Network Working Group S. Dasgupta
Request for Comments: 5468 J. de Oliveira
Category: Informational Drexel University
JP. Vasseur
Cisco Systems
April 2009
Network Working Group S. Dasgupta
Request for Comments: 5468 J. de Oliveira
Category: Informational Drexel University
JP. Vasseur
Cisco Systems
April 2009
Performance Analysis of Inter-Domain Path Computation Methodologies
域间路径计算方法的性能分析
Status of This Memo
关于下段备忘
This memo provides information for the Internet community. It does not specify an Internet standard of any kind. Distribution of this memo is unlimited.
本备忘录为互联网社区提供信息。它没有规定任何类型的互联网标准。本备忘录的分发不受限制。
Copyright Notice
版权公告
Copyright (c) 2009 IETF Trust and the persons identified as the document authors. All rights reserved.
版权所有（c）2009 IETF信托基金和确定为文件作者的人员。版权所有。
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents in effect on the date of publication of this document (http://trustee.ietf.org/license-info). Please review these documents carefully, as they describe your rights and restrictions with respect to this document.
本文件受BCP 78和IETF信托在本文件出版之日生效的与IETF文件有关的法律规定的约束(http://trustee.ietf.org/license-info). 请仔细阅读这些文件，因为它们描述了您对本文件的权利和限制。
Abstract
摘要
This document presents a performance comparison between the per-domain path computation method and the Path Computation Element (PCE) Architecture-based Backward Recursive Path Computation (BRPC) procedure. Metrics to capture the significant performance aspects are identified, and detailed simulations are carried out on realistic scenarios. A performance analysis for each of the path computation methods is then undertaken.
本文介绍了基于逐域路径计算方法和基于路径计算元素（PCE）体系结构的反向递归路径计算（BRPC）过程的性能比较。确定了捕获重要性能方面的指标，并在实际场景中进行了详细的模拟。然后对每种路径计算方法进行性能分析。
Table of Contents
目录
1. Introduction ....................................................2
2. Terminology .....................................................3
3. Evaluation Metrics ..............................................4
4. Simulation Setup ................................................5
5. Results and Analysis ............................................6
5.1. Path Cost ..................................................7
5.2. Crankback/Setup Delay ......................................7
5.3. Signaling Failures .........................................8
5.4. Failed TE-LSPs/Bandwidth on Link Failures ..................8
5.5. TE LSP/Bandwidth Setup Capacity ............................8
6. Security Considerations .........................................9
7. Acknowledgment ..................................................9
8. Informative References ..........................................9
1. Introduction ....................................................2
2. Terminology .....................................................3
3. Evaluation Metrics ..............................................4
4. Simulation Setup ................................................5
5. Results and Analysis ............................................6
5.1. Path Cost ..................................................7
5.2. Crankback/Setup Delay ......................................7
5.3. Signaling Failures .........................................8
5.4. Failed TE-LSPs/Bandwidth on Link Failures ..................8
5.5. TE LSP/Bandwidth Setup Capacity ............................8
6. Security Considerations .........................................9
7. Acknowledgment ..................................................9
8. Informative References ..........................................9
The IETF has specified two approaches for the computation of inter-domain (Generalized) Multi-Protocol Label Switching ((G)MPLS) Traffic Engineering (TE) Label Switched Paths (LSP): the per-domain path computation approach defined in [RFC5152] and the PCE-based approach specified in [RFC4655]. More specifically, we study the PCE-based path computation model that makes use of the BRPC method outlined in [RFC5441]. In the rest of this document, we will call PD and PCE the per-domain path computation approach and the PCE path computation approach, respectively.
IETF规定了计算域间（通用）多协议标签交换（（G）MPLS）流量工程（TE）标签交换路径（LSP）的两种方法：[RFC5152]中定义的每域路径计算方法和[RFC4655]中指定的基于PCE的方法。更具体地说，我们研究了基于PCE的路径计算模型，该模型使用了[RFC5441]中概述的BRPC方法。在本文档的其余部分中，我们将PD和PCE分别称为每域路径计算方法和PCE路径计算方法。
In the per-domain path computation approach, each path segment within a domain is computed during the signaling process by each entry node of the domain up to the next-hop exit node of that same domain.
