8032v2和v1,v3对照学习笔记 - v2篇 下载本文

G.8032/Y.1344 2010/03

Summary APS

protection switching mechanisms(characteristics, architectures) ETH layer Ethernet ring topologies incorporates Amendment 1 (2010),

Appendix X (minimizing segmentation in interconnected rings)

multiple faults that could cause segmentation in a network of interconnected rings. uses additional management information and configuration of MEP Appendix XI (end-to-end service resilience) ring protection to interact with the full service protection ITU-T G.8031/Y.1342

1 Scope APS

Protection switching mechanisms for ETH layer Ethernet Ring topologies The ETH layer ring maps to the physical layer ring structure.

2 References [ITU-T G.805] [ITU-T G.806] 2009版 [ITU-T G.808.1] [ITU-T G.809] [ITU-T G.870] [ITU-T G.8010] [ITU-T G.8021] [ITU-T Y.1731] [IEEE 802.1Q]

3 Definitions

3.1 Terms defined elsewhere

和1相比,唯一的区别是G.8021中多定义了一个ETHDi/ETH adaptation function (ETHDi/ETH_A)

3.2 Terms defined in this Recommendation 3.2.1 Ethernet ring 由ring node形成的一个物理上闭合的回路,每个ring node和邻接的ring node相连。注意表述中说的是ethernet ring和ethernet ring node, 并非erp ring或erp ring

node,即讨论物理层,Ethernet层面,环的组成和连接,erp ring 和node 映射到该ring物理结构上,实际情况中,如果其中有非支持erp的ring node,应视为透明的link;由此产生的单通状态【如cc单通,或一侧link单通】按erp单通逻辑处理。 3.2.2 Ethernet ring node A network element which implements at least the following functionalities: ETH_C+ ETH_FF for forwarding APS Two ring ports, including ETHDi/ETH adaptation function ERP control process controlling the blocking and unblocking of traffic over the ring ports. 上述几个function在G.8021中定义,需要进一步学习 3.2.3 ERP instance 3.2.4 interconnection node 表述相当拗口,逻辑还是简单的,简单说互联节点,就是连接多个ring的节点;多个ring中,一个是major ring,每个互联节点有2个ring port参与,其它都是sub-ring,每个互联节点有且只有一个ring port参与. Be common to 与…共有 互联节点和共享节点不同,共享节点是多个ring共享该节点,各个ring彼此独立,每个ring,有2个ring port参与 在没有定义major ring和sub ring的时候抛出interconnection node的概念,又没有图说明,是不妥当滴。 3.2.5 major ring 物理和逻辑上都闭合的环,共享link属于major ring,每个共享节点有2个ring port参与 3.2.6 R-APS virtual channel 叙述相当稀烂;ERPS中有2种通道,traffic channel和R-APS channel,在多环的情况下,子环物理和逻辑上未闭合,v1a1中引入virtual channel的概念,让sub ring的R-APS message在主环中透传,从而实现子环逻辑上的闭合 3.2.7 ring MEL 3.2.8 ring protection link (RPL) 3.2.9 RPL neighbour node: 逻辑很容易理解,但引入neighbour不过是能 block rpl的另一端,有什么现实意义?减少无效流量侵占带宽? 3.2.10 RPL owner node Owner如何相应FS/MS请求? 3.2.11 sub-ring 3.2.12 sub-ring link Span这里如何翻译?node之间的间距? 3.2.13 wait to block timer

4 Abbreviations 新增BPR,FS,MS,ID,TCM,VPLS 5 Conventions 5.1 Representation of octets

多个8位元组表达一个2进制数,低位在前,高位在后

BIN HEX 0000 001001101100 6C 02

8位元组内部,高位在前,低位在后

6 Introduction Mechanisms, APS apply on ETH ring wide-area multipoint connectivity,moreeconomically(reduced number of links) highly reliable and stable protection,never form loops ring protection switching architecture loop avoidance at any time, traffic may flow on all but one of the ring links. Ethernet flow forwarding function (ETH_FF) this Recommendation shall be applicable for a multi-ring/ladder network, if the following principles are adhered to:

a) R-APS channels are not shared across Ethernet ring interconnections;[独立RAPS通道] b) On each ring port, each traffic channel and each R-APS channel are controlled (e.g., for blocking or flushing) by the Ethernet ring protection control process (ERP control process) of only one Ethernet ring;[各环有独立控制的ring port, tc, rc,不共享] c) Each major ring or sub-ring must have its own RPL.[独立RPL]

7 Ring protection characteristics

7.1 Monitoring methods and conditions Switching trigger condition: Condition就是SF, SF or OK,其它方式没有定义

SF trigger switching

SF defined in 8021,SF or no failed, not defined other condition, i.e. SD SF defined on transport entity of each ring link SF检测有three monitoring methods Inherent --- physical layer trail Sub-layer --- tandem connection monitoring (TCM). Test trail --- extra test trail (CCM?) 对于ERP进程而言,不知道采用的何种defects monitoring 方式,只知道传输实体(ring link)SF或OK

7.2 Ethernet traffic and bandwidth consideration 无论ERP出于何种状态,都应该能提供足够的带宽保证业务和RAPS Care should be taken that ring link capacity can continue to support all ring APS (R-APS) and service traffic that is protected after protection switching 7.3 Ethernet ring protection switching performance Single ring(switch completion time = transfer time)<50ms

without congestion all nodes in idle status

<1200km fibre circumference, <16node

single link failure

所有其它情况不保证倒换时间小于50ms

特别提到子环虚通道场景,性能部分受主环性能影响,如果节点数,距离超

过上述标准,则达不到50ms

8 Ring protection conditions and commands

Conditions,这里conditions如何翻译?条件,状态或其它?似乎状态比较合理,ring 可以出于2种状态,故障和无故障状态,从宏观上说环的状态用idle和protection表述更合适,并且ring也不止idle和protection状态,不知道这里用2种消息来表述环的状态是什么意思,当然另一种理解是,当环上检测到SF,则SF消息在环上传播遍历所有节点,恢复时NR消息传遍所有节点,这是从微观角度看环的状态

