ITSY: Initial Trigger-Based PFC Deadlock Detection in the Data Plane
ITSY: Initial Trigger-Based PFC Deadlock Detection in the Data Plane
复制标题
DOI：
10.1109/infocomwkshps51825.2021.9484601
复制
发表时间：
2021-05
期刊：
影响因子：
--
通讯作者：
X. Wu;Eugene Ng
中科院分区：
文献类型：
--
作者：
X. Wu;Eugene Ng
无损网络在数据中心和云环境中的高性能应用中越来越受欢迎。为了在以太网中实现无损网络，采用基于优先级的流量控制（PFC）协议来保证零丢包。然而，PFC会引起网络内死锁，严重时会导致整个网络阻塞。现有的解决方案侧重于避免僵局的策略;不幸的是，它们并非万无一失。因此，死锁检测是必要的。在本文中，我们提出了ITSY，一个新的系统，正确地检测和解决死锁完全在数据平面。它不需要对网络拓扑和路由算法做任何假设。ITSY的独特之处在于使用了死锁初始触发器，这有助于有效的死锁检测和死锁复发预防。我们实现ITSY的可编程开关在P4语言。初步评估表明，ITSY能以最小的开销快速检测死锁，并有效地减少死锁的重复发生。
Lossless networks are increasingly popular for high-performance applications in data centers and cloud environments. To realize a lossless network in Ethernet, the Priority-based Flow Control (PFC) protocol is adopted to guarantee zero packet loss. PFC, however, can induce in-network deadlocks and in severe cases cause the entire network to be blocked. Existing solutions have focused on deadlock avoidance strategies; unfortunately, they are not foolproof. Therefore, deadlock detection is a necessity. In this paper, we propose ITSY, a novel system that correctly detects and solves deadlocks entirely in the data plane. It does not require any assumptions on network topologies and routing algorithms. Unique to ITSY is the use of deadlock initial triggers, which contributes to efficient deadlock detection and deadlock recurrence prevention. We implement ITSY for programmable switches in the P4 language. Preliminary evaluations demonstrate that ITSY can detect deadlocks rapidly with minimal overheads and mitigate the recurrence of the same deadlocks effectively.