在每域路径计算方法中，域内的每个路径段在信令过程中由域的每个入口节点计算到同一域的下一跳出口节点。
In contrast, the PCE-based approach and, in particular, the BRPC method defined in [RFC5441] rely on the collaboration between a set of PCEs to find to shortest inter-domain path after the computation of which the corresponding TE LSP is signaled: path computation is undertaken using multiple PCEs in a backward recursive fashion from the destination domain to the source domain. The notion of a Virtual Shortest Path Tree (VSPT) is introduced. Each link of a VSPT represents the shortest path satisfying the set of required constraints between the border nodes of a domain and the destination LSR. The VSPT of each domain is returned by the corresponding PCE to create a new VSPT by PCEs present in other domains. [RFC5441] discusses the BRPC procedure in complete detail.
相反，基于PCE的方法，尤其是[RFC5441]中定义的BRPC方法依靠一组PCE之间的协作，在计算完相应的TE LSP后找到最短的域间路径：以向后递归的方式使用多个PCE从目标域到源域进行路径计算。引入了虚拟最短路径树（VSPT）的概念。VSPT的每个链路表示满足域边界节点和目标LSR之间所需约束集的最短路径。每个域的VSPT由相应的PCE返回，以通过其他域中的PCE创建新的VSPT。[RFC5441]详细讨论了BRPC程序。
This document presents some simulation results and analysis to compare the performance of the above two inter-domain path computation approaches. Two realistic topologies with accompanying traffic matrices are used to undertake the simulations.
本文给出了一些仿真结果和分析，以比较上述两种域间路径计算方法的性能。两个真实的拓扑和伴随的流量矩阵被用来进行模拟。
Note that although the simulation results discussed in this document have used inter-area networks, they also apply to Inter-AS cases.
请注意，尽管本文中讨论的模拟结果使用了区域间网络，但它们也适用于区域间AS情况。
Disclaimer: although simulations have been made on different and realistic topologies showing consistent results, the metrics shown below may vary with the network topology.
免责声明：尽管已经在不同和真实的拓扑上进行了模拟，显示了一致的结果，但下面显示的指标可能会因网络拓扑而异。
Note that this document refers to multiple figures that are only available in the PDF version.
请注意，本文档引用了多个仅在PDF版本中可用的图。
Terminology used in this document:
本文件中使用的术语：
TE LSP: Traffic Engineered Label Switched Path.
TE LSP：流量工程标签交换路径。
CSPF: Constrained Shortest Path First.
CSPF：约束最短路径优先。
PCE: Path Computation Element.
PCE：路径计算元素。
BRPC: Backward Recursive PCE-based Computation.
BRPC：基于PCE的反向递归计算。
AS: Autonomous System.
AS：自治系统。
ABR: Routers used to connect two IGP areas (areas in OSPF or levels in IS-IS).
ABR：用于连接两个IGP区域（OSPF中的区域或IS-IS中的级别）的路由器。
ASBR: Routers used to connect together ASes of a different or the same Service Provider via one or more Inter-AS links.
ASBR：用于通过一个或多个AS间链路将不同或相同服务提供商的ASE连接在一起的路由器。
Border LSR: A border LSR is either an ABR in the context of inter-area TE or an ASBR in the context of Inter-AS TE.
边界LSR：边界LSR是区域间TE上下文中的ABR或区域间TE上下文中的ASBR。
VSPT: Virtual Shortest Path Tree.
虚拟最短路径树。
LSA: Link State Advertisement.
链接状态广告。
LSR: Label Switching Router.
标签交换路由器。
IGP: Interior Gateway Protocol.
IGP：内部网关协议。
TED: Traffic Engineering Database.
交通工程数据库。
PD: Per-Domain
PD：每个域
This section discusses the metrics that are used to quantify and compare the performance of the two approaches.
