理由：
1、eigrp協議可控性差，網絡衝突比較嚴重？
2、eigrp穩定性差，任何一個網點的錯誤配置會導致路由環路，甚至全網癱瘓，ospf就不會。？
3、ospf有安全認證，eigrp沒有。
4、eigrp配置，使得網點的路由表龐大，佔用大量線路帶寬？
甲方：
1.eigrp協議可控性差，網絡衝突比較嚴重？
what is "網絡衝突"?
2.eigrp穩定性差，任何一個網點的錯誤配置會導致路由環路，甚至全網癱瘓，ospf就不會。？
Yes in theory."任何一個網點的錯誤配置會導致路由環路" may be that, but not always. Also in low bandwidth links and unstable links, may lead to "stuck in active".But the convergent time of ospf is not as fast as I think.
3.ospf有安全認證，eigrp沒有。
No. eigrp also has md5 authentication.
4.eigrp配置，使得網點的路由表龐大，佔用大量線路帶寬？
eigrp can also summarize routes, but not as flexible as ospf.
Since you have 300 nodes, I also think ospf is better.And ospf is more standard.But nowadays eigrp still is used and also you can use eigrp stub feature to minimize the records in routing table.
Generally ospf is good for networks that over 100 nodes.
乙方：
“eigrp can also summarize routes, but not as flexible as ospf.”
I m NOT with you.
eigrp can summarized by Interface, while OSPF has to summarize at ABR or ASBR. for example R1----R2----R3 all in the same OSPF area, then you can not summarized anywhere, while Eigrp is much more flexible.
1. "任何一個網點的錯誤配置會導致路由環路", when you use any routing protocol, you will have the same problem, if you 錯誤配置.
2. OSPF use LSA, when the number of router increase, LSA will use a lot of memory plus SPF.
3. when link flapping, when you use OSPF, you will need to be careful for SPF recalculating on every routers, except you use STUB/NSSA.
4. OSPF only supports equal cost load balance, while EIGRP i can do unequal cost
5. EIGRP using 90/170 AD to avoid routing loop, while OSPF treat it the same.
6. OSPF have to bring in area 0 concept for large network, and area 0 have to full redundant or at least some what degree redundant. EIGRP does not have such restriction.
7. EIGRP stuck in active, just wondering, is there anyone really see yourself happening in real life? and how many time?
甲方
Yes. But how about the feature of OSPF Incremental SPF since 12. (2)18s to lower the SPF load?
And in eigrp using interface summary, I just can summarize the routes with classical prefix, e.g. I just can summarize 192.168.0.0/24, but can not summarize with 192.168.0.0/26. But ospf can witch "area x range" command (Although the config will be more complex).
Actually I have not met SIA in eigrp, I just have found some cases in web site. I remember the web site mentioned that unstable link might cause SIA. In ospf, that will cause SPF recalculation, But we can setup the spf timers with "timer spf" command in low bandwidth wan links.
Also ospf protocol is improving, but I am not sure if eigrp has new versions.
with your experience which protocol runs better?
乙方：
“And in eigrp using interface summary, I just can summarize the routes with classical prefix, e.g. I just can summarize 192.168.0.0/24, but can not summarize with 192.168.0.0/26. But ospf can witch "area x range" command (Although the config will be more complex。-甲方”
interface Ethernet0
ip summary-address eigrp 1 192.168.0.0 255.255.255.192
“But we can setup the spf timers with "timer spf" command in low bandwidth wan links.-甲方”
timer spf is used to ajust/relay spf calculation, it does not resolve the sutiation when it is flapping/unstable. also it delay the convergence time. in large network, when you delay convergence time, have higher chance cause the LOOP.
“Yes. But how about the feature of OSPF Incremental SPF since 12.(2)18s to lower the SPF load?-甲方”
lower the SPF load, because it NEEDS it.
“Actually I have not met SIA in eigrp, I just have found some cases in web site. I remember the web site mentioned that unstable link might cause SIA.-甲方”
actually i did see this SIA things happen. when SIA, the flapping route will disappear from the down stream router's routing table, it is actually BETTER for down stream routers; all the down stream routers will think the flapping link/route is NOT reachable.
甲方：
“interface Ethernet0
ip summary-address eigrp 1 192.168.0.0 255.255.255.192-乙方”
To my strange, I remember in my lab, when I summary address in rip or eigrp with 192.168.0.0/26 or like this,their will be an error message. But today I tried with my new routers and layer3 switches but they can.
