QOS学习笔记

NBAR :network-based application recongnition 基于网络的应用识别

 

需要QOS的情况：
1.speed mismatch
2.many to one
3.aggregation
产生的后果：
1.delay
2.jitter
3.packet loss
 
best-effort services 尽力而为
  FIFO
integrated services 集成
  RSVP:resource reservation protocol
  发送数据前先发信令请求预留带宽，发送信令也需要带宽
  动态，需全网配置，基于每个数据流的控制，需要信令信息多，端到端的
  ！！配置参考：cisco configuring RSVP

  接口??：ip RSVP bandwitch / no sh /sh run int ??/ sh ip rsvp int /sh ip rsvp detail ----75%默认

   ip rsvp sender ? ? ?　／　sh ip rsvp sender detail 类似于ospf还需要建立邻居,来传递rsvp信令

differentiated services 区分
  静态，局部配置，基于类的控制，先分类，再加入队列，下一跳的，易于实现
 
   access-list 100 permit ip any any precedence ？ 可以看到tos 0~7的优先级，普通为0，协议为7
 
     r1(config)#access-list 100 per ip  an an precedence ?
  <0-7>           Precedence value
  critical        Match packets with critical precedence (5)
  flash           Match packets with flash precedence (3)
  flash-override  Match packets with flash override precedence (4)
  immediate       Match packets with immediate precedence (2)
  internet        Match packets with internetwork control precedence (6)
  network         Match packets with network control precedence (7)
  priority        Match packets with priority precedence (1)
  routine         Match packets with routine precedence (0)
 
  access-list 100 permit ip any any dscp ？有四种：af：保证转发assured forwarding，cs，default，ef：快速转发expedited forwarding
    r1(config)#access-list 100 per ip  an an dscp ?
  <0-63>   Differentiated services codepoint value
  af11     Match packets with AF11 dscp (001010)
  af12     Match packets with AF12 dscp (001100)
  af13     Match packets with AF13 dscp (001110)
  af21     Match packets with AF21 dscp (010010)
  af22     Match packets with AF22 dscp (010100)
  af23     Match packets with AF23 dscp (010110)
  af31     Match packets with AF31 dscp (011010)
  af32     Match packets with AF32 dscp (011100)
  af33     Match packets with AF33 dscp (011110)
  af41     Match packets with AF41 dscp (100010)
  af42     Match packets with AF42 dscp (100100)
  af43     Match packets with AF43 dscp (100110)
  cs1      Match packets with CS1(precedence 1) dscp (001000)
  cs2      Match packets with CS2(precedence 2) dscp (010000)
  cs3      Match packets with CS3(precedence 3) dscp (011000)
  cs4      Match packets with CS4(precedence 4) dscp (100000)
  cs5      Match packets with CS5(precedence 5) dscp (101000)
  cs6      Match packets with CS6(precedence 6) dscp (110000)
  cs7      Match packets with CS7(precedence 7) dscp (111000)
  default  Match packets with default dscp (000000)
  ef       Match packets with EF dscp (101110)
 
    二层的qos主要用于区分ip phone ，mls qos cos

队列机制：二层交换机无法做queuing
FIFO:cisco认为接口速率>2.048m的默认的队列类型
PQ：priority queuing ：分四类，高优先级的发完，低的才能发（只应用在路由器上）
     配置：priority-list 1 protocol ip high tcp telnet
           priority-list 1 queue-limit 1000 ----队列的限制大小
            int e0
              priority-group 1   只用在出方向！！
            show Int e0
            show queuing
            show queuing Internet e0
            debug priority
            
CQ：custom queuing 分类1~16 + 0（系统队列相当于PQ，优先发送）轮询转发
             queue-list 1 protocol ip 0 tcp telnet ----每包50B
             queue-list 1 protocol ip 1 list 100   ----每包100B
             acc 100 per icmp any any
             queue-list 1 queu 0 byte-count 100 ------默认为1500
             queue-list 1 queu 1 byte-count 100

             !!!!queue-list 1 lowest-custon 2 -----0,1队列执行PQ，2以后的队列执行CQ（默认0为PQ）
                                            0 ---则没有PQ了？？好像不对，估计0肯定是PQ（因为0是系统队列），待验证！！
              int e0
                 custom-queue-list 1

              show int e0
              show queuing
              show queuing int e0
              debug custom-queue ---ping 
              debug priority --telnet

LLQ:low latency queuing
相当于PQ与WFQ得结合
    原理：分两类，高优先级的和普通的；首先调动高优先级的队列，使用RTP协议先传；普通流量按WFQ执行
          接口下#ip rtp priority 2000 4000 100 ---2000和4000为udp端口号范围2000~6000（=2000+4000），100为保留带宽
                 show interface ？？  发现刚分配的100k已经扣除了。。。     
  ！！！预留带宽默认是75%，可以更改  接口下#max-reserved-bandwidth ？%

WFQ:weighted fair queuing
         分类机制：将ip包头中的目的ip、源ip、协议及TOS字段和后面tcp或udp的目的端口、源端口这六项运行hash算法，