arXiv subjects

Ankit Singla

Publications and source records attributed to Ankit Singla.

At least 19 recordsLinked to original sources

arXiv2022

New primitives for bounded degradation in network service

Certain new ascendant data center workloads can absorb some degradation in network service, not needing fully reliable data transport and/or their fair-share of network bandwidth. This opens up opportunities for superior network and infrastructure multiplexing by having this flexible traffic cede capacity under congestion to regular traffic with stricter needs. We posit there is opportunity in network service primitives which permit degradation within certain bounds, such that flexible traffic still receives an acceptable level of service, while benefiting from its weaker requirements. We propose two primitives, namely guaranteed partial delivery and bounded deprioritization. We design a budgeting algorithm to provide guarantees relative to their fair share, which is measured via probing. The requirement of budgeting and probing limits the algorithm's applicability to large flexible flows. We evaluate our algorithm with large flexible flows and for three workloads of regular flows of small size, large size and a distribution of sizes. Across the workloads, our algorithm achieves less speed-up of regular flows than fixed prioritization, especially for the small flows workload (1.25x vs. 6.82 in the 99th %-tile). Our algorithm provides better guarantees in the workload with large regular flows (with 14.5% vs. 32.5% of flexible flows being slowed down beyond their guarantee). However, it provides not much better or even slightly worse guarantees for the other two workloads. The ability to enforce guarantees is influenced by flow fair share interdependence, measurement inaccuracies and dependency on convergence. We observe that priority changes to probe or to deprioritize causes queue shifts which deteriorate guarantees and limit possible speed-up, especially of small flows. We find that mechanisms to both prioritize traffic and track guarantees should be as non-disruptive as possible.

cs.NI↗

arXiv2022

Prepare your video for streaming with Segue

We identify new opportunities in video streaming, involving the joint consideration of offline video chunking and online rate adaptation. Due to a video's complexity varying over time, certain parts are more likely to cause performance impairments during playback with a particular rate adaptation algorithm. To address such an issue, we propose Segue, which carefully uses variable-length video segments, and augment specific segments with additional bitrate tracks. The key novelty of our approach is in making such decisions based on the video's time-varying complexity and the expected rate adaptation behavior over time. We propose and implement several methods for such adaptation-aware chunking. Our results show that Segue substantially reduces rebuffering and quality fluctuations, while maintaining video quality delivered; Segue improves QoE by 9% on average, and by 22% in low-bandwidth conditions. Finally, we view our problem framing as a first step in a new thread on algorithmic and design innovation in video streaming, and leave the reader with several interesting open questions.

cs.NI↗

arXiv2021

Order P4-66: Characterizing and mitigating surreptitious programmable network device exploitation

Substantial efforts are invested in improving network security, but the threat landscape is rapidly evolving, particularly with the recent interest in programmable network hardware. We explore a new security threat, from an attacker who has gained control of such devices. While it should be obvious that such attackers can trivially cause substantial damage, the challenge and novelty are in doing so while preventing quick diagnosis by the operator. We find that compromised programmable devices can easily degrade networked applications by orders of magnitude, while evading diagnosis by even the most sophisticated network diagnosis methods in deployment. Two key observations yield this result: (a) targeting a small number of packets is often enough to cause disproportionate performance degradation; and (b) new programmable hardware is an effective enabler of careful, selective targeting of packets. Our results also point to recommendations for minimizing the damage from such attacks, ranging from known, easy to implement techniques like encryption and redundant requests, to more complex considerations that would potentially limit some intended uses of programmable hardware. For data center contexts, we also discuss application-aware monitoring and response as a potential mitigation.

cs.CR↗

arXiv2021

Towards Demystifying Serverless Machine Learning Training

The appeal of serverless (FaaS) has triggered a growing interest on how to use it in data-intensive applications such as ETL, query processing, or machine learning (ML). Several systems exist for training large-scale ML models on top of serverless infrastructures (e.g., AWS Lambda) but with inconclusive results in terms of their performance and relative advantage over "serverful" infrastructures (IaaS). In this paper we present a systematic, comparative study of distributed ML training over FaaS and IaaS. We present a design space covering design choices such as optimization algorithms and synchronization protocols, and implement a platform, LambdaML, that enables a fair comparison between FaaS and IaaS. We present experimental results using LambdaML, and further develop an analytic model to capture cost/performance tradeoffs that must be considered when opting for a serverless infrastructure. Our results indicate that ML training pays off in serverless only for models with efficient (i.e., reduced) communication and that quickly converge. In general, FaaS can be much faster but it is never significantly cheaper than IaaS.