Signal fail (SF) No request (NR) Commands Forced switch (FS)[FS或MS可以同时配多处吗?假如可以同时多处配FS或MS,那么环上同时多个堵点,从实际工作上说同时配多处没有意义,但又可能不做检查允许这样做]

Manual switch (MS) Clear a) Clearing an active local administrative command (e.g., forced switch or manual switch).[clear优先级高, 假定可以同时多配FS/MS,仍然清除所有FS/MS的配置]

b) Triggering reversion before the WTR or WTB timer expires in case of revertive operation.[在WTR或WTR超时前触发恢复]

c) Triggering reversion in case of non-revertive operation. 下面3种命令没有实现

Lockout of protection – This command disables the protection group.

Replace the RPL – This command moves the RPL by blocking a different ring link and unblocking the RPL permanently.

Exercise signal – Exercise of the R-APS protocol. The signal is chosen so as not to modify the position of the blocked ring port.

9 Ring protection architectures

9.1 Revertive and non-revertive switching Switch conditions cleared(SF is removed, ms/fs is cleared) Revertive(WTR->switch over to working transport entity) Advantage: working transport entity may be optimized Disadvantage: the expense of an additional traffic interruption Non-revertive(no switch) Advantage: not the expense of an additional traffic interruption Disadvantage: may use the non-optimizedtransport entity 9.2 Protection switching triggers 3个倒换触发来源,SF,RAPS, operator control。ERP控制进程消息输入包括本地消息和外部消息,各种消息优先级不同,任意时候,只响应最高优先级的消息

a. SF, it has a higher priority b. the received R-APS message requests to switch and it has a higher priority than any otherlocal request(i.e.假定故障恢复,wtr超时,某个故障邻接node将收到owner的NRRB消息,此时如果NRRB消息优先级最高,就会触发倒换) c. initiated by operator control (e.g., forced switch, manual switch) if it has a higher prioritythan any other local request or far-end request 9.2.1 Signal fail declaration conditions

SF is defined in 8021

9.3 Protection switching models on a single Ethernet ring

举了一个4 nodes单环的例子,每个环节点抽象为应用了若干function ETHcharacteristic information (ETH_CI)即以太的报文 ETH_FF某个特定的vlan

Ethernet flow point (ETH_FP)某个端口

对图中的圆圈可以理解为某个ETH_FF,若干端口加入其中,上面承载的是ETH_CI的业务,ERP Process控制ETH_FF实现对ring端口block或unblock,flush db等 图1 正常情况 图2 SF

图3 引入CCM, MEP 图4 解释CCM如何工作

If a MEP detects a defect, which contributes to an SF defect condition, it informs the ERP control process that a failure condition has been detected.

An ERP control function uses the ETH_CI_SSF information, forwarded from the ETHx/ETH-m_A_Sk, to assert the SF condition of the ring link. R-APS protocol用R-APS message交换消息

ETHDi/ETH_A功能模块负责解析和生成R-APS message To ERP process: R-APS to ETH_CI_RAPS, To ring port: ETH_CI_RAPS to R-APS

注意:ERP process收到的为ETH_CI_RAPS除了送ERP处理,还会通过RAPS-FF向下游转发

A received R-APS message is also forwarded to the ETH_FF

9.4 Traffic channel blocking

Blocking traffic is supported by excluding the connection point from the ETH_FF functions for theone or more VLAN IDs of the traffic channel controlled by the ERP Instance

Each ERP instance shall only block orunblock the VLAN IDs of the traffic channels of the set of VLANs assigned for protection by thatERP instance. 9.5 R-APS channel blocking

R-APS channel VLAN traffic forwarding is always blocked at the same ring ports where the trafficchannel is blocked, except on sub-rings without R-APS virtual channel

除了子环w/o vc情况下,traffic channel和raps channel block不同以外,其它情况blocking或unblocking都保持一致,并且二者blocking的逻辑相同 但blocking时一些细节不同

对于traffic channel,承载的是业务数据,若在某个block的端口【ingress: 线路到端口】收到,由于该端口已经从vlan中excluded,所以不会转发,同理试图通过该节点转发的数据【egress: switch到port】也会被block,不存在进一步处理的可能 对于R-APS, 不转发收到的R-APS,但收到的R-APS送ERP处理

即使端口block,但ERP仍可通过block的端口发送自己产生的R-APS

On sub-rings without R-APS virtual channel, the R-APS channel is never blocked on any of itssub-ring nodes. However, in this case, the R-APS channel is terminated at theinterconnectionnodes. 9.6 FDB flush

removing MAC addresses

flush only the FDB for the VLAN IDs

9.7 Ethernet ring protection switching models for interconnection 6点双环,很简单

9.7.1 Ring interconnection model with an R-APS virtual channel

图9.7是一个sub ring port, 一头连子环,另一侧的network是主环,基本逻辑和单环是一样的,区别在于多了一个topology notice,即tc,子环发生倒换,通知主环动作;另外子环的RAPS会穿过主环 疑问:为什么要vc?

interconnection node 提供了子环到其它网络的连接

An R-APS virtual channel provides R-APS connectivity between thisinterconnection node and the other interconnection node of the same sub-ring, over the network

Care must be taken to ensure that the local R-APSmessages of the sub-ring being transported over the R-APS virtual channel into the interconnectednetwork can be uniquely disambiguated from those of other interconnected ring R-APS messages.

This can be achieved by, for example, using separate VIDs for the R-APS virtual channels of different sub-rings.

