History of Congestion Due to Link Over-utilization
Congestion in networks is as old as the networks themselves. It’s been a topic of interest within the research community since the early days. It is not surprising to observe congestion when a network port is expected to send more traffic than its capacity. 网络拥塞与网络本身一样古老。从早期开始，它就一直是研究界关注的话题。当网络端口预计发送的流量超过其容量时，出现拥塞就不足为奇了。
But in the production storage networks, this issue was not widely seen until the middle of the 2010 decade (around the year 2015). This is when the storage industry went through a historic transformation because of all-flash technology. When All-Flash Arrays were deployed in production storage networks, the storage links were utilized at their full capacity, a condition that was rare earlier. By this time, Fibre Channel had undergone multiple speed upgrades of 1GFC, 2GFC, 4GFC, 8GFC, and 16GFC. Most people followed the years of best practices that they had used while upgrading through many speed generations. But this time they saw a phenomenon that they had not seen earlier. 但在生产存储网络中，这一问题直到 2010 年代中期（2015 年左右）才被广泛关注。此时，由于全闪存技术的出现，存储行业经历了一次历史性的变革。在生产存储网络中部署全闪存阵列后，存储链路被充分利用，而这种情况在以前是很少见的。此时，光纤通道已经经历了 1GFC、2GFC、4GFC、8GFC 和 16GFC 的多次速度升级。大多数人都沿用了多年来的最佳实践，进行了多代速度升级。但这一次，他们看到了以前从未见过的现象。
In the 2015-2016 timeframe, understanding and finding the reason for this type of congestion took a while. Although the effect on the fabric due to over-utilization of the edge links is similar to the congestion caused by a slow-drain device, both the detection of it and the reasons for it are different. Hence, both det