cs.DC↗

arXiv2020

Performance-Driven Internet Path Selection

Internet routing can often be sub-optimal, with the chosen routes providing worse performance than other available policy-compliant routes. This stems from the lack of visibility into route performance at the network layer. While this is an old problem, we argue that recent advances in programmable hardware finally open up the possibility of performance-aware routing in a deployable, BGP-compatible manner. We introduce ROUTESCOUT, a hybrid hardware/software system supporting performance-based routing at ISP scale. In the data plane, ROUTESCOUT leverages P4-enabled hardware to monitor performance across policy-compliant route choices for each destination, at line-rate and with a small memory footprint. ROUTESCOUT's control plane then asynchronously pulls aggregated performance metrics to synthesize a performance-aware forwarding policy. We show that ROUTESCOUT can monitor performance across most of an ISP's traffic, using only 4 MB of memory. Further, its control can flexibly satisfy a variety of operator objectives, with sub-second operating times.

cs.NI↗

arXiv2020

FatPaths: Routing in Supercomputers and Data Centers when Shortest Paths Fall Short

We introduce FatPaths: a simple, generic, and robust routing architecture that enables state-of-the-art low-diameter topologies such as Slim Fly to achieve unprecedented performance. FatPaths targets Ethernet stacks in both HPC supercomputers as well as cloud data centers and clusters. FatPaths exposes and exploits the rich ("fat") diversity of both minimal and non-minimal paths for high-performance multi-pathing. Moreover, FatPaths uses a redesigned "purified" transport layer that removes virtually all TCP performance issues (e.g., the slow start), and incorporates flowlet switching, a technique used to prevent packet reordering in TCP networks, to enable very simple and effective load balancing. Our design enables recent low-diameter topologies to outperform powerful Clos designs, achieving 15% higher net throughput at 2x lower latency for comparable cost. FatPaths will significantly accelerate Ethernet clusters that form more than 50% of the Top500 list and it may become a standard routing scheme for modern topologies.

cs.NI↗

arXiv2020

High-Performance Routing with Multipathing and Path Diversity in Ethernet and HPC Networks

The recent line of research into topology design focuses on lowering network diameter. Many low-diameter topologies such as Slim Fly or Jellyfish that substantially reduce cost, power consumption, and latency have been proposed. A key challenge in realizing the benefits of these topologies is routing. On one hand, these networks provide shorter path lengths than established topologies such as Clos or torus, leading to performance improvements. On the other hand, the number of shortest paths between each pair of endpoints is much smaller than in Clos, but there is a large number of non-minimal paths between router pairs. This hampers or even makes it impossible to use established multipath routing schemes such as ECMP. In this work, to facilitate high-performance routing in modern networks, we analyze existing routing protocols and architectures, focusing on how well they exploit the diversity of minimal and non-minimal paths. We first develop a taxonomy of different forms of support for multipathing and overall path diversity. Then, we analyze how existing routing schemes support this diversity. Among others, we consider multipathing with both shortest and non-shortest paths, support for disjoint paths, or enabling adaptivity. To address the ongoing convergence of HPC and "Big Data" domains, we consider routing protocols developed for both HPC systems and for data centers as well as general clusters. Thus, we cover architectures and protocols based on Ethernet, InfiniBand, and other HPC networks such as Myrinet. Our review will foster developing future high-performance multipathing routing protocols in supercomputers and data centers.

cs.NI↗

arXiv2020

Understanding video streaming algorithms in the wild

While video streaming algorithms are a hot research area, with interesting new approaches proposed every few months, little is known about the behavior of the streaming algorithms deployed across large online streaming platforms that account for a substantial fraction of Internet traffic. We thus study adaptive bitrate streaming algorithms in use at 10 such video platforms with diverse target audiences. We collect traces of each video player's response to controlled variations in network bandwidth, and examine the algorithmic behavior: how risk averse is an algorithm in terms of target buffer; how long does it takes to reach a stable state after startup; how reactive is it in attempting to match bandwidth versus operating stably; how efficiently does it use the available network bandwidth; etc. We find that deployed algorithms exhibit a wide spectrum of behaviors across these axes, indicating the lack of a consensus one-size-fits-all solution. We also find evidence that most deployed algorithms are tuned towards stable behavior rather than fast adaptation to bandwidth variations, some are tuned towards a visual perception metric rather than a bitrate-based metric, and many leave a surprisingly large amount of the available bandwidth unused.

