FineComb: Measuring Microscopic Latencies and Losses in the Presence of Reordering

Modern trading and cluster applications require microsecond latencies and almost no losses in data centers. This paper introduces an algorithm called FineComb that can estimate fine-grain loss and latency at monitoring devices between edges of a data center. Such a mechanism can allow managers to distinguish between latencies and loss singularities caused by servers and those caused by the network. Compared to prior work such as LDA that need to be deployed in hardware at every router and link, FineComb can be deployed today using commodity boards. The challenge for FineComb is to deal with persistent reordering that can occur in a data center network using equal cost path splitting. Without care, a loss estimation algorithm can confound loss and reordering; further, any attempt to aggregate delay estimates in the presence of reordering results in severe errors. FineComb deals with these problems using order-agnostic packet digests and a simple new idea we call stash recovery. Our evaluation demonstrates that FineComb can easily be implemented in software or hardware, and provides significantly more accurate loss and delay estimates in the presence of reordering than LDA.