本节讨论用于量化和比较两种方法性能的指标。
o Path Cost. The maximum and average path costs are observed for each TE LSP. The distributions for the maximum and average path costs are then compared for the two path computation approaches.
o 路径成本。观察每个TE LSP的最大和平均路径成本。然后比较两种路径计算方法的最大和平均路径成本分布。
o Signaling Failures. Signaling failures may occur in various circumstances. With PD, the head-end LSR chooses the downstream border router (ABR, ASBR) according to some selection criteria (IGP shortest path, ....) based on the information in its TED. This ABR then selects the next ABR using its TED, continuing the process till the destination is reached. At each step, the TED information could be out of date, potentially resulting in a signaling failure during setup. In the BRPC procedure, the PCEs are the ABRs that cooperate to form the VSPT based on the information in their respective TEDs. As in the case of the PD approach, information in the TED could be out of date, potentially resulting in signaling failures during setup. Also, only with the PD approach, another situation that leads to a signaling failure is when the selected exit ABR does not have any path obeying the set of constraints toward a downstream exit node or the TE LSP destination. This situation does not occur with the BRPC. The signaling failure metric captures the total number of signaling failures that occur during initial setup and re-route (on link failure) of a TE LSP. The distribution of the number of signaling failures encountered for all TE LSPs is then compared for the PD and BRPC methods.
o 信号故障。在各种情况下都可能发生信号故障。对于PD，前端LSR根据其TED中的信息根据一些选择标准（IGP最短路径等）选择下游边界路由器（ABR、ASBR）。然后，该ABR使用其TED选择下一个ABR，继续该过程直到到达目的地。在每个步骤中，TED信息都可能过期，可能导致设置过程中的信令故障。在BRPC程序中，PCE是ABR，它们根据各自TED中的信息协作形成VSPT。与PD方法的情况一样，TED中的信息可能过时，可能导致设置期间的信令故障。此外，仅在PD方法中，导致信令失败的另一种情况是当所选出口ABR没有任何路径服从朝向下游出口节点或TE LSP目的地的约束集时。BRPC不会出现这种情况。信令故障度量捕获TE LSP初始设置和重新路由（链路故障）期间发生的信令故障总数。然后比较PD和BRPC方法中所有TE LSP遇到的信令故障数的分布。
o Crankback Signaling. In this document, we made the assumption that in the case of PD, when an entry border node fails to find a route in the corresponding domain, boundary re-routing crankback [RFC4920] signaling was used. A crankback signaling message propagates to the entry border node of the domain and a new exit border node is chosen. After this, path computation takes place to find a path segment to a new entry border node of the next domain. This causes an additional delay in setup time. This metric captures the distribution of the number of crankback signals and the corresponding delay in setup time for a TE LSP when using PD. The total delay arising from the crankback signaling is proportional to the costs of the links over which the signal travels, i.e., the path that is setup from the entry border node of a domain to its exit border node (the assumption was made
o 回退信号。在本文中，我们假设在PD情况下，当入口边界节点无法在相应域中找到路由时，使用边界重路由回退[RFC4920]信令。回退信令消息传播到域的入口边界节点，并选择新的出口边界节点。在此之后，进行路径计算，以查找到下一个域的新入口边界节点的路径段。这会导致设置时间的额外延迟。当使用PD时，该度量捕获了TE LSP的拖转信号数量分布和相应的设置时间延迟。回退信令产生的总延迟与信号所经过的链路的成本成比例，即，从域的入口边界节点到其出口边界节点的路径（假设已作出）
that link metrics reflect propagation delays). Similar to the above metrics, the distribution of total crankback signaling and corresponding proportional delay across all TE LSPs is compared.
链路度量反映了传播延迟）。与上述指标类似，比较了所有TE LSP上的总拖转信号和相应比例延迟的分布。
o TE LSPs/Bandwidth Setup Capacity. Due to the different path computation techniques, there is a significant difference in the amount of TE LSPs/bandwidth that can be set up. This metric captures the difference in the number of TE LSPs and corresponding bandwidth that can be set up using the two path computation techniques. The traffic matrix is continuously scaled and stopped when the first TE LSP cannot be set up for both the methods. The difference in the scaling factor gives the extra bandwidth that can be set up using the corresponding path computation technique.
o TE LSPs/带宽设置容量。由于不同的路径计算技术，可以设置的TE lsp/带宽量存在显著差异。此度量捕获TE LSP数量的差异以及可使用两条路径计算技术设置的相应带宽。当两种方法都无法设置第一个TE LSP时，流量矩阵将持续缩放并停止。比例因子的差异提供了额外的带宽，可以使用相应的路径计算技术设置带宽。
o Failed TE LSPs/Bandwidth on Link Failure. Link failures are induced in the network during the course of the simulations conducted. This metric captures the number of TE LSPs and the corresponding bandwidth that failed to find a route when one or more links lying on its path failed.
o 失败的TE LSP/链路故障时的带宽。在进行模拟的过程中，网络中会出现链路故障。此度量捕获当路径上的一个或多个链路发生故障时，无法找到路由的TE LSP数量和相应带宽。
A very detailed simulator has been developed to replicate a real-life network scenario accurately. Following is the set of entities used in the simulation with a brief description of their behavior.
