当前位置:网站首页>An enhanced dynamic packet buffer management. The core part of the paper
An enhanced dynamic packet buffer management. The core part of the paper
2022-08-10 12:06:00 【is blooming】
DADT cache management strategy:
The DADT algorithm is similar to the dynamic algorithm, except that it has multiple thresholds instead of the single control threshold in the dynamic algorithm.
![]()
where αi is a proportionality constant, different for each cohort.Compared with the static thresholding scheme, the dynamic and DADT algorithm has two main advantages: 1) It can adaptively change according to the traffic conditions.Once the output queue becomes active, it starts receiving packets and its queue length increases.This increases the total buffer occupancy, and the control threshold decreases accordingly.If the current queue length is greater than the control threshold, incoming packets will be discarded until the output queue is naturally exhausted; 2) Easy to implement in hardware.The only requirements are a queue length counter and a comparator with a shift register.
The DADT algorithm has similar properties to the dynamic algorithm (i.e. the control threshold of the queue is proportional to the remaining space in the buffer), but the threshold is different for each queue.The DADT algorithm was developed specifically for packet buffers, while the dynamic algorithm was originally developed for ATM switches.ATM switches have a fixed size of incoming ATM packets.This is why all output queues behave similarly when encountering uniform traffic.Therefore, a queue cannot monopolize buffer space when encountering uniform traffic.However, for the packet buffer on the NIC, the output queue has various incoming packet sizes.
Since each queue has different packet sizes, all queues have a threshold that may allow one queue to monopolize the entire buffer space.Queues with larger incoming packet sizes tend to fill up buffer space faster.Using a single threshold for all queues results in excessive packet loss for other output queues.According to our simulation results, by controlling the threshold of each queue separately, the proposed DADT algorithm can effectively reduce the packet loss of other queues.
边栏推荐
- VSCode远程连接服务器报错:Could not establish connection to “xxxxxx”的可能错误原因及解决
- LCD驱动端与设备端名称匹配过程分析(Tiny4412)
- 英特尔推送20220809 CPU微码更新 修补Intel-SA-00657安全漏洞
- 【LeetCode】640. 求解方程
- HDU 6040 Hints of sd0061 (技巧)
- Licking Exercise - 59 From Binary Search Trees to Greater Sum Trees
- 微信小程序提交审核历史版本记录从哪里查看
- 力扣练习——59 从二叉搜索树到更大和树
- Configuration swagger
- leetcode 823. Binary Trees With Factors(因子二叉树)
猜你喜欢

机器学习之暴力调参案例

Analysis of the implementation principle of UUID from the perspective of source code

Nocalhost - Making development more efficient in the cloud-native era

Flutter气泡框实现

LeetCode50天刷题计划(Day 18—— 搜索旋转排序数组(8.50-12.00)

7、Instant-ngp

Does your child lack self-discipline?Ape Counseling: Pay attention to "blank" in the schedule to give children more control

Since the media hot style title how to write?Taught you how to write the title

再有人问你分布式事务,把这篇扔给他

three.js模糊玻璃效果
随机推荐
使用哈工大LTP测试分词并且增加自定义字典
ssm框架搭建过程[通俗易懂]
网络基础(第一节)
Flutter气泡框实现
LeetCode 138. Copy a linked list with random pointers
Analysis of the name matching process between the LCD driver and the device (Tiny4412)
力扣练习——58 验证二叉搜索树
LeetCode50天刷题计划(Day 16—— 两两交换链表中的节点(9.10-10.30)
Cannot find symbol log because lombok is not found
不止跑路,拯救误操作rm -rf /*的小伙儿
HDU 4372:Count the Buildings (Stirling数)
LeetCode 237. Delete a node in a linked list
LeetCode 82. Remove Duplicate Elements in Sorted List II
【mysql】explain介绍[通俗易懂]
态路小课堂丨如何为CXP光模块选择光纤跳线?
Servlet---解决post请求中中文乱码问题
LeetCode 25. A set of K flipped linked lists
codevs 2370 小机房的树 (LCA)
Clicking Exercise - 64 Longest Harmonic Subsequences
LeetCode 369. Plus One Linked List