R-APS在主环通过不同的VID传输

主环R-APS->ETHDi/ETH_A------>ETHDi/ETH_A-----?ERP1

子环R-APS--?ETHDi/ETH_A------>R-APS_2_FF----?ETHDi/ETH_A-----?ERP2 Topology_Change signal 当子环发生倒换(捕捉条件:flush DB),子环发送该信号给主环,主环可以选择做相应动作

从图9.8上看,多环的ERP process之间只有子环tc的一个单向通讯通道,而子环的RAPS在主环上只是一个普通的vlan

are separated in ERP1 using different R-APS VIDs应该是are separated in ERPx using different R-APS VIDs

9.7.2 Ring interconnection model without R-APS virtual channel(P.25)

无虚通道的模型,R-APS在互联节点终结,如果没有vc,则子环存在一个R-APS通道隔离的问题,因为子环互联节点之间既无普通R-APS通道,也无虚通道,为了解决这个问题,子环上R-APS blocking要配为不生效,这样互联节点之间就可以通讯了。 该段关于为什么要用w/o vc,和如何用w/o vc,以及行为细节都没有说明 自我总结如下:

1. Vc需要穿通互联的其它网络域,带来额外的配置复杂性和倒换性能的不确定性问 题

2. 所有节点要配置w/o vc,而不是仅仅在互联节点上配置,这样才能保证R-APS channel的畅通,事实上在w/o vc场景下,互联节点其中一个作用是终结子环的r-aps 3. 除了r-aps不阻断,其行为类似with vc

9.7.3 Guidelines for using ring interconnection model with or without R-APS virtual Channel

a) With virtual channel, R-APS message is to tunnel(vid) between interconnect node 特点:R-APS channel blocking机制单环多环一致

优点:支持连接多主环(包括非ITU-T 和g8032的环),如下图,主环不需重新配置成子环

注意: 带宽占用,每条vc需分配不同的vid且(或)环id以区分之 Switching time受R-APS经过虚链路时间影响,比如经过时间很长 Flush 的组合 主环倒换不会相互影响,也就不会出发flush 子环的倒换可以出发主环flush

Without virtual channel, R-APS message is to terminal at interconnect node.

无R-APS message在互联节点被导入或导出,所以此种模型无需考虑带宽,VID/环id 子环倒换时间和互联环配置无关 互联环是树形拓扑无需考虑环路

缺点: R-APS blocking机制不同(见10.1.14) 两个主环通过一个子环连接,其中一个主环需要重新配置为子环,这样会引起服务中断和倒换

10 Protection control protocol Ring protection but loop avoidance

At any time traffic may flow on all but one of ring links

如果一个端口block,它可以unblock,前提条件是它知道环中至少有一个端口是block的 比如

收到NRRB消息,表示RPL端口block

收到SF消息,表明环上有break的link,进而有block的端口 10.1 Principles of operations

ERP处理的基本逻辑是从各种消息源搜集输入并比较,得到最高优先级的输入,并结合节点自身目前的状态,配置,输出相应的动作,比如flush db, 发送R-APS,block/unblock端口等:

输入包括本地请求,本地端口状态和远端R-APS中的请求

本地请求,即ExtCMD, 包括manual, force,clear命令,由于可能同时存在多个,所有cmd先送本地优先级排序逻辑LocalPriorityLogic,再送优先级逻辑二次排序 Ring口上产生的本地SF,SF clear送本地故障检查模块评估,holdoff,然后送优先级排序逻辑 远端R-APS Validity check,guard timer,priority logic Note: Guard timer在转,R-APS不送priority logic WTR timer在转,WTR Running signal插入 priority logic WTR timer过期,WTR Expires signal传到priority logic WTB类似WTR is identified as the current top priority request signaled to theR-APS request processing logic.(P29)-- 这句话表述有误,应该是说本地节点收到的RAPS消息,是远端节点的最高优先级的request

priority logic的消息来源:

a) the R-APS request/state and status information b) status and events from the WTR timer c) status and events from the WTB timer

d) status of the local Ethernet ring node's ring ports e) toppriority local request

f) the current node state from the R-APSrequest processing

R-APS request processing接受最高优先级的请求,并处理,输出和收到的最高优先级请求,配置,当前状态有关 消息:

Priority logic 配置:

MI_RAPS_RPL_Owner_Node(指示该节点是否是RPL,相应端口)

MI_RAPS_RPL_Neighbour_Node(指示该节点是否是RPL- neighbor,相应端口) 处理方式:

输入:current top priority request加local Ethernet ring node state 动作:

Transmissionof R-APS messages blocking or unblocking ring ports flushing the FDB

starting or stoppingthe timers.

flush logic根据收到的R-APS判断是否发生倒换,并由此触发flush

The topology change propagation process只存在于interconnection node上,子环拓扑改变发出Topology_Change signal通知主环处理。

interconnection flush logic从其它连接实体接收拓扑改变信息: 子环的ERP, and MI_RAPS_Propagate_TC 输出:flush ring port FDB and may trigger transmission of R-APS event requests

to both ring ports

Note: 该逻辑只存在于互联节点 backward compatibility logic 检查配置和请求

R-APS block logic(只存在于子环节点) 输入:

R-APS request processing

the top priority request from the priority logic

MI_RAPS_Sub_Ring_Without_Virtual_Channel signal. 输出:

block or unblocktraffic channel and/or the R-APS channel on ring port Backward compatibility logic模块参见10.1.13 10.1.1 Priority logic

所有输入都称作requests, 输入有多个来源(7),不同的request有不同的优先级,最高优先级的request叫top request

某一时刻有一个top request, 每当收到远端R-APS或本地request发生变化,top request要重新评估。

Clear,FS,remote fs,local sf,local sf clear,remote sf,remote ms, ms, wtr-e,wtr-r,wtb-e,wtb-r,nrrb,nr 注意:对于本地请求,如fs,sf,ms会一直在priority logic中保持,直到有更高优先级的请求或被清除