已经开发了一个非常详细的模拟器，以准确地复制现实生活中的网络场景。以下是模拟中使用的一组实体及其行为的简要说明。
+------------+-------+-------+--------+--------+---------+----------+
| Domain | # of | # of | OC48 | OC192 | [0,20) | [20,100] |
| Name | nodes | links | links | links | Mbps | Mbps |
+------------+-------+-------+--------+--------+---------+----------+
| D1 | 17 | 24 | 18 | 6 | 125 | 368 |
| D2 | 14 | 17 | 12 | 5 | 76 | 186 |
| D3 | 19 | 26 | 20 | 6 | 14 | 20 |
| D4 | 9 | 12 | 9 | 3 | 7 | 18 |
| MESH-CORE | 83 | 167 | 132 | 35 | 0 | 0 |
| (backbone) | | | | | | |
| SYM-CORE | 29 | 377 | 26 | 11 | 0 | 0 |
| (backbone) | | | | | | |
+------------+-------+-------+--------+--------+---------+----------+
+------------+-------+-------+--------+--------+---------+----------+
| Domain | # of | # of | OC48 | OC192 | [0,20) | [20,100] |
| Name | nodes | links | links | links | Mbps | Mbps |
+------------+-------+-------+--------+--------+---------+----------+
| D1 | 17 | 24 | 18 | 6 | 125 | 368 |
| D2 | 14 | 17 | 12 | 5 | 76 | 186 |
| D3 | 19 | 26 | 20 | 6 | 14 | 20 |
| D4 | 9 | 12 | 9 | 3 | 7 | 18 |
| MESH-CORE | 83 | 167 | 132 | 35 | 0 | 0 |
| (backbone) | | | | | | |
| SYM-CORE | 29 | 377 | 26 | 11 | 0 | 0 |
| (backbone) | | | | | | |
+------------+-------+-------+--------+--------+---------+----------+
Table 1. Domain Details and TE LSP Size Distribution
表1。域详细信息和TE LSP大小分布
o Topology Description. To obtain meaningful results applicable to present-day Service Provider topologies, simulations have been run on two representative topologies. They consists of a large backbone area to which four smaller areas are connected. For the first topology named MESH-CORE, a densely connected backbone was obtained from RocketFuel [ROCKETFUEL]. The second topology has a
o 拓扑描述。为了获得适用于当今服务提供商拓扑的有意义的结果，在两种具有代表性的拓扑上进行了仿真。它们由一个较大的主干区域组成，四个较小的区域连接到该主干区域。对于第一个名为MESH-CORE的拓扑，从RocketFuel[RocketFuel]获得了一个密集连接的主干。第二个拓扑具有
symmetrical backbone and is called SYM-CORE. The four connected smaller areas are obtained from [DEF-DES]. Details of the topologies are shown in Table 1 along with their layout in Figure 1. All TE LSPs set up on this network have their source and destinations in different areas and all of them need to traverse the backbone network. Table 1 also shows the number of TE LSPs that have their sources in the corresponding areas along with their size distribution.
对称主干，称为SYM-CORE。四个连接的较小区域从[DEF-DES]获得。拓扑的详细信息如表1所示，其布局如图1所示。在此网络上设置的所有TE LSP的源和目的地位于不同的区域，并且所有这些LSP都需要穿越主干网。表1还显示了在相应区域具有来源的TE LSP的数量及其大小分布。
o Node Behavior. Every node in the topology represents a router that maintains states for all the TE LSPs passing through it. Each node in a domain is a source for TE LSPs to all the other nodes in the other domains. As in a real-life scenario, where routers boot up at random points in time, the nodes in the topologies also start sending traffic on the TE LSPs originating from them at a random start time (to take into account the different boot-up times). All nodes are up within an hour of the start of simulation. All nodes maintain a TED that is updated using LSA updates as outlined in [RFC3630]. The flooding scope of the Traffic Engineering IGP updates are restricted only to the domain in which they originate in compliance with [RFC3630] and [RFC5305].