cs.NI↗

arXiv2019

Measuring and exploiting the cloud consolidation of the Web

We present measurements showing that the top one million most popular Web domains are reachable within 13ms (in the median) from a collective of just 12 cloud data centers. We explore the consequences of this Web "consolidation", focusing on its potential for speeding up the evolution of the Web. That most popular services reside in or near a small number of data centers implies that new application and transport technologies can be rapidly deployed for these Web services, without the involvement of their operators. We show how this may be achieved by orchestrating a handful of reverse proxies deployed in the same data centers, with new technologies deployed at these proxies being nearly as effective as deploying them directly to the Web servers. We present early measurements of this approach, demonstrating a >50% reduction in Web page load times for users with high latencies to Web servers. We also show that this model, using a small number of proxies, can be surprisingly competitive with extensive CDN deployments, especially in geographies with high last-mile latencies.

cs.NI↗

arXiv2019

Distributed Learning over Unreliable Networks

Most of today's distributed machine learning systems assume {\em reliable networks}: whenever two machines exchange information (e.g., gradients or models), the network should guarantee the delivery of the message. At the same time, recent work exhibits the impressive tolerance of machine learning algorithms to errors or noise arising from relaxed communication or synchronization. In this paper, we connect these two trends, and consider the following question: {\em Can we design machine learning systems that are tolerant to network unreliability during training?} With this motivation, we focus on a theoretical problem of independent interest---given a standard distributed parameter server architecture, if every communication between the worker and the server has a non-zero probability $p$ of being dropped, does there exist an algorithm that still converges, and at what speed? The technical contribution of this paper is a novel theoretical analysis proving that distributed learning over unreliable network can achieve comparable convergence rate to centralized or distributed learning over reliable networks. Further, we prove that the influence of the packet drop rate diminishes with the growth of the number of \textcolor{black}{parameter servers}. We map this theoretical result onto a real-world scenario, training deep neural networks over an unreliable network layer, and conduct network simulation to validate the system improvement by allowing the networks to be unreliable.

cs.DC↗

arXiv2018

Dissecting Latency in the Internet's Fiber Infrastructure

The recent publication of the `InterTubes' map of long-haul fiber-optic cables in the contiguous United States invites an exciting question: how much faster would the Internet be if routes were chosen to minimize latency? Previous measurement campaigns suggest the following rule of thumb for estimating Internet latency: multiply line-of-sight distance by 2.1, then divide by the speed of light in fiber. But a simple computation of shortest-path lengths through the conduits in the InterTubes map suggests that the conversion factor for all pairs of the 120 largest population centers in the U.S.\ could be reduced from 2.1 to 1.3, in the median, even using less than half of the links. To determine whether an overlay network could be used to provide shortest paths, and how well it would perform, we used the diverse server deployment of a CDN to measure latency across individual conduits. We were surprised to find, however, that latencies are sometimes much higher than would be predicted by conduit length alone. To understand why, we report findings from our analysis of network latency data from the backbones of two Tier-1 ISPs, two scientific and research networks, and the recently built fiber backbone of a CDN.

cs.NI↗

arXiv2018

Enabling Efficient RDMA-based Synchronous Mirroring of Persistent Memory Transactions

Synchronous Mirroring (SM) is a standard approach to building highly-available and fault-tolerant enterprise storage systems. SM ensures strong data consistency by maintaining multiple exact data replicas and synchronously propagating every update to all of them. Such strong consistency provides fault tolerance guarantees and a simple programming model coveted by enterprise system designers. For current storage devices, SM comes at modest performance overheads. This is because performing both local and remote updates simultaneously is only marginally slower than performing just local updates, due to the relatively slow performance of accesses to storage in today's systems. However, emerging persistent memory and ultra-low-latency network technologies necessitate a careful re-evaluation of the existing SM techniques, as these technologies present fundamentally different latency characteristics compared than their traditional counterparts. In addition to that, existing low-latency network technologies, such as Remote Direct Memory Access (RDMA), provide limited ordering guarantees and do not provide durability guarantees necessary for SM. To evaluate the performance implications of RDMA-based SM, we develop a rigorous testing framework that is based on emulated persistent memory. Our testing framework makes use of two different tools: (i) a configurable microbenchmark and (ii) a modified version of the WHISPER benchmark suite, which comprises a set of common cloud applications. Using this framework, we find that recently proposed RDMA primitives, such as remote commit, provide correctness guarantees, but do not take full advantage of the asynchronous nature of RDMA hardware. To this end, we propose new primitives enabling efficient and correct SM over RDMA, and use these primitives to develop two new techniques delivering high-performance SM of persistent memories.

