返回列表 发帖

Re: mpls lsp 如何克服ip 路由查找的hop by hop 问题?

我的理解是正如楼上所说,MPLS通过标签查找避免了IP的路由下一跳查找。而且,基于CEF,LFIB生成后根本不看IP,只看进站标记和对应的出站标记。

TOP

Re: mpls lsp 如何克服ip 路由查找的hop by hop 问题?

IP is looked as "hop by hop" because the route to destination can DIFFER in each router along the path (to the destination). Therefore, the route actually taken from source to destination can be different from what the SOURCE router would like. In other words, each router (along the path) is only taking the path they thought would be the best.

? This situation should only occur when using distance vector routing protocols because link state protocols have a complete view of the topology.

? MPLS is also known as source-based protocol because packet is send on the path exactly desired by the source router.

Regards,

Angela

TOP

Re: mpls lsp 如何克服ip 路由查找的hop by hop 问题?

angela ,source 知道路径中经过的每一个节点的标签吗?在途经的LSR上不进行标签查找吗?

TOP

Re: mpls lsp 如何克服ip 路由查找的hop by hop 问题?

Source doesn't know the label assigned/used at each LSR, but the label used on source will also result in the path (LSP) on the next hop LSR. Label swapping/popping/pushing is still done at each LSR, but the point is, the route from source to destination has already been determined on the source and will not be changed by the intermediate routers.

For instance, Router A (source) wants reach Router C (destination). Router B (intermediate LSR) has 2 ways to reach Router C, one to go from its Serial link and another via Router D from the GE interface. Router A wants to go from the Serial interface. If using IP, the packet will go directly to Router C (because of the faster bandwidth), which is not the desired result. You certainly can use features such as PBR to change this, but in a complex network, this will be error-prone and tiresome

? MPLS will establish an LSP from Router A to Router C and the other way around (so data can be exchanged). Suppose the route of Router A->Router B->Router D->Router C is assigned label 21 while route Router A->Router B->Router C is assigned label 20, Router A will use label 20 instead of label 21 to achieve the desired result, this eliminate the need of PBR and many other tweaking.

Regards,

Angela

TOP

Re: mpls lsp 如何克服ip 路由查找的hop by hop 问题?

这个问题的核心点是:有没有确定路径,而不是对比两种处理 的查找。

ip routing 在每一跳,源到目的地的路径都没有确定,都需要查找路由,来确定 下一跳

mpls 的lsp ,已经根据routing 确定了源到目的地的路径,在LSR上,知道下一跳是谁,只查找对应的标签进行。

angela , this? ok

TOP

Re: mpls lsp 如何克服ip 路由查找的hop by hop 问题?

MPLS 的最大特点就是她其实是 "switching" 而不是 "routing" 就如她的名字一样是 MPL "switching" .? 在ISP P 设备里通过 MPLS 会生成 Label, 当数据包通过时, 只是进行简单的 Label 查询 ( 与switch 进行MAC地址查询相似). 这个查询是不用耗什么CPU 资源的基于HW base 的查询, 所以, 根本不会用上 Routing 的那些方式(packet 送到CPU 去做routing table 查询等等). ... 当然, 如果, 出于任何原因, 在通往数据包目的地址的path上, Label没有生成, 那又是另外一回事啦.... 至于, MPLS 怎么生成Label, 怎么样propagate, 是MPLS 的技术核心所在, 如果想了解深入, 可以从CCIP- mpls-study-guide 开始

TOP

Re: mpls lsp 如何克服ip 路由查找的hop by hop 问题?

补充两点,

在Ingress router(PE)上,其实差不多.

在P router,MPLS关键是Label/Path pre-allocated. 主要是table小(因为用BGP next-hop,所以只要对ISP主干上的节点做mapping).

TOP

Re: mpls lsp 如何克服ip 路由查找的hop by hop 问题?

进行路由查找的时候,每经过一个路由器都要拆到3层查看IP报头,再查路由器转发表,然后进行转发。

在进行标签转发的时候,只是拆到2.5层看标签。我们可以把建立的LSP看作是一条河道,LSP的首端只需查看IP报头,决定扔到哪一条LSP,那么数据就会从LSP首端发到LSP的尾端。中间的路由器不用去查IP报头来决定向哪一个路由器转发。

TOP

返回列表 回复 发帖