o 节点行为。拓扑中的每个节点都代表一个路由器，它维护通过它的所有TE LSP的状态。域中的每个节点都是其他域中所有其他节点的TE LSP的源。在现实生活中，路由器在随机时间点启动，拓扑中的节点也开始在随机启动时间（考虑不同的启动时间）在TE LSP上发送来自它们的流量。所有节点在模拟开始后一小时内启动。所有节点都维护一个TED，该TED使用[RFC3630]中概述的LSA更新进行更新。流量工程IGP更新的泛洪范围仅限于符合[RFC3630]和[RFC5305]的域。
o TE LSP Setup. When a node boots up, it sets up all TE LSPs that originate from it in descending order of size. The network is dimensioned such that all TE LSPs can find a path. Once set up, all TE LSPs stay in the network for the complete duration of the simulation unless they fail due to a link failure. Even though the TE LSPs are set up in descending order of size from a head-end router, from the network perspective, TE LSPs are set up in random fashion as the routers boot up at random times.
o TE LSP设置。当一个节点启动时，它会按大小降序设置所有源于它的TE LSP。对网络进行尺寸标注，以便所有TE LSP都能找到路径。一旦设置完毕，所有TE LSP将在整个模拟期间留在网络中，除非它们因链路故障而失败。尽管TE LSP是从前端路由器按大小降序设置的，但从网络的角度来看，TE LSP是以随机方式设置的，因为路由器在随机时间启动。
o Inducing Failures. For thorough performance analysis and comparison, link failures are induced in all the areas. Each link in a domain can fail independently with a mean failure time of 24 hours and be restored with a mean restore time of 15 minutes. Both inter-failure and inter-restore times are uniformly distributed. No attempt to re-optimize the path of a TE LSP is made when a link is restored. The links that join two domains never fail. This step has been taken to concentrate only on how link failures within domains affect the performance.
o 导致失败。为了进行全面的性能分析和比较，在所有方面都会导致链路故障。域中的每个链路都可以独立地发生故障，平均故障时间为24小时，恢复时间为15分钟。内部故障和内部恢复时间都是均匀分布的。恢复链路时，不会尝试重新优化TE LSP的路径。连接两个域的链接从未失败。这一步只关注域内的链路故障如何影响性能。
Simulations were carried out on the two topologies previously described. The results are presented and discussed in this section. All figures are from the PDF version of this document. In the figures, "PD-Setup" and "PCE-Setup" represent results corresponding
在前面描述的两种拓扑上进行了仿真。本节介绍并讨论了结果。所有数字均来自本文件的PDF版本。在图中，“PD设置”和“PCE设置”表示相应的结果
to the initial setting up of TE LSPs on an empty network using the per-domain and the PCE approach, respectively. Similarly, "PD-Failure" and "PCE-Failure" denote the results under the link failure scenario. A period of one week was simulated and results were collected after the transient period. Figure 2 and Figure 3 illustrate the behavior of the metrics for topologies MESH-CORE and SYM-CORE, respectively.
分别使用每个域和PCE方法在空网络上初始设置TE LSP。类似地，“PD故障”和“PCE故障”表示链路故障场景下的结果。模拟一周的时间，并在过渡期后收集结果。图2和图3分别说明了拓扑MESH-CORE和SYM-CORE的度量行为。
Figures 2a and 3a show the distribution of the average path cost of the TE LSPs for MESH-CORE and SYM-CORE, respectively. During the initial setup, roughly 40% of TE LSPs for MESH-CORE and 70% of TE LSPs for SYM-CORE have path costs greater with PD (PD-Setup) than with the PCE approach (PCE-Setup). This is due to the ability of the BRPC procedure to select the inter-domain shortest constrained paths that satisfy the constraints. Since the per-domain approach to path computation is undertaken in stages where every entry border router to a domain computes the path in the corresponding domain, the most optimal (shortest constrained inter-domain) route is not always found. When failures start to take place in the network, TE LSPs are re-routed over different paths resulting in path costs that are different from the initial costs. PD-Failure and PCE-Failure in Figures 2a and 3a show the distribution of the average path costs that the TE LSPs have over the duration of the simulation with link failures occurring. Similarly, the average path costs with the PD approach are much higher than the PCE approach when link failures occur. Figures 2b and 3b show similar trends and present the maximum path costs for a TE LSP for the two topologies, respectively. It can be seen that with per-domain path computation, the maximum path costs are larger for 30% and 100% of the TE LSPs for MESH-CORE and SYM-CORE, respectively.