对于R-APS的request的决策是一次性的,不保存

local sf clear是被动的,仅在SF condition不存在,且在其他端口没有更高优先级的请求时存在

如果收到的消息的RING ID和自己的相同,忽略 10.1.2 R-APS request processing(p31-38) 逻辑: R-APS request processing接收当前优先级最高的request和当前节点的状态,设置共同决定,执行一系列动作,节点迁移到新的状态 输入:

最高优先级的request

本地设置,兼容性【v1->revertive mode】,owner,neighbor 输出: Guard timer start/stop Node status -> priority logic Wtr/wtb -> priority logic RAPS message transmission/stop Block/unblock ring ports& flush db[应是调用flush db logic]

初始状态: ---,初始化,分支1,动作,下一个状态e Initialization 31

IDLE Protection ManualSwitch ForcedSwitch Pending 状态 B A C D E 编号 58

Pending – clear – idle

Clear 只在owner上生效吗?首先clear是本地命令,不会有消息发出;其次,clear不只针对owner生效,一切使能fs/ms的节点都会生效;具体的动作和当前节点状态以及角色,配置有关 比如

idle,protection -> no action

ms/fs ->清block端口,起guard,发NR,如果是owner且revertive模式,起wtb… pending ->只针对owner动作 3

Idle – FS – FS Owner 不stop wtr/wtb?, idle状态wtr/wtb必然没有启动 Row 1

初始化,普通节点会block一个端口,unblock另一个端口,比较模糊,各种设备如何处理? 对于pvx 10.3,初始状态下内建的CC也不工作,所以起来时,2个端口都报SF,无法观察到真实状态 问题

1. 对于non-revertive mode,环上节点初始化后为何不存在2个阻塞点?

两个原因,node-id是sw的mac,且结合条件2,初始态下,1个端口阻塞,一个端口unblock,假定某个node 的node id最大,pending态下所有小node id阻塞点将被unblock,并停发RAPS,一轮下来,只有node id最大的点继续发NR,且阻塞点只有一个,在0号端口 2. Block one ring port/Unblock the other ring port

…Unblocks traffic channel and R-APS channel on the secondring port where the port is not unblocked, 是否是暗示port0阻塞,port1 unblock

10.1.3 R-APS message transmission【P39】

R-APS message发或不发,发什么内容,由R-APSrequest process发起,Tx R-APS动作触发,stop R-APS停止

内容包括request/state和status, 所有ring port同时发,同时停发其它当前在发的R-APS消息,但event消息除外【独占性】

如果待传输的R-APS消息发生改变,最初的3个消息将以burst的方式发送,间隔3.33ms, 以后5s发一次

Event message 是一种特殊消息,它是一次性的,且可以和其他消息并行发送,目前只定义了一种Flush db event

10.1.4 Delay timers(P40)

所谓delay timers, 包括2种WTR和WTB, 共同特征是2者都只在owner上生效,都将从pending切换到idle状态,都将会触发阻塞RPL的动作,但各自针对不同的应用场景, WTR: revertivemode of operation 需要是revertive mode, protection下,local clear SF或收到RAPS-NR触发,防止间歇性的链路故障可能导致的频繁保护倒换

WTB:before reverting to idle state (state A) after clearing operator commands (FS,MS) 2个目的

防环路,比如意外收到outdated ms/fs消息

防无意义倒换,在有多个FS的场景,当发起一个clear时,不必倒换,清除节点/owner短暂进入pending,会因为收到FS而恢复到FS状态(line 60)

注意无论何种delay timer, 都包括2种触发情况,1个是触发直接在owner上进行,第二个是接收网络消息NR触发

例如对于wtr,本质是故障恢复触发,如果恢复点在owner,则直接触发wtr,如果恢复点在其他位置,则ower还处于protection状态下通过接收nr消息触发

对于wtb,本质是发起clear触发,如果恢复点在owner,则直接触发wtr,如果恢复点在其他位置,则ower还处于fs/ms状态下通过接收nr消息触发 注意:line 36不是很清楚,貌似不可能发生啊 WTB = 5s + guard timer

Delay timer包括start,running,stop, expires 如果已经running,则新的start不会重置timer 如果stop,则停止且重置timer

Running 持续产生WTB Running signal Expires产生Expires signal

10.1.5 Guard timer

R-APS传送方式,是复制,转发,在某些情况下如果节点收到过期的R-APS,会导致错误的节点状态并可能形成loop, Guard timer的作用是用来防止节点响应过期的RAPS。 这里说的某些情况是指清本地倒换请求,local switching request包括local SF和FS/MS The guard timer is activated whenever an Ethernet ring node receives an indication that a localswitching request has cleared (i.e., local clear SF, Clear) local clear SF可能引起loop, 解释如下

SF是双向传播,在有local SF情况下,不会收到对端的SF,但当故障恢复时,相应端口上就有可能收到对端发送的SF消息【因为现在线路已经恢复】,如果没有guard time,综合考虑本端local clear/block, 收到RAPS-SF,认为远端有故障,并已block,它将打开non-failed blocked端口,远端也会如此做,从而形成了环路;

对于clear,没有一个明确的理由能解释为什么要启动guard time,在clear之前,整个环上只有自己发的FS或MS;clear之后,进入pending/block,有可能仍收到自己发的FS/MS,或其它节点发的FS,有什么必要启动guard呢?或端口状态有趋势unblock,习惯性忽略线路上的消息,保护一下

