Memory-Efficient Performance Monitoring on Programmable Switches with Lean Algorithms

Abstract

Network performance problems are notoriously difficult to diagnose. Prior profiling systems collect performance statistics by keeping information about each network flow, but maintaining per-flow state is not scalable on resource-constrained NIC and switch hardware. Instead, we propose sketch-based performance monitoring using memory that is sublinear in the number of flows. Existing sketches estimate flow monitoring metrics based on flow sizes. In contrast, performance monitoring typically requires combining information across pairs of packets, such as matching a data packet with its acknowledgment to compute a round-trip time. We define a new class of lean algorithms that use memory sublinear in both the size of input data and the number of flows. We then introduce lean algorithms for a set of important statistics, such as identifying flows with high latency, loss, out-of-order, or retransmitted packets. We implement prototypes of our lean algorithms on a commodity programmable switch using the P4 language. Our experiments show that lean algorithms detect ∼82% of top 100 problematic flows among real-world packet traces using just 40KB memory.

An Exhaustive Survey on P4 Programmable Data Plane Switches: Taxonomy, Applications, Challenges, and Future Trends

“…Based on the inferred variables, Dapper can identify the root cause of the bottleneck. Similarly, the authors in [89] monitored conditions such as retransmissions, packet loss, round-trip-time, out-of-order packets to identify the top-k problematic flows. Furthermore, Blink detects failures based on the predictable behavior of TCP, which retransmits packets at epochs exponentially spaced in time, in the presence of failure.…”

“…The key idea is to have every switch maintain fine-grained telemetry data for a short period of time, and upon detecting performance degradation (e.g., increased delay), the information is offloaded to a collector. Liu et al [81] proposed a memory-efficient approach for network performance monitoring. This solution only monitors the top-k problematic flows.…”

A survey on stateful data plane in software defined networks

“…Applications that are sensitive to memory size would be affected or even infeasible. For example, load balancing [28] lapses into slower, accuracy of sketching or monitoring applications declines, and even network diagnosis [8] that relies on per-flow or per-packet monitoring would be infeasible if the number of con-nections is large [117]. Hence, most of the applications need to make tradeoffs between performance and memory usage.…”

Joltik

“…Instead, as shown in Table 2, Joltik complements these solutions by providing a generalized analytics framework that does not rely on assumptions about the sensed data and can operate in a centralized manner. Sketching for Data Analytics: Sketching algorithms for aggregate statistics have been explored in various contexts, including stream data processing [4,10,18,44,51], database [16,19,25] and network telemetry [36,43,45,46,72,75]. Perhaps the closest related work is UnivMon [45], which enables real-time general network telemetry.…”