图2a和3a分别显示了MESH-CORE和SYM-CORE的TE lsp的平均路径成本分布。在初始设置期间，大约40%的MESH-CORE TE LSP和70%的SYM-CORE TE LSP的PD（PD设置）路径成本高于PCE方法（PCE设置）。这是由于BRPC程序能够选择满足约束的域间最短约束路径。由于路径计算的每域方法是在每个域的入口边界路由器计算相应域中的路径的阶段中进行的，因此并不总是找到最优（最短约束域间）路由。当网络中开始发生故障时，TE LSP通过不同的路径重新路由，导致路径成本与初始成本不同。图2a和3a中的PD故障和PCE故障显示了TE LSP在发生链路故障的模拟期间的平均路径成本分布。类似地，当链路发生故障时，PD方法的平均路径成本远高于PCE方法。图2b和3b显示了类似的趋势，并分别给出了两种拓扑的TE LSP的最大路径成本。可以看出，对于MESH-CORE和SYM-CORE，对于每域路径计算，最大路径开销分别为TE-lsp的30%和100%。
Due to crankbacks that take place in the per-domain approach of path computation, TE LSP setup time is significantly increased. This could lead to Quality-of-Service (QoS) requirements not being met, especially during failures when re-routing needs to be quick in order to keep traffic disruption to a minimum (for example in the absence of local repair mechanisms such as defined in [RFC4090]). Since crankbacks do not take place during path computation with a PCE, setup delays are significantly reduced. Figures 2c and 3c show the distributions of the number of crankbacks that took place during the setup of the corresponding TE LSPs for MESH-CORE and SYM-CORE, respectively. It can be seen that all crankbacks occurred when failures were taking place in the networks. Figures 2d and 3d illustrate the "proportional" setup delays experienced by the TE LSPs
由于路径计算的每域方法中发生的回退，TE LSP设置时间显著增加。这可能导致服务质量（QoS）要求得不到满足，尤其是在故障期间，当需要快速重新路由以将流量中断降至最低时（例如，在缺乏[RFC4090]中定义的本地修复机制的情况下）。由于在使用PCE进行路径计算时不会发生回转，因此设置延迟会显著减少。图2c和3c分别显示了在设置MESH-CORE和SYM-CORE的相应TE LSP期间发生的回退次数分布。可以看出，当网络发生故障时，所有的回退都会发生。图2d和3d说明了TE LSP经历的“成比例”设置延迟
due to crankbacks for the two topologies. It can be observed that for a large proportion of the TE LSPs, the setup delays arising out of crankbacks are very large, possibly proving to be very detrimental to QoS requirements. The large delays arise out of the crankback signaling that needs to propagate back and forth from the exit border router of a domain to its entry border router. More crankbacks occur for SYM-CORE as compared to MESH-CORE as it is a very "restricted" and "constrained" network in terms of connectivity. This causes a lack of routes and often several cycles of crankback signaling are required to find a constrained path.
由于两种拓扑的回退。可以观察到，对于大部分TE-lsp，由回退引起的设置延迟非常大，可能证明对QoS需求非常不利。大延迟产生于需要从域的出口边界路由器来回传播到其入口边界路由器的回退信令。与MESH-CORE相比，SYM-CORE出现了更多的回退，因为它在连接方面是一个非常“受限”和“受限”的网络。这会导致缺少路由，通常需要几个回退信号周期才能找到受约束的路径。
As discussed in the previous sections, signaling failures occur either due to an outdated TED or when a path cannot be found from the selected entry border router. Figures 2e and 3e show the distribution of the total number of signaling failures experienced by the TE LSPs during setup. About 38% and 55% of TE LSPs for MESH-CORE and SYM-CORE, respectively, experience a signaling failures with per-domain path computation when link failures take place in the network. In contrast, only about 3% of the TE LSPs experience signaling failures with the PCE method. It should be noted that the signaling failures experienced with the PCE correspond only to the TEDs being out of date.