缺省500ms,10ms to 2s, step 10ms, 大小应大于raps traverse整个ring的时间

guide timer在所有节点上运行,启动后,所有收到的R-APS(除了Request/State field = \)被block

10.1.6 Validity check

检查R-APS的有效性,所谓有效性,是指接收的R-APS消息在Table 10-3中有定义 未定义的忽略

有个特殊的消息flush indication single即

Request/State field = \\

此消息是一个flush标示,不经过guard time, 直接送flush logic

Validity check模块2个端口关联的输出中一路箭头指向flush logic,即捕捉到此消息时,直接送flush logic

10.1.7 Local defect logic

两个ring port各自处理收到的ETH_CI_SSF消息,

收到的ETH_CI_SSF消息,持续产生的SF信令, holdoff延迟后,确认SF状态并报告给priority logic

ETH_CI_SSF results in producing the clear SF signal

10.1.8 Holdoff timer

这里所说的multiple layers,server layer, client layer都不是很清晰

大概的意思应该是当故障发生时,可以先等一个holdoff时间,也许该故障其他逻辑层可以修复或自愈,则不必触发倒换;等待holdoff时间后,如果故障仍存在则出发倒换,否则则不触发倒换;该故障可以是原故障或新产生的故障 0 to 10 seconds in steps of 100 ms, with an accuracy of ±5 ms, 缺省为0

10.1.9 Local priority logic 该模块有2个输入,1个输出

一个输入是来自用户的operator cmd【A】, 另一个是来自priority logic的Top priority request【B】;该模块综合评估两个输入,确定其输出;即比较二者优先级关系,如果A优先级高于B,则A作为该模块输出,称之为Top PriorityLocal Request,发送给priority logic模块 另外operator cmd是一次性的,一旦被高优先级request覆盖过,将不再继续生效

10.1.10 Flush logic 输入

从来源上区分有来自线路上的和来自内部RAPS request processing模块的block/unblock动作触发的flush请求

线路上分两类,一种是flush event,可以绕过Validitycheck直接送flush logic模块;另一种是通过Validitycheck后送flush logic模块的

内部输入由RAPS request processing模块发起,当有flush需求[Block/Unblock ring port]通知flush logic 输出

输出包括2个部分,一个是执行flush动作;另一个是发送Topology ChangePropagation消息[子环到主环]

flush logic为每个ring port保存一组数据(Node ID, BPR),这个数据在初始化的时候为 Node ID: 00:00:00:00:00:00 BPR:0

此后记录的是最近一个RAPS消息中block节点的node id和block的端口号 更新(Node ID, BPR):

每当收到的R-APS消息通过validity check,一方面送往guard, 一方面送往flush logic模块 提取(node id,BPR)并和以前的记录做比较,如果不同则更新,并触发flush

and if it is different from或和另一端口已存储的(node id,BPR)不匹配,则触发flush(除非R-APS包含DNF标志,或收到自己的node id) 两种情况删除(node id,BPR):

1. 如果收到NR消息,接收端口删除现存的(node id,BPR)且不存储NR带的该信息 2. 如果某个端口block,清除所有ring port的(node id,BPR)

10.1.11 Interconnection flush logic 输入 某个子环发生倒换,即Topology_Change[1..M]从disable toggle 为enable 输出 flush FDB action is triggered on the ring port of the target ERP instance,即如果子环倒换会触发目标ERP的ring port会触发flush【目标ERP可以是主环或子环】

同时,如果相应的MI_RAPS_Propagate_TC[1..M]enable,three R-APS \messages被触发,并在target ERP instance RAPS channel中propagation MI_RAPS_PropagaInterconnection Flush Logic应disabled

10.1.12 Topology change propagation

逻辑是如果子环发生倒换[flush db],使能TC信号,作为interconnection flush logic的输入,持续10ms

输入:子环触发的flush db

输出:Topology_Change signal

10.1.13 Backward compatibility logic

根据一个本地设置值MI_RAPS_Compatible_Version,决定ERP的工作版本,1或2,进而决定extcmd和revertive的处理方式

如果MI_RAPS_Compatible_Version为1,则 忽略FS,MS,MI_RAPS_Revertive置为1 或2, FS,MS forward, MI_RAPS_Revertive根据实际设置值

10.1.14 R-APS block logic【43】

输入: block/unblock ring ports (0/1) signal from the R-APS requestprocessing Top priority request from the priority logic MI_RAPS_Sub_Ring_Without_Virtual_Channel signal Flush FDB【未说明用途】 输出: 如果MI_RAPS_Sub_Ring_Without_Virtual_Channel signal disable,则both the traffic channel and the R-APS channelare blocked, when the block/unblock indicates the need to block a ring port MI_RAPS_Sub_Ring_Without_Virtual_Channel is enabled,则看the top priority request 是否为 local SF orlocal FS 是:depending on the value of the block/unblock ring port(0/1) signal, both the traffic channel and the R-APS channel are blocked for the appropriate ringport. 否:the traffic channel is blocked on the appropriate ring port (0/1) based on theblock/unblock ring port (0/1) signal, however the R-APS channel is not blocked

比如收到SF,MS, 此时会决策打开non-failed blocked的端口,所以此时traffic/raps channel都是unblocked

如果存在local ms,相应口traffic channel block,但r-aps channel unblock, 有何现实意义?

