TCP Throughput Profiles Using Measurements over Dedicated Connections

Abstract

Wide-area data transfers in high-performance computing infrastructures are increasingly being carried over dynamically provisioned dedicated network connections that provide high capacities with no competing tra c. We present extensive TCP throughput measurements and time traces over a suite of physical and emulated 10 Gbps connections with 0-366 ms round-trip times (RTTs). Contrary to the general expectation, they show signi cant statistical and temporal variations, in addition to the overall dependencies on the congestion control mechanism, bu er size, and the number of parallel streams. We analyze several throughput pro les that have highly desirable concave regions wherein the throughput decreases slowly with RTTs, in stark contrast to the convex pro les predicted by various TCP analytical models. We present a generic throughput model that abstracts the ramp-up and sustainment phases of TCP ows, which provides insights into qualitative trends observed in measurements across TCP variants: (i) slow-start followed by wellsustained throughput leads to concave regions; (ii) large bu ers and multiple parallel streams expand the concave regions in addition to improving the throughput; and (iii) stable throughput dynamics, indicated by a smoother Poincaré map and smaller Lyapunov exponents, lead to wider concave regions. ese measurements and analytical results together enable us to select a TCP variant and its parameters for a given connection to achieve high throughput with statistical guarantees.

Measurements and analytics of wide-area file transfers over dedicated connections

“…They represent two contrasting profiles: they differ evidently in peak throughput values, namely, 10 and 4.5 Gbps; additionally, the shape of the former is concave in contrast to the convex shape of the latter. The peak of a throughput profile is a direct indicator of the performance, and furthermore, its shape is subtler but also an important indicator, particularly, in projecting measurements to obtain throughput estimates [9,15]. Consider projectinĝ Θ E (τ ) based on throughput measurements collected at RTTs τ 1 and τ 2 , for τ 1 < τ < τ 2 .…”

“…On the other hand, for convex profiles, the only guarantee that can be provided is that Θ E (τ ) is above the minimum ofΘ E (τ 1 ) andΘ E (τ 2 ). At a deeper level, the shape of Θ E (τ ) reflects the time dynamics of file transfers, and in particular its concavity requires fast ramp-up followed by stable throughput [15].…”

“…Overall, the current and past measurements indicate thatΘ E (τ ) is typically concave for smaller RTT and then switches to convex as RTT is increased [15][16][17][18]. Consider the memory transfer profilê Θ T (τ ) of the underlying TCP transport, which itself exhibits such a dual property, typically with a wider concave region [18].…”

“…Highperformance protocols, such as Hamilton TCP [21] and Scalable TCP [8], provide wider concave regions compared to other TCP variants. Furthermore, increasing configuration parameters, such as the buffer size and number of parallel streams, leads to expanded concave regions [15], which represents a more effective transport method. The file transfer profileΘ E (τ ) is a modulated version of Θ T (τ ) by file, IO and host systems, typically resulting in smaller concave regions, as shown subsequently.…”

“…In this paper, we show the dependencies ofΘ E (τ ) on both network and file IO parameters, and highlight the differences among various file transfer methods, and then propose systematic methods to compare them. We mainly focus on analytics of measurements under various configurations that have described in detail in [14][15][16][17]19], which we briefly summarize here in various sections to provide the context and illustrate specific performance parameters.…”