cs.DC↗

arXiv2018

cISP: A Speed-of-Light Internet Service Provider

Low latency is a requirement for a variety of interactive network applications. The Internet, however, is not optimized for latency. We thus explore the design of cost-effective wide-area networks that move data over paths very close to great-circle paths, at speeds very close to the speed of light in vacuum. Our cISP design augments the Internet's fiber with free-space wireless connectivity. cISP addresses the fundamental challenge of simultaneously providing low latency and scalable bandwidth, while accounting for numerous practical factors ranging from transmission tower availability to packet queuing. We show that instantiations of cISP across the contiguous United States and Europe would achieve mean latencies within 5% of that achievable using great-circle paths at the speed of light, over medium and long distances. Further, we estimate that the economic value from such networks would substantially exceed their expense.

cs.NI↗

arXiv2016

Measuring and Understanding Throughput of Network Topologies

High throughput is of particular interest in data center and HPC networks. Although myriad network topologies have been proposed, a broad head-to-head comparison across topologies and across traffic patterns is absent, and the right way to compare worst-case throughput performance is a subtle problem. In this paper, we develop a framework to benchmark the throughput of network topologies, using a two-pronged approach. First, we study performance on a variety of synthetic and experimentally-measured traffic matrices (TMs). Second, we show how to measure worst-case throughput by generating a near-worst-case TM for any given topology. We apply the framework to study the performance of these TMs in a wide range of network topologies, revealing insights into the performance of topologies with scaling, robustness of performance across TMs, and the effect of scattered workload placement. Our evaluation code is freely available.

cs.NI↗

arXiv2015

Towards a Speed of Light Internet

In principle, a network can transfer data at nearly the speed of light. Today's Internet, however, is much slower: our measurements show that latencies are typically more than one, and often more than two orders of magnitude larger than the lower bound implied by the speed of light. Closing this gap would not only add value to today's Internet applications, but might also open the door to exciting new applications. Thus, we propose a grand challenge for the networking research community: building a speed-of-light Internet. Towards addressing this goal, we begin by investigating the causes of latency inflation in the Internet across the network stack. Our analysis reveals that while protocol overheads, which have dominated the community's attention, are indeed important, infrastructural inefficiencies are a significant and under-explored problem. Thus, we propose a radical, yet surprisingly low-cost approach to mitigating latency inflation at the lowest layers and building a nearly speed-of-light Internet infrastructure.

cs.NI↗

arXiv2014

High Throughput Data Center Topology Design

With high throughput networks acquiring a crucial role in supporting data-intensive applications, a variety of data center network topologies have been proposed to achieve high capacity at low cost. While this literature explores a large number of design points, even in the limited case of a network of identical switches, no proposal has been able to claim any notion of optimality. The case of heterogeneous networks, incorporating multiple line-speeds and port-counts as data centers grow over time, introduces even greater complexity. In this paper, we present the first non-trivial upper-bound on network throughput under uniform traffic patterns for any topology with identical switches. We then show that random graphs achieve throughput surprisingly close to this bound, within a few percent at the scale of a few thousand servers. Apart from demonstrating that homogeneous topology design may be reaching its limits, this result also motivates our use of random graphs as building blocks to explore the design of heterogeneous networks. Given a heterogeneous pool of network switches, through experiments and analysis, we explore how the distribution of servers across switches and the interconnection of switches affect network throughput. We apply these insights to a real-world heterogeneous data center topology, VL2, demonstrating as much as 43% higher throughput with the same equipment.

cs.NI↗

arXiv2013

Scalable Routing on Flat Names

We introduce a protocol which routes on flat, location-independent identifiers with guaranteed scalability and low stretch. Our design builds on theoretical advances in the area of compact routing, and is the first to realize these guarantees in a dynamic distributed setting.

cs.NI↗

arXiv2012

On the Resilience of Routing Tables

Many modern network designs incorporate "failover" paths into routers' forwarding tables. We initiate the theoretical study of the conditions under which such resilient routing tables can guarantee delivery of packets.

cs.DC↗