如前几节所述，由于过时的TED或无法从所选入口边界路由器中找到路径，会发生信令故障。图2e和3e显示了TE LSP在设置期间经历的信令故障总数的分布。当网络中发生链路故障时，MESH-CORE和SYM-CORE的TE LSP中，分别有38%和55%的TE LSP在计算每域路径时会遇到信令故障。相比之下，只有约3%的TE LSP使用PCE方法经历信令故障。应注意，PCE经历的信令故障仅与TED过时相对应。
Figures 2f and 3f show the number of TE LSPs and the associated required bandwidth that fail to find a route when link failures are taking place in the topologies. For MESH-CORE, with the per-domain approach, 395 TE LSPs failed to find a path corresponding to 1612 Mbps of bandwidth. For PCE, this number is lesser at 374 corresponding to 1546 Mbps of bandwidth. For SYM-CORE, with the per-domain approach, 434 TE LSPs fail to find a route corresponding to 1893 Mbps of bandwidth. With the PCE approach, only 192 TE LSPs fail to find a route, corresponding to 895 Mbps of bandwidth. It is clearly visible that the PCE allows more TE LSPs to find a route thus leading to better performance during link failures.
图2f和3f显示了当拓扑中发生链路故障时，无法找到路由的TE LSP的数量和相关的所需带宽。对于MESH-CORE，使用每域方法，395个TE LSP无法找到对应于1612 Mbps带宽的路径。对于PCE，该数字较小，为374，对应于1546 Mbps的带宽。对于SYM-CORE，使用每域方法，434个TE LSP无法找到对应于1893 Mbps带宽的路由。在PCE方法中，只有192个TE LSP找不到路由，对应于895 Mbps的带宽。显然，PCE允许更多TE LSP找到路由，从而在链路故障期间获得更好的性能。
Since PCE and the per-domain path computation approach differ in how path computation takes place, more bandwidth can be set up with PCE. This is primarily due to the way in which BRPC functions. To observe the extra bandwidth that can fit into the network, the traffic matrix was scaled. Scaling was stopped when the first TE LSP failed to set up with PCE. This metric, like all the others discussed above, is topology dependent (therefore, the choice of two topologies for this
由于PCE和每域路径计算方法在路径计算发生方式上不同，因此可以使用PCE设置更多带宽。这主要是由于BRPC的工作方式。为了观察网络中能够容纳的额外带宽，对流量矩阵进行了缩放。当第一个TE LSP无法使用PCE设置时，缩放停止。与上面讨论的所有其他度量一样，该度量依赖于拓扑（因此，为此选择两种拓扑）
study). This metric highlights the ability of PCE to fit more bandwidth in the network. For MESH-CORE, on scaling, 1556 Mbps more could be set up with PCE. In comparison, for SYM-CORE, this value is 986 Mbps. The amount of extra bandwidth that can be set up on SYM-CORE is lesser due to its restricted nature and limited capacity.
研究）。该指标强调了PCE在网络中适应更多带宽的能力。对于MESH-CORE，在扩展时，可以使用PCE设置1556 Mbps以上。相比之下，对于SYM-CORE，该值为986 Mbps。由于SYM-CORE的受限性质和容量有限，因此可以在其上设置的额外带宽较少。
This document does not raise any security issues.
本文件不涉及任何安全问题。
The authors would like to acknowledge Dimitri Papadimitriou for his helpful comments to clarify the text.
作者感谢Dimitri Papadimitriou为澄清文本所作的有益评论。
[DEF-DES] J. Guichard, F. Le Faucheur, and J.-P. Vasseur, "Definitive MPLS Network Designs", Cisco Press, 2005.
[DEF-DES]J.Guichard，F.Le Faucheur和J.-P.Vasseur，“最终MPLS网络设计”，思科出版社，2005年。
[RFC5152] Vasseur, JP., Ed., Ayyangar, A., Ed., and R. Zhang, "A Per-Domain Path Computation Method for Establishing Inter-Domain Traffic Engineering (TE) Label Switched Paths (LSPs)", RFC 5152, February 2008.
