sack dsack 相关介绍

　　之前做过关于 sack dsack相关笔记：tcp dsack 

其实关于sack的不好之处 也是比较明显的：在一定程度上来说，SACK可能就像DOS攻击一样，每次遍历都要消耗大量CPU，时间复杂度为O（n^2），n为packets in flight的数量。

SACK Advantages

  • Receiver informs the sender about all the PDUs that have been received successfully.
  • Sender retransmits only the PDUs that have not been Acked.
  • Implemented using two TCP options.

SACK –Permitted Option

  • First option: enabling option(2 bytes) sent in a SYN
  • Indicates SACK option can be used once connection established
    • ---Receiver may return SACK’s
    • ----Sender can process SACK’s

 SACK Rules

  • SACK does not change the meaning of ACK field.
  • SACK cannot be sent unless SACK permitted option has been received.
  • If SACKs are sent, they should be included in all TCP-PDUs when out-of-order data has been buffered.
  • First SACK must ack most recently received out-of-order PDU.

SACK Rules Receiver

  • Receiver returns as many distinct SACKs as possible.
  • SACK option is filled out by repeating most recently reported SACK blocks.
  • There may be some data in receiver’s queue which should be SACKed but is not.

SACK Rules Sender

  • Records the SACK for future reference.
  • Maintains a retransmission queue containing unacknowledged PDUs.
  • Turns SACK bit on when sender receives a SACK.
  • Skips SACK PDUs during retransmission.
  • Retransmits the PDUs not SACKed so far and less than the highest SACKed PDU.
  • Turns off SACK bit after retransmission time out.

Reneging

  • Reneging is a case when the receiver SACK some data and later discards that data. Such discarding is discouraged, but permitted if the receiver runs out of the buffer space.

If Reneging occurs

  • 　　The first SACK should reflect the newest PDU even if the PDU is going to be discarded.
  • 　　Except for the newest PDU, all SACK blocks must not Ack any data which is no longer held by receiver.

Consequences of Reneging

  • Sender must maintain normal TCP timeouts. A PDU cannot be considered acknowledged until its ACK by cumulative acknowledgement.
  • PDU’s cannot be removed from the transmit buffer until ACK by the cumulative acknowledgement.

Duplicate SACK (DSACK)Extension to SACK

  • When D-SACK is used, the first block of the SACK option should specify the duplicate PDU.
  • D-SACK block is used to report duplicate contiguous sequence of data received by the receiver in the most recent packet.
  • Each duplicate is reported at most once.
  • Allows the sender TCP to determine when a retransmission is not necessary.

D-SACK TCP Rules

  • If D-SACK block reports duplicate PDU from (possibly larger) block of data in the receiver buffer above the cumulative acknowledgement, the second SACK block (the first non D-SACK block) should specify this block.
  • As only, the first SACK block is considered as D-SACK block, if multiple sequences are duplicated, only the first is contained in the D-SACK block.

D-SACK TCP & Retransmissions

  • D-SACK allows TCP to determine when retransmission is not necessary and thereby undo congestion control measures.
  • D-SACK allows TCP to determine if the network is duplicating packets.
  • D-SACK does not allow a sender to determine if both the original and retransmitted PDU are received, or the original is lost and the retransmitted PDU is duplicated by the network.

SACK & D-SACK Interaction

  • No difference between SACK & D-SACK, except that first SACK block is used to report a duplicate PDU in D-SACK.
  • No separate negotiation/options for D-SACK.
  • D-SACK is compatible with current implementations of SACK option in TCP.

参考：

  • ftp://ftp.rfc-editor.org/in-notes/rfc2018.txt
  • ftp://ftp.rfc-editor.org/in-notes/rfc2883.txt
  • Professor Amer’s Slide
  • http://www.cse.nd.edu/~surendar/teach/spr03/cse598N/Lectures/Lecture13.pdf

以上可以在内容来自https://www.eecis.udel.edu/~amer/856/sack.04f.ppt 

引入D-SACK的目的是使TCP进行更好的流控，具体来说有以下几个好处：

1. 让发送方知道，是发送的包丢了，还是返回的ACK包丢了；

2. 网络上是否出现了包失序；

3. 数据包是否被网络上的路由器复制并转发了

4. 是不是自己的timeout太小了，导致重传

通过D-SACK这种方法，发送方可以更仔细判断出当前网络的传输情况，可以发现数据段被网络复制、错误重传、ACK丢失引起的重传、重传超时等异常的网络状。

D-SACK使用了SACK的第一个段来做标志，如何判断D-SACK：

1. 如果SACK的第一个段的范围被ACK所覆盖，那么就是D-SACK

2. 如果SACK的第一个段的范围被SACK的第二个段覆盖，那么就是D-SACK

D-SACK的规则如下：

1. 第一个block将包含重复收到的报文段的序号

2. 跟在D-SACK之后的SACK将按照SACK的方式工作

3. 如果有多个被重复接收的报文段，则D-SACK只包含其中第一个
linuxtcpip

相关

标签