10.2 Protection switching behavior(43) 10.2.1 Protection switching – Link signal fail 前提:无higher priority request exists,

a. node检测到SF,故障口block traffic&RAPS channel b. 故障node2个方向持续发SF消息

c.local FDB flush(假定之前处于idle状态) 其它node收到SF消息,

d.开非故障阻塞端口,如rpl e.停发其它R-APS

f.如果消息不带DNF标志,执行flush

单向链路故障,只会引起一端节点触发保护; 节点故障的处理等同于两条链路故障

10.2.2 Protection switching – Signal degrade on link --未定义,further study

10.2.3 Protection switching – Recovery

初始态,节点有1-2个口上存在SF,1-2口上的SF清除

1. 至少保持1个口处于block状态,直到RPL block 2. 新入其它高优先级请求

Link恢复时,link两端node发送NR,并把node id作为优先级信息相互比较,收到高优先级信息的node[link的对端或ring上其它节点]unblock端口,这样的结果是ring上只有一条链路一端block

收到NR,RB或其它高优先级请求时,停发NR

10.2.3.1 Revertive behavior a. RPL收到NR触发WTR

b. WTR可能由于收到或本地产生的高等级请求而cancel c. WTR超时,RPL block,发NR,RB,flush

d. 其它节点unblock非故障口,flush(假设无DNF)

10.2.3.2 Non-revertive behaviour a.RPL owner不响应NR

b.其它健康节点也不响应NR

c. clear命令触发rpl owner block rpl,发NR,RB d.其它节点unblock非故障口,flush(假设无DNF)

10.2.4 Protection switching – Manual switch a.执行ms的端口被block b.持续发MS

c. trigger a local flush 其它节点接收到ms, d.unblock port e.停发r-aps f. flush

其它情况:

a. 如果环中已存在一个ms请求,新的ms请求将reject

b. 如果已处于ms的节点收到其它节点发送的ms,则清自己的ms,发NR,进入pending或ms状态【p34,row36】

c. 如果产生或收到更高的请求,则清ms,处理高等级请求 10.2.4.1 Manual switch – Clearing

在发起ms的node,issue clear命令清除ms

清除后端口继续block,直到RPL block或收到higher request 持续发NR,直到收到NRRB,或higher request

如果收到NR with higher node id,开端口,停发消息[什么情况会发生?]

Revertive behavior

a. Owner收NR,开始wtb,期间不接受其它ms

b. Wtb超时,产生WTB Expires signal,initiates reversion:block RPL, NRRB,flush c. 接受者unblock block的端口,flush(无dnf) Non-revertive behaviour a. Owner 收到NR不响应

b. Clear at owner,block rpl,trans nrrb,flush c. 接收者,unblock un-failed block port, flush

10.2.5 Protection switching – Forced switch [P47]

a. 目标节点对应端口关数据,RAPS通道;开其它端口 b. 持续双向发FS(假定优先级最高)

c.收端如无本地高优先级请求开blocked端口 d.收端如无本地高优先级请求停发R-APS e. 收端按需执行flush DB Note:环上可以存在多个FS,相互独立,可能导致segmentation的问题,需要使用者自行维护 10.2.5.1 Forced switch – Clearing Clear 命令,端口保持block,直到RPL block;持续发NR,直到收到NRRB; Revertive behavior Owner收到NR,进入WTB,WTB超时,block RPL,发NRRB,flush DB 其它节点收到NRRB,开blocked-no-failed端口,flush db Non-revertive behavior Owner收到NR,无动作,收到clear命令,block rpl port,发nrrb,fush db 其它节点收到NRRB,开blocked-no-failed端口,flush db

10.3 R-APS format

DA:01-19-A7-00-00-[Ring ID], SA: Bridge mac or port mac? Ethernet OAM OpCode 40

R-APS info 32 octets,用了8个,保留24个octets

首字节高4bit,request/state

首字节低4位– sub-code,对request/state进一步解释

1. 如request/state是1110 event

a. Sub-code 0000 表示flush request b. 其它值保留

2. 如request/state是其他值,sub-code传0,忽略接收

第二个字节status

RB 标示RPL是否block,1 block,0 unblock

DNF标示收到消息是否执行flush DB,1 不执行,0执行

BPR,Blocked port reference,标识那个端口block,0 端口0, 1端口1,If both ring links are blocked, the encoded value can be either value Reserved 5 bit 保留

3-8字节Node id,即节点的mac

Reserved 24 bit 保留,传0,接收忽略

10.4 Failure of protocol defect

收到的R-APS包可能会有错,比如owner收到NRRB但node id又不是自己,---配置错,环上配了多个RPL owner,有个在8021里定义的检测机制,failureof protocol – provisioning mismatch (FOP-PM)捕捉并通知设备故障管理进程。

Appendix IRing protection network objectives

No loop,keep monitoring eth layer ring kink, inform SD or SF;Server layer SD, SF inform;not contend with server layer;recovery from sing link failure;recovery sing node failure;multi-failure;operate under all network load conditions;shall be independent of the capability of the server layer; Support multi ring OAM message format

The protection process shall be deterministic Tiny BW consuming

No specific requirement for relay and filter No limitation on num. of ring node(16-255) administratively triggered switchover Revertive mode Non-revertive mode

Single node or link,switch time < 50ms Holdoff time

configurable wait-to-restore times reversion time less than 50ms

administratively triggered switchover,switch time <50ms

Appendix IIEthernet ring network objectives

描述ERPS的目标,没有太多实质性内容

物理层可以有一些保护机制,如SDH VCs 用GFP mapping,LAG,8032和其无关。

Appendix IIIRing protection scenarios Scenario A – Single link failure(p56) Failure:

A.既然RPL口block了,A会G发的NR,RB,消息吗?从图上看,会,既然RPL端口都block了,R-APS channel都block,怎么还会收到NR,RB呢?