[RFC5152]Vasseur，JP.，Ed.，Ayyangar，A.，Ed.，和R.Zhang，“用于建立域间流量工程（TE）标签交换路径（LSP）的每域路径计算方法”，RFC 5152，2008年2月。
[RFC5441] Vasseur, JP., Zhang, R., Bitar, N., and JL. Le Roux, "A Backward Recursive PCE-Based Computation (BRPC) Procedure to Compute Shortest Constrained Inter-Domain Traffic Engineering Label Switched Paths", RFC 5441, April 2009.
[RFC5441]Vasseur，JP.，Zhang，R.，Bitar，N.，和JL。Le Roux，“计算最短约束域间流量工程标签交换路径的基于PCE的反向递归计算（BRPC）过程”，RFC 54412009年4月。
[RFC3630] Katz, D., Kompella, K., and D. Yeung, "Traffic Engineering (TE) Extensions to OSPF Version 2", RFC 3630, September 2003.
[RFC3630]Katz，D.，Kompella，K.，和D.Yeung，“OSPF版本2的交通工程（TE）扩展”，RFC 3630，2003年9月。
[RFC5305] Li, T. and H. Smit, "IS-IS Extensions for Traffic Engineering", RFC 5305, October 2008.
[RFC5305]Li，T.和H.Smit，“交通工程的IS-IS扩展”，RFC 5305，2008年10月。
[RFC4090] Pan, P., Ed., Swallow, G., Ed., and A. Atlas, Ed., "Fast Reroute Extensions to RSVP-TE for LSP Tunnels", RFC 4090, May 2005.
[RFC4090]Pan，P.，Ed.，Swallow，G.，Ed.，和A.Atlas，Ed.，“LSP隧道RSVP-TE快速重路由扩展”，RFC 40902005年5月。
[RFC4655] Farrel, A., Vasseur, J.-P., and J. Ash, "A Path Computation Element (PCE)-Based Architecture", RFC 4655, August 2006.
[RFC4655]Farrel，A.，Vasseur，J.-P.，和J.Ash，“基于路径计算元素（PCE）的体系结构”，RFC 46552006年8月。
[RFC4920] Farrel, A., Ed., Satyanarayana, A., Iwata, A., Fujita, N., and G. Ash, "Crankback Signaling Extensions for MPLS and GMPLS RSVP-TE", RFC 4920, July 2007.
[RFC4920]Farrel，A.，Ed.，Satyanarayana，A.，Iwata，A.，Fujita，N.，和G.Ash，“MPLS和GMPLS RSVP-TE的回退信令扩展”，RFC 4920，2007年7月。
[ROCKETFUEL] N. Spring, R. Mahajan, and D. Wehterall, "Measuring ISP Topologies with Rocketfuel", Proceedings of ACM SIGCOMM, 2002.
[火箭燃料]N.Spring，R.Mahajan和D.Wehteral，“用火箭燃料测量ISP拓扑”，ACM SIGCOMM会议录，2002年。
Authors' Addresses
作者地址
Sukrit Dasgupta Drexel University Dept of ECE, 3141 Chestnut Street Philadelphia, PA 19104 USA
Sukrit Dasgupta Drexel大学欧洲经委会系，美国宾夕法尼亚州费城切斯特纳特街3141号，邮编：19104
Phone: 215-895-1862
EMail: sukrit@ece.drexel.edu
URI: www.pages.drexel.edu/~sd88
Phone: 215-895-1862
EMail: sukrit@ece.drexel.edu
URI: www.pages.drexel.edu/~sd88
Jaudelice C. de Oliveira Drexel University Dept. of ECE, 3141 Chestnut Street Philadelphia, PA 19104 USA
Jaudelice C.de Oliveira Drexel大学欧洲经委会系，美国宾夕法尼亚州费城栗街3141号，邮编：19104
Phone: 215-895-2248
EMail: jau@ece.drexel.edu
URI: www.ece.drexel.edu/faculty/deoliveira
Phone: 215-895-2248
EMail: jau@ece.drexel.edu
URI: www.ece.drexel.edu/faculty/deoliveira
JP Vasseur Cisco Systems 1414 Massachussetts Avenue Boxborough, MA 01719 USA
JP Vasseur Cisco Systems 1414马萨诸塞州博克斯伯勒大道，马萨诸塞州01719
EMail: jpv@cisco.com
EMail: jpv@cisco.com