“timer spf is used to ajust/relay spf calculation, it does not resolve the sutiation when it is flapping/unstable. also it delay the convergence time. in large network, when you delay convergence time, have higher chance cause the LOOP.-乙方”
the timer spf command will delay convergence time,but it can lower the spf calculation in unstable links. I think it's good to use the command just on routers that connected to unstable WAN links.
“actually i did see this SIA things happen. when SIA, the flapping route will disappear from the down stream router's routing table, it is actually BETTER for down stream routers; all the down stream routers will think the flapping link/route is NOT reachable.-乙方”
But at that time, it means that all eigrp routers are recomputing. but in ospf, the recomputing just within the area.
In eigrp, when there is a topology change and there is no feasible successor, eigrp needs to send queries to all of the eigrp routers, and waits for reply, which is the reason of SIA. And in OSPF, topology change just need to flood the LSA in the same area. And in eigrp, topology change must flood to all routers, but in ospf, it's not the case if in totally stub area.
eigrp is less cpu intensive than ospf, simple to config, unequal load-balance which ospf does not have, and also route summarized feature.But I will choose ospf (may be just my preference), because I like to use stub area, totally stub area and NSSA to minimize the records in routing table. And it's not cisco proprietary protocol.
Maybe that's because of my concept is not clear enough and lack of experience, I want to ask y7975 a question: How will you choose routing protocol when you design a network and reason?
乙方：
“But at that time, it means that all eigrp routers are recomputing. but in ospf, the recomputing just within the area.-甲方”
ok, if the route/link that lost is E2 in OSPF, which is using LSA type 5, it will go to all areas except stub area, does not mean "all" routers recomputing? LSA type 3, does not go to other AREAs? Then force you to configure all areas (except 0) to STUB/NSSA, just to compensate the draw back of LSA.
“In eigrp, when there is a topology change and there is no feasible successor, eigrp needs to send queries to all of the eigrp routers, and waits for reply, which is the reason of SIA.-甲方”
If EIGRP can get stuck in active(query without answer), for the same situation, you think OSPF will received LSA?
“question: How will you choose routing protocol when you design a network and reason?-甲方”
1. $
2. vendors platform (cisco or other vendors)
3. physical connection, topology and number of nodes(layer 1 first)
4. expandabliity (connect with others)
5. other routing protocol may involve? ex. BGP
6. Firewall (such as PIX only support RIP and OSPF now)
7. *** (such as *** concentrator only have RIP and OSPF now)
8. delay sensative application (ex. VOICE)
甲方：
Thank you. I am more clear now.And it's very important that if other routing protocols are involved, eigrp should be better for it's simplicity, eigrp need not think about ABR,ASBR. That's what I ignored before. Now I realized why one of our HongKong customer, use eigrp instead of ospf, and there were about a hundred of routes in the routing table.(May be because most of networks I meet are typical hub and spoke networks, which make me tend to use ospf)Also varied router roles in OSPF make the design complex, and high loading in ASR,ASBR,and core routers,also the size of ospf area needs to be concerned. But I still have 2 questions:
1.
----------------------------------------------------------------------
In eigrp, when there is a topology change and there is no feasible successor, eigrp needs to send queries to all of the eigrp routers, and waits for reply, which is the reason of SIA.
----------------------------------------------------------------------
If EIGRP can get stuck in active(query without answer), for the same situation, you think OSPF will received LSA?
If in the following topology:
LAN1--R1--R2--R3--LAN2
If LAN1 fails, link between R2 and R3 is unstable. R1 will send LSA to R2, R2 will send to R3 but may fail. I think that's no problem in OSPF but may lead SIA in EIGRP, because R1 is waiting for the reply from R3.
2. For the eigrp stub feature. In cisco document: "A stub router should not have any EIGRP neighbors other than distribution routers." So I think this means that the remote site just can have 1 router so it's an impediment to implement eigrp stub, and ospf stub area can include serveral routers.
乙方：
“If in the following topology:
LAN1--R1--R2--R3--LAN2
If LAN1 fails, link between R2 and R3 is unstable. R1 will send LSA to R2, R2 will send to R3 but may fail. I think that's no problem in OSPF but may lead SIA in EIGRP, because R1 is waiting for the reply from R3.---甲方”
not a problem, SIA will just time out. (default 3 mins ?)
“2. For the eigrp stub feature. In cisco document: "A stub router should not have any EIGRP neighbors other than distribution routers." So I think this means that the remote site just can have 1 router so it's an impediment to implement eigrp stub, and ospf stub area can include serveral routers.——甲方”
correct. the be close to what OSPF STUB doing, EIGRP needs to do "ip summary-addr 0.0.0.0 0.0.0.0"