https://efb97021.aww-3cz.pages.dev/tags/wred/ Recent content in Wred on Aaron's Worthless Words Hugo en Wed, 03 Feb 2010 00:00:00 +0000 https://efb97021.aww-3cz.pages.dev/posts/2010/02/ont-notes-congestion-avoidance-policing-shaping-and-link-efficiency/ Wed, 03 Feb 2010 00:00:00 +0000 https://efb97021.aww-3cz.pages.dev/posts/2010/02/ont-notes-congestion-avoidance-policing-shaping-and-link-efficiency/ <ul> <li>Tail drop drawbacks <ul> <li>TCP synchronization - Dropping TCP packets from different flows can cause them all to window down and back up again at the same time in cycles.</li> <li>TCP starvation - Non-TCP or aggressive flows can starve everyone else out when TCP throttles back.</li> <li>No differentiated drop - Tail drop doesn&rsquo;t care who you are, so you get dropped if the queue is full.</li> </ul> </li> <li>RED - Random Early Detection <ul> <li>Avoids tail drop by randomly dropping packets from the queue before it gets full</li> <li>Only dropped TCP flows slow down instead of everyone who has sent a packet since the queue filled</li> <li>Queues are smaller.</li> <li>Link utilization is more efficient</li> <li>Configured with <ul> <li>Minimum threshold - start dropping when the queue is this size</li> <li>Maximum threshold - if the queue is this big, start tail dropping</li> <li>Mark probability denominator (MPD) - 1/MPD is the ratio of packets to drop when between the thresholds</li> </ul> </li> </ul> </li> <li>WRED - Weighted RED <ul> <li>Based on IP precedence or DSCP values</li> <li>Less-important packets are dropped more aggressively than important packets</li> <li>Applied to an interface, VC or a class within a policy map</li> </ul> </li> <li>CBWRED - Class based WRED <ul> <li>Configured with CBWFQ</li> </ul> </li> <li>Policing <ul> <li>Limits subrate bandwidth (give you 100kbps on a T1)</li> <li>Limits traffic of certain applications</li> <li>Any traffic that exceeds police is dropped or re-classified; it&rsquo;s a hard limit</li> <li>Inbound or outbound</li> </ul> </li> <li>Shaping <ul> <li>Sets a limit but buffers any in excess</li> <li>Requires memory to store the buffer</li> <li>Buffers = delay and/or jitter</li> <li>Outbound only</li> <li>Can respond to network signals like BECNs and FECNs</li> </ul> </li> <li>Token and bucket <ul> <li>The queue is a bucket; if a byte of data needs to be sent, it needs a token.</li> <li>If there are enough tokens, the traffic is considered conforming.</li> <li>If there aren&rsquo;t enough tokens, the traffic is considered exceeding, which triggers the drop (policing), re-classify (policing), or buffer (shaping).</li> </ul> </li> <li>Frame relay traffic shaping (FRTS) <ul> <li>Only controls frame relay traffic</li> <li>Applied on subif or DLCI</li> <li>Support fragmentation and interleaving</li> <li>Reacts to FECNs and BECNs</li> </ul> </li> <li>Compression <ul> <li>Removed redundancy and patterns in data</li> <li>Less data = less latency</li> <li>Hardware compression or hardware-assisted compression does not involve the main CPU</li> <li>Software compression does</li> <li>Payload compression</li> <li>Header compression</li> </ul> </li> <li>Link fragmentation and interleaving <ul> <li>Small data might be waiting for larger data pieces to finish sending</li> <li>Chunks data into smaller fragments so they don&rsquo;t have to wait</li> <li>Interleaving shuffles flows in the Tx queue</li> </ul> </li> </ul>