't correct in the context it was presented. There was no mention of the variance command in the explanation or in the summary. I guess the better statement would have been "IGRP can use unequal cost load balancing" or something to that effect. It's splitting hairs, but that is what we do here. We contemplate every possible scenario backwards and forwards, no matter how ridiculous circumstances and try to understand the outcome. I think the point was to eliminate confusion about IGRP or EIGRP load balancing. And of course all our obsessing on this topic is mute. In a production network you don't want your routers load balancing packets between a 64k link and a T-1. Why? Jitter. It will cause unpredictable delivery of the packets, which we try to avoid.

Feel the love

Netstudent

Thats totally understandable. I just didn't get the point of signing up at techexams just to tell someone they were wrong when in actuality they weren't "completely" wrong, but just worded it kinda wrong. And yes, disecting each scenario no matter what the outcome is very prevalent here.

cristi.grigore

Sorry if I seemed a bit harsh, but I felt that I had to make the point as it wasn't obvious form the explanation that (E)IGRP does not use unequal load balancing unless instructed to (as long as you were all talking about splitting hairs...).

I apologize again if I was too blunt, my intention was to clarify rather than attack

Hope my post helped.

mikej412

cristi.grigore wrote:

Hope my post helped.

I thought it was a "good catch" -- Welcome to the forum!

Netstudent

cristi.grigore wrote:

Sorry if I seemed a bit harsh, but I felt that I had to make the point as it wasn't obvious form the explanation that (E)IGRP does not use unequal load balancing unless instructed to (as long as you were all talking about splitting hairs...).

I apologize again if I was too blunt, my intention was to clarify rather than attack

Hope my post helped.

it's all good and your clarification is spot on. The more knowledgable people we have around here the better learning environment it is for everyone. Hope to see you sharing some good info here for all of us cisco junkies!

tech-airman

cristi.grigore wrote:

tech-airman wrote:

So in summary:

1. RIP uses equal cost load balancing
2. IGRP uses unequal cost load balancing

I hope this helps.

That is not at all correct.

All routing protocols running in Cisco IOS use equal cost load balancing by deafult. The default number of paths used is 4, for all IGP's (RIP, IGRP, EIGRP, OSPF and IS-IS), and 1 for BGP. The maximum number of paths is 16 (starting from IOS 12.3T - it was 6 before that).

IGRP and EIGRP, beside being able to do equal cost load balancing, can also do unequal cost load balancing, by using the variance command. It works like this: (E)IGRP determines the lowest metric to the destination (let's say 150); if you set the variance to a value of 2, all the paths that don't contain a loop and have the metric greater than the minimum metric and lower than, or equal to, the minimum metric times the variance is selected for load balancing, i.e. all routes that have a metric between 150 and 150x2=300.

By default, the variance is set to 1, which means equal cost load balancing for both IGRP and EIGRP - the interval for the above example would be 150 to 150x1=150.

totts, bear in mind that IGRP is no longer present in IOS 12.4 or in the new CCNA certification exam (640-802).

So in summary:

1. RIP, IGRP and EIGRP use equal cost load balancing by default
2. IGRP and EIGRP can use unequal cost load balancing

cristi.grigore,

My response was based within the scope of the post title of "RIP and IGRP load balancing." The OP didn't ask about EIGRP, OSPF, IS-IS, or BGP. Therefore, mentioning those other Interior Gateway Protocols (IGP) and Exterior Gateway Protocol (EGP) may lead to increased confusion instead of trying to help reduce confusion by not containing the response within the scope of the question. So I based my reply on the outline of:

1. RIP
2. IGRP

Then, within the scope of that outline and the OP's original statement of "...having difficulty in differentiating load balancing between RIP and IGRP. Can anyone help clear this up for me? " , I expanded by trying to show how RIP executes "load balancing" of the type "equal cost" and how IGRP differs by using "load balancing" of the type "unequal cost." So I was trying to show the difference in how these routing protocols operate therefore trying to keep the discussion to the technology level. Without a proper understanding of the technology, trying to configure the technology may be that much more difficult and confusing. Also, I know that there are many levels of configuration for a given technology, e.g. OSPF. However, within the scope of the CCNA, you only need to know the single area case. In BSCI which is a professional level exam towards the CCNP/CCIP/CCDP, OSPF in multiple areas is fair game. At the CCIE level, the scope is that much wider. If you need a higher/deeper understanding of the technology to help you understand the "basic level" for the CCNA, then that's your learning style. However, there are those who can get or are "lost" because they delved too deeply into a topic that they don't come back to progress in studying for the CCNA therefore why I was trying to contain the scope of my reply.

From the perspective of nitpicking, my statement of "IGRP uses unequal cost load balancing" was a summary within the context of my post and wasn't intended to be an all encompassing statement. I would have used the key word "only" if "unequal" was the only load balancing method for IGRP for example "IGRP uses unequal cost load balancing only" which in that case I would be wrong. Logically, my statement was intedended to be equivalent to saying "my car uses 87 octane gasoline." It is also true that "my car CAN use 92 octane gasoline." So is my first statement false due to omission of the second statement? I don't think so. So I believe my statement is still true as long as I don't claim that it's the ONLY option which I haven't. As a reiteration, it was originally asked "...difficulty in differentiating load balancing ..." so that's the question I was answering.

totts

Well done Netstudent for keeping this one alive!!! I went away to Oxford for a wedding over weekend to come home and discover 17 posts! Thanks to everyone who's posted, it looks like some excellent debate and information has taken place. I'd better get my head around this lot, thanks

totts

Another question on this subject if you would... If you do use the variance command with
(E)IGRP to say utilise the bandwidth of a 512k line alongside the more desirable T1 line, then will it try to put equal amounts of traffic down each link?

I've just been reading about OSPF and misleading it to believe that two paths are of equal cost by altering either the bandwidth or cost. But because it believes they're of equal cost, it'll treat them that way and try to transmit equal amounts of data. Does this hold true when you're using the variance command with (E)IGRP?

mikej412

http://www.cisco.com/warp/public/103/19.html#traffic

EIGRP not only provides unequal cost path load balancing, but also intelligent load balancing, such as traffic sharing. In order to control how traffic is distributed among routes when there are multiple routes for the same destination network that have different costs, use the traffic-share balanced command. With the keyword balanced, the router distributes traffic proportionately to the ratios of the metrics that are associated with different routes. This is the default setting:

    router eigrp 1 
    network x.x.x.x 
    variance 2 
    traffic-share balanced

ITdude

Hey, splitting hairs is what we are all about here aren't we?