C. flush DB,针对的对象应是所有被保护的vlan,其它vlan应不受影响。,故障口还会清掉node id, bpr

D.收到NR,清node id, bpr

F. 其它节点会收到2次故障节点发送的SF消息,近端和远端,每次收到都会触发flush FDB.

发生单断,整个环要经历C,D,E,F,G五个阶段最后稳定到G状态下 B发生单断 故障2端经历4个动作,阻塞端口(after hold time expires),发送SF,flush db,清空nodeid,bpr,跨越C,D,E三个阶段,可以理解为,先阻塞端口,后继三个动作并行完成 E阶段,所有收到SF的节点flush db,rpl owner和neighbor开阻塞端口 F阶段,SF消息继续传播,穿过原来阻塞的节点,所有节点收到SF时,将再次flush db

Reversion:

E: 断点link恢复后

故障相关节点:

进入guard timer,各自发送NR,期间不响应接收到的R-APS,端口保持阻塞 清node id ,bpr

Non-revertive恢复,在owner上执行clear命令,block rpl, send nrrb,flush fdb,其它和revertive一样

Scenario B – Single unidirectional link failure(P59) 单断,故障和恢复类似双断 单断,即收无光,只在故障节点动作,邻接节点无动作 注意步骤F, 当owner和neighbor开block端口后,所有node会第二次收到SF, 对于其他节点node id,bpr没有改变,所以不发生第二次flush DB,但对于故障节点,由于它之前发生故障时清空了node id ,bpr,此时收到自己的SF,会再次flush DB.

Scenario C – RPL failure 区别在于SF with DNF

Scenario D – Multiple failure case – Recovery NR,guard time,先开2个小端口 收到SF,block-unfailed端口打开 收到SF,Wtr停止

Guard timer后,开始处理收到的NR, 并根据NR中Node id和自己node id大小

比对,决定是否unblock端口;如果收到的node id比自己大,则开放端口,停发NR

Node id大的节点收到NRRB后,停止发NR, 开端口

其它节点: 所有节点收到NR后,清node id, bpr Owner触发WTR,wtr超时,阻塞端口,发送NRRB 所有节点收到NRRB后,flush, neighbor收到nrrb,阻塞端口

Appendix IVConsiderations for the different timers(P64)

IV.1 State machine use of timers

描述各种会用到时钟及何种时钟的场景

e.描述owner进pending状态的两种情况,不知何意 1.收到NR,此时如果是 Revertive,触发WTR,进pending 进pending,直到clear 2.WTR触发 Clear ms,sf?还有其它情况吗?

f. clear after,ms/fs, pending ->protection之前启动guard g. have a ms, receive ms, 进pending之前启动guard h.owner has a ms,收到一个ms或clear,启动wtb

IV.2 Guard timer use to block outdated R-APS messages

举了一个为什么要用guard的例子,简单归结为用guard过滤掉outdated信息 否则可能引起环路

比如断点恢复,断点可能收到过期的SF,开non-failed的端口,导致环路

Appendix VInterconnected rings example

V.1 Configuration for interconnected rings

主子环的2种组合,需要注意的是3环,即2个主环中间用一个子环连接

The Sub-Ring is composed of at least one Sub-Ring link and one R-APS Virtual Channel in order toallocate the RPL on a Sub-Ring ---不知道什么意思 V.2 Topology examples for interconnected Ethernet rings a-g多种互联拓扑模型,写cases有用(P69) a. Location of the RPL for a sub-ring

b. Intermediate Ethernet ring node(s) between interconnection nodes c. Multiple sub-rings connected to a major ring d. Sub-ring(s) interconnection

e. A sub-ring connected to multiple Ethernet rings f. A sub-ring attached to multiple major rings

g. A sub-ring connected to a network that supports any technology network

Appendix VIProtection switching for multiple ERP instances

VI.1 Multiple ERP instances

同一个物理环可以创建多个ERP实例,每个实例相互独立,并各自维护一个映射到traffic channel的VLAN的子集

目的:提高链路利用率,RPL链路可以被利用起来;

注意:各实例的VLAN子集互斥,neighbor配合保证链路带宽利用率

VI.2 Applying protection mechanisms to multiple ERP instances 如1.1各实例单独配,RPL一般要配在不同的link上

VI.2.1 Addressing of multiple ERP instances

01-19-A7-00-00-01,其中01是ring-id, 同一个物理环中,ring-id一样,不同实例用不同的R-APS vlan区分

R-APS messages of different ERP instances are differentiated by the use of different R-APS VIDs.

VI.2.2 Protection switching – Signal failure

各实例独立响应SF

VI.2.3 Protection switching – Revertive and non-revertive

各实例re,non-re可能不一致,独立处理

VI.2.4 Protection switching – Manual switch and forced switch

各实例独立处理

VI.3 Protection switching model for multiple ERP instances

所谓Protection switching model for multiple ERP instances是指,各ERP实例独立运行,但故障通告是共享的

The MEP adaptation function is de-multiplexed VI.4 Multiple instances of interconnected rings

互联环多实例vlan定义 1.vlan限制在一个环

2.vlan跨环,多环多实例 图中G2是主环

Appendix VIIGuidelines for the configuration of VIDs and Ring-IDs of R-APS channels

VII.1 Sub-ring with R-APS virtual channel

VII.1.1 Example 1: R-APS channel with different VIDs, and R-APS channel of sub-ringand R-APS virtual channel having different VIDs

主子环,不同实例各用不同的VID区分,虚通道也另用一和子环不同的VID。 缺点:需要的VID多

优点:子环的VID可重用,主环,或不共互联节点的子环【with virtual channel, 互联节点存在vlan转换,共存则产生歧义,例如sub1, 2-21, sub2 2-31, vid 2不可在同一互联节点共用】

注意:VID重用应只考虑已作为RAPS VLAN的VID,如果某VID已在某环中做RAPS-VLAN 还考虑作为Traffic channel则往往带来SPAN的问题,意义不大。

注意2:子环with virtual channel,主子环之间是有消息交互的,子环收到的RAPS消息会通过RAPS-2-FF转给主环,内容包含在ETH_C里,例如2->21, 再由主环转发。P75,参考图Figure 9-8

VII.1.2 Example 2: R-APS channel with different VIDs, and R-APS channel of sub-ringand R-APS virtual channel having the same VID

虚通道和子环RAPS channel共用一个VID 优点:管理简单

缺点:VID不能被其它环重用,总消耗的VID和example 1是一样的

VII.2 Example 3: Sub-ring without R-APS virtual channel model; each R-APS channel withdifferent VIDs

w/o虚通道的场景,主子环各实例vid,ring id独立配置,互不相干 模型参考figure 9-10, 子环erp不会借助主环erpforward raps消息 问题1:ring id有什么用 问题2:为何vid不能重用?

