Gautam Kumar
Gautam Kumar is a Software Engineer at Google working on datacenter congestion-control, network-performance and clock-synchronization. He is the recipient of SIGCOMM 2020 Best Paper Award. He received an M.S. from University from California at Berkeley and a B.Tech. from Indian Institute of Technology (IIT), Kharagpur where he received the President of India Gold medal.
Authored Publications
Sort By
Poseidon: An Efficient Congestion Control using Deployable INT for Data Center Networks
Weitao Wang
Masoud Moshref
T. S. Eugene Ng
NSDI (2023)
Preview abstract
The difficulty in gaining visibility into the fine-time scale hop-level congestion state of networks has been a key challenge faced by congestion control protocols for decades. How-ever, the emergence of commodity switches supporting in-network telemetry (INT) enables more advanced congestion control. In this paper, we presentPoseidon, a novel congestion control protocol that exploits INT to address blind spots of end-to-end algorithms and realize several fundamentally advantageous properties. Specifically, Poseidon realizes congestion control for the actual bottleneck hop. In the steady state,Poseidon realizes network-wide max-min fair bandwidth al-location. Furthermore, Poseidon decouples the bandwidth fairness requirement from the traditional AIMD control law, making it possible for Poseidon to converge fast and smooth out bandwidth oscillations. Equally important, Poseidon is de-signed to be amenable to incremental brownfield deployment in networks that mix INT and non-INT switches. Our testbed and simulation experiments show that compared to a widely-deployed state-of-the-art non-INT protocol, Swift, Poseidon improves op latency up to 10x in some percentiles (61% in average), lowers fabric RTT by more than 50%, reduces congestion window ramp up time by 40% while decreasing the throughput variation for flows with small windows by 94%.Finally, it is robust to reverse-path and multi-hop congestion.
View details
Preview abstract
Bolt is a congestion-control algorithm designed to providesingle-digit microsecond tail network-queuing at near-linerate utilization. Motivated by the need for ultra-low latencyto support applications such as NVMe, as line rates reach200G and beyond, most transfers fit within a single BDP en-tailing that transfer times predominantly become a functionof queuing and propagation delays. Bolt is an attempt topush congestion-control to its theoretical limits by harness-ing the power of programmable dataplanes such as Tofinoand Trident3+ chips. Bolt is founded on three key ideas, (i)Sub-RTT reaction (SRR): reacting to congestion faster thanRTT control-loop delay, (ii) Proactive Ramp-up (PRU): bytracking future flow-completions, and (iii) Supply matching(SM): leveraging Network Calculus concepts to maximizeutilization. Our current results achieve a 75% reduction inqueuing-delays over Swift with upto 3x improvement incompletion times for short transfers.
View details
Aequitas: Admission Control for latency-critical RPCs in Datacenters
Yiwen Zhang
Priyaranjan Jha
Mosharaf Chowdhury
SIGCOMM (2022)
Preview abstract
A modern datacenter hosts thousands of services with a mix of latency-sensitive, throughput-intensive, and best-effort traffic with high degrees of fan-out and fan-in patterns. Maintaining low tail latency under high overload conditions is difficult, especially for latency-sensitive (LS) RPCs. In this paper, we consider the challenging case of providing service-level objectives (SLO) to LS RPCs when there are unpredictable surges in demand. We present Aequitas, a distributed sender-driven admission control scheme that is anchored on the key conceptual insight: Weighted-Fair Quality of Service (QoS) queues, found in standard NICs and switches, can be used to guarantee RPC level latency SLOs by a judicious selection of QoS weights and traffic-mix across QoS queues. Aequitas installs cluster-wide RPC latency SLOs by mapping LS RPCs to higher weight QoS queues, and coping with overloads by adaptively apportioning LS RPCs amongst QoS queues based on measured completion times for each queue. When the network demand spikes unexpectedly to 25× of provisioned capacity, Aequitas achieves a latency SLO that is 3.8× lower than the state-of-art congestion control at the 99.9th-p and admits 15× more RPCs meeting SLO target compared to pFabric when RPC sizes are not aligned with priorities.
View details
Understanding Host Interconnect Congestion
Khaled Elmeleegy
Masoud Moshref
Rachit Agarwal
Saksham Agarwal
Sylvia Ratnasamy
Association for Computing Machinery, New York, NY, USA (2022), 198–204
Preview abstract
We present evidence and characterization of host congestion in production clusters: adoption of high-bandwidth access links leading to emergence of bottlenecks within the host interconnect (NIC-to-CPU data path). We demonstrate that contention on existing IO memory management units and/or the memory subsystem can significantly reduce the available NIC-to-CPU bandwidth, resulting in hundreds of microseconds of queueing delays and eventual packet drops at hosts (even when running a state-of-the-art congestion control protocol that accounts for CPU-induced host congestion). We also discuss implications of host interconnect congestion to design of future host architecture, network stacks and network protocols.
