BGP策略的几篇经典文章(1)

 

1 《Detecting BGP Configuration Faults with Static Analysis》NSDI2005

主要是一个router configuration checker，并给出了其实现。

站在一个自治系统的角度来静态检查路由器配置，通过抽象出的两类正确配置原则：path visibility 和 route validity，and map this specification

to conditions on the BGP configuration that must hold

for those aspects of correctness to hold来检查configuration fault.

 

将rrc应用于实际网络中，发现配置错误导致的fault主要有三类：

These include: (1) faults that could

have caused network partitions due to errors in how external

BGP informationwas being propagated to routers inside

an AS, (2) faults that cause invalid routes to propagate inside

an AS, and (3) faults in policy expression that caused

routers to advertise, and hence potentially forward packets,

in a manner inconsistent with the AS’s desired policies.

 

分析配置错误的原因主要有三种

First, the mechanisms for propagating routes within a network

are overly complex. The main techniques used to

propagate routes scalably within a network (e.g., “route re-

flection with clusters”) are easily misconfigured. Second,

even simple policy specifications (e.g., controlling route

propagation) require configuration fragments on many (if

not all) routers in the network, and each fragment often

involves several layers of indirection (e.g., “route maps”,

“community lists”, etc.). Finally, operators lack a systematic

process for configuring functions such as filtering.

 

他的数据源就是由路由器的实际配置语言而转化成的一种中间的厂商无关的表示形式.

 

2 《In Search for an Appropriate Granularity to Model Routing Policies》sigcom 2007

1 为什么需要quasi-router?

本质上是为了捕捉路径选择的多样性，所以构建quasi-router的时候就是根据如果一个自治系统有关于一个前缀的不同路径选择时，就用quasi-router来划分具有不同选路决定的铜一个自治系统中的路由器们。

2 对于使用商业关系作为BGP policy的模型的评价

商业关系对于BGP路径选择的影响体现在两个方面：preference与non-transit。其中preference表示customer routes are preferred over peering routes, and peering routes over provider routes.而non-transit则表示valley-free的特性。

一方面，本文中说，放置filter的很多流行位置都与valley-free特性中的valley是相符合的。但是另一方面，商业关系中的preference所表达的信息还是太过粗糙，以至于an AS may learn multiple customer, provider or peering routes for the same prefix. In such a case, business relationships cannot tell which one of the equally good routes should be selected as best，即还是不能全面完整的模拟BGP路径选择（所谓BGP路径选择，即which path to select as best from a set of equivalent paths, all permitted by policies.）。

3 关于next-hop atom

在BGP policy建模中，既然使用BGP atom太细，商业关系太粗，那么他自然就提出了结合两者的一种定义了，A next-hop atom captures the coarsest granularity (across prefixes)

at which an AS chooses its best paths in distinct ways (among its neighbors)（因为别忘了next-hop atom的定义就是基于BGP atom来进行定义的）。在他们的实验中发现，很多的next-hop atom是只有一个邻居的，这说明了per-neighbor path preferences appear quite common in the Internet, a non-negligible fraction of the path choices are made not by routing policies, but by tie-breaking within the BGP decision process.

3  

《Analyzing BGP Policies: Methodology and Tool》 infocom 2004

1) 

1 BGP策略的概括

The filter can include a list

of routes, a list of regular expressions on the AS PATH, a

list of communities, or any possible combination of these

three

2)能够有效地读取IRR中的使用RPSL描述的策略，所谓有效就是可以比较好的提取出策略关系、去除错误的不准确的等。
然后基于这个来推测AS之间的常规relationship,并用RouteView等公网路由表进行验证。
个人感觉缺点是：首先近两年都已经把relationship不仅仅局限于那么简单的三类了，而是更细粒度的，比如quasi-router等，
从这个意义上讲有点滞后；第二，因为策略不光是用作商业关系的，还有用在流量工程、安全等方面，从这个角度讲对于策略
的验证也是不全面的。