However, the same number of VIDs as in Example 2 is used, and cannot be reassigned. P77

VII.3 Example 4: Co-existence on an Ethernet ring of Ethernet ring nodes which support this Recommendation (v2) and the previous version (v1) of this Recommendation

v1,v2对接 v2做互联节点,ring-id必须设为1,只支持单实例,子环要用with virtual channel方式

On the other hand, when a subring without R-APS virtual channel is used, the behaviour of the blocked ring port is different between ITU-T G.8032v1 and ITU-T G.8032v2 as specified in clause 10.1.14. Therefore, when Ethernet ring nodes running ITU-T G.8032v1 and ITU-T G.8032v2 co-exist on an Ethernet ring, the sub-ring should be deployed with the R-APS virtual channel. P77

--如果子环不用neighbor port功能,则所有子环节点可以正确地响应任何protection switchingevent, 所以不存在上述限制。

Appendix VIIIFlush optimization

VIII.1 Flushing FDB consideration

发生倒换就得FLUSH,但某些情况下逻辑拓扑未变化,则不必flush,flush有潜在风暴的可能,这就是为什么考虑避免不必要的flush。

VIII.2 Scenarios of unnecessary FDB flushing a) Do not flush when RPL fails or recovers.

b) Do not flush when the RPL owner node or the RPL neighbour node fails or recovers. c) Do not flush when the currently blocked ring port fails or recovers in non-revertive mode. d) Do not flush when a request that results in blocking an already blocked ring link is issued (e.g., MS on RPL owner node).

VIII.3 Example of FDB flush optimization 引入了2个新概念

1. RPL next-neighbour node连接owner或neighbor的node

2. RPL next-neighbourportsRPL next-neighbour node和owner或neighbor的端口 Rule 1: If detecting RPL link failure in [idle state], transmit R-APS (SF, DNF).

Rule 2: When detecting a failure from an RPL next-neighbour port, in idle state, transmit R-APS (SF) message only on the RPL next-neighbour port and do not transmit R-APS messages on the other ring port.

Rule 3: If the RPL recovers, transmit R_APS (NR, RB, DNF) message from the RPL owner node

Rule 4: If the RPL owner node detects ring recovery in the R-APS (SF, DNF) condition, transmit R_APS (NR, RB, DNF) after the WTR timer expires.

VIII.4 Additional definition for ERP control process model and state machine 上述规则1,3基本ERP逻辑已实现,2,4需要修改ERP状态机 1. 加DNF的状态,供查询 2. 加next-neighbour port VIII.5 DNF status

Appendix IXGuidelines for management procedures

IX.1 An example procedure for removing an Ethernet ring node 在待移除节点的邻接节点用FS或MS

IX.2 Management procedures to exit the FS state in case of failure of an Ethernet ring node under an FS condition

描述了一种如果直接在待移除节点执行FS,移除节点,导致的异常情况,和恢复办法,邻接节点执行FS,clear

IX.3 Replacing a v1 Ethernet ring node with a v2 Ethernet ring node Owner要先升级,否则有如下问题

unidirectional failure on the non-RPL ring link attachedto the RPL owner node

Appendix XMinimizing segmentation in interconnected rings

X.1 Characterization of the segmentation issue

Dualfailures in one of the rings

不可避免,但可规避 X.1.1 Problem statement

X.1.2 Relationship to interconnection models

上述segmentation无论有否virtual channel都会存在

X.2 Class of double faults addressed

X.2.1 Detection of interconnection segmentation 用up mep

X.3 Procedure for minimization of segmentation

X.3.1 Management configuration

为了实现避免网络隔离的目标,需要额外的管理配置信息于子环ERP进程的互联节点上

– ETH_C_MI_RAPS_Interconnection_Node ? Values: \ ? Default value: \

标示是否是互联节点,pri或sec都可以,如果配为互联节点,则需要配a tandem connection,即2个互联节点之间的up mep.

– ETH_C_MI_RAPS_Multiple_Failure

? Values: \. ? Default value: \

X.3.2 Block indication logic procedure

从过程看不是很清楚

故障时比较上面设定的两个值是否一致,即ETH_C_MI_RAPS_Interconnection_Node和 ETH_C_MI_RAPS_Multiple_Failure,如果相同则执行一个MS,锁sub-ring port. 这里有2个疑问:

1. 两个参数到底代表什么含义,该怎么设,为什么他们一致就认为双断

a) 推测就是给两个节点指定一个角色,pri或sec,其实是随意指定的,某个端口侦测

到双断时,SSF消息和设定的端口角色消息进行比对,一致,则认为双断了

2. 如果双断两个端口那个端口执行MS

a) 协议中未明确说明,700中可能是先发NS比较,node id小的端口阻断?

Appendix XIEnd-to-end service resilience

XI.1 Generic end-to-end service resilience

端到端保护,可以在access link上用8031,中间用8032

XI.2 Layering ITU-T G.8031 protection over ITU-T G.8032

假定端到端保护用的是8031

XI.2.1 Basic guidelines for the layering of ITU-T G.8031 over ITU-T G.8032 Rule:

Working/protection每个路径Node2个,一进一出 ERP部分视为一条Link,具体路径由ERP决定 8031的hold off要设置足够大

Working/protection使用的vid要被ERP保护

XI.3 End-to-end service that traverses interconnected rings

规则类似单环

V2 neighbor, block RPL at both end It totally isolates RPL from the traffic channel