View details
PLB: Congestion Signals are Simple and Effective for Network Load Balancing
Abdul Kabbani
Junhua Yan
Kira Yin
Masoud Moshref
Mubashir Adnan Qureshi
Qiaobin Fu
Van Jacobson
SIGCOMM (2022)
Preview abstract
We describe our experience with PLB, a host-based load balancing design for modern networks. PLB randomly changes the paths of connections that experience congestion, preferring idle periods to minimize transport interactions. It does so by changing the IPv6 FlowLabel on the packets of a connection, which switches include as part of the ECMP flow hash. Across many hosts, this action drives down the number of hotspots in the network, while separating short RPCs from elephant flows to keep their completion time low.
We use PLB fleetwide in Google networks for TCP and PonyExpress (RDMA-like) traffic. We find it to be simple, general, and effective. It was easy to deploy, co-existing with other traffic, requiring only small transport modifications and little of switches, and needing no application changes. And it has produced large gains across the board, for multiple transports and from datacenter through backbone networks. After deploying PLB, the median utilization imbalance of busy switches in Google datacenter networks fell by 60\% and packet drops correspondingly fell by 33\%. At hosts, the tail latency (99$^{th}$ percentile) of short RPCs fell by up to 25\%.
View details
Swift: Delay is Simple and Effective for Congestion Control in the Datacenter
Keon Jang
Kevin Springborn
Mike Ryan
SIGCOMM 2020 (2020)
Preview abstract
We report on experiences deploying Swift congestion control in Google datacenters. Swift relies on hardware timestamps in modern NICs, and is based on AIMD control with a specified end-to-end delay target. This simple design is an evolution of earlier protocols used at Google. It has emerged as a foundation for excellent performance, when network distances are well-known, that helps to meet operational challenges. Delay is easy to decompose into fabric and host components to separate concerns, and effortless to deploy and maintain as a signal from switches in changing datacenter environments. With Swift, we obtain low flow completion times for short RPCs, even at the 99th-percentile, while providing high throughput for long RPCs. At datacenter scale, Swift achieves 50$\mu$s tail latencies for short RPCs while sustaining a 100Gbps throughput per-server, a load close to 100\%. This is much better than protocols such as DCTCP that degrade latency and loss at high utilization.
View details
Sundial: Fault-tolerant Clock Synchronization for Datacenters
Hema Hariharan
Dave Platt
Simon Sabato
Minlan Yu
Prashant Chandra
14th USENIX Symposium on Operating Systems Design and Implementation (OSDI 20), USENIX Association (2020), pp. 1171-1186
Preview abstract
Clock synchronization is critical for many datacenter applications such as distributed transactional databases, consistent snapshots, and network telemetry. As applications have increasing performance requirements and datacenter networks get into ultra-low latency, we need submicrosecond-level bound on time-uncertainty to reduce transaction delay and enable new network management applications (e.g., measuring one-way delay for congestion control). The state-of-the-art clock synchronization solutions focus on improving clock precision but may incur significant time-uncertainty bound due to the presence of failures. This significantly affects applications because in large-scale datacenters, temperature-related, link, device, and domain failures are common. We present Sundial, a fault-tolerant clock-synchronization system for datacenters that achieves ~100ns time-uncertainty bound under various types of failures. Sundial provides fast failure detection based on frequent synchronization messages in hardware. Sundial enables fast failure recovery using a novel graph-based algorithm to precompute a backup plan that is generic to failures. Through experiments in a >500-machine testbed and large-scale simulations, we show that Sundial can achieve ~100ns time-uncertainty bound under different types of failures, which is more than two orders of magnitude lower than the state-of-the-art solutions. We also demonstrate the benefit of Sundial on applications such as Spanner and Swift congestion control.
View details
Snap: a Microkernel Approach to Host Networking
Jacob Adriaens
Sean Bauer
Carlo Contavalli
Mike Dalton
William C. Evans
Nicholas Kidd
Roman Kononov
Carl Mauer
Emily Musick
Lena Olson
Mike Ryan
Erik Rubow
Kevin Springborn
Valas Valancius
In ACM SIGOPS 27th Symposium on Operating Systems Principles, ACM, New York, NY, USA (2019) (to appear)
Preview abstract
This paper presents our design and experience with a microkernel-inspired approach to host networking called Snap. Snap is a userspace networking system that supports Google’s rapidly evolving needs with flexible modules that implement a range of network functions, including edge packet switching, virtualization for our cloud platform, traffic shaping policy enforcement, and a high-performance reliable messaging and RDMA-like service. Snap has been running in production for over three years, supporting the extensible communication needs of several large and critical systems.
Snap enables fast development and deployment of new networking features, leveraging the benefits of address space isolation and the productivity of userspace software development together with support for transparently upgrading networking services without migrating applications off of a machine. At the same time, Snap achieves compelling performance through a modular architecture that promotes principled synchronization with minimal state sharing, and supports real-time scheduling with dynamic scaling of CPU resources through a novel kernel/userspace CPU scheduler co-design. Our evaluation demonstrates over 3x Gbps/core improvement compared to a kernel networking stack for RPC workloads, software-based RDMA-like performance of up to 5M IOPS/core, and transparent upgrades that are largely imperceptible to user applications. Snap is deployed to over half of our fleet of machines and supports the needs of numerous teams.
View details
