Cloud data centers host diverse applications, mixing workloads that require small predictable latency with others requiring large sustained throughput. In this environment, today's state-of-the-art TCP protocol falls short. We present measurements of a 6000 server production cluster and reveal impairments that lead to high application latencies, rooted in TCP's demands on the limited buffer space available in data center switches. For example, bandwidth hungry "background" flows build up queues at the switches, and thus impact the performance of latency sensitive "foreground" traffic.To address these problems, we propose DCTCP, a TCP-like protocol for data center networks. DCTCP leverages Explicit Congestion Notification (ECN) in the network to provide multi-bit feedback to the end hosts. We evaluate DCTCP at 1 and 10Gbps speeds using commodity, shallow buffered switches. We find DCTCP delivers the same or better throughput than TCP, while using 90% less buffer space. Unlike TCP, DCTCP also provides high burst tolerance and low latency for short flows. In handling workloads derived from operational measurements, we found DCTCP enables the applications to handle 10X the current background traffic, without impacting foreground traffic. Further, a 10X increase in foreground traffic does not cause any timeouts, thus largely eliminating incast problems.

/pdf/data-center-tcp-dctcp-4d3pj8xdal.pdf

Data center TCP (DCTCP)

The Transmission Control Protocol.

Presentacio sobre l'Oficina de Proteccio de Dades Personals de la UAB i la politica Open Science. Va formar part de la conferencia "Les politiques d'Open Data / Open Acces: Implicacions a la recerca" orientada a investigadors i gestors de projectes europeus que va tenir lloc el 20 de setembre de 2018 a la Universitat Autonoma de Barcelona

/pdf/general-data-protection-regulation-2z11cloyp7.pdf

General data protection regulation

Many existing data center network (DCN) flow scheduling schemes minimize flow completion times (FCT) based on prior knowledge of flows and custom switch functions, making them superior in performance but hard to use in practice. By contrast, we seek to minimize FCT with no prior knowledge and existing commodity switch hardware.

To this end, we present PIAS, a DCN flow scheduling mechanism that aims to minimize FCT by mimicking Shortest Job First (SJF) on the premise that flow size is not known a priori. At its heart, PIAS leverages multiple priority queues available in existing commodity switches to implement a Multiple Level Feedback Queue (MLFQ), in which a PIAS flow is gradually demoted from higher-priority queues to lower-priority queues based on the number of bytes it has sent. As a result, short flows are likely to be finished in the first few high-priority queues and thus be prioritized over long flows in general, which enables PIAS to emulate SJF without knowing flow sizes beforehand.

We have implemented a PIAS prototype and evaluated PIAS through both testbed experiments and ns- 2 simulations. We show that PIAS is readily deployable with commodity switches and backward compatible with legacy TCP/IP stacks. Our evaluation results show that PIAS significantly outperforms existing information-agnostic schemes. For example, it reduces FCT by up to 50% and 40% over DCTCP [11] and L2DCT [27] respectively; and it only has a 4.9% performance gap to an ideal information-aware scheme, pFabric [13], for short flows under a production DCN workload.

/pdf/information-agnostic-flow-scheduling-for-commodity-data-1uu4bys2e7.pdf

Information-agnostic flow scheduling for commodity data centers

Traffic optimizations (TO, e.g. flow scheduling, load balancing) in datacenters are difficult online decision-making problems. Previously, they are done with heuristics relying on operators' understanding of the workload and environment. Designing and implementing proper TO algorithms thus take at least weeks. Encouraged by recent successes in applying deep reinforcement learning (DRL) techniques to solve complex online control problems, we study if DRL can be used for automatic TO without human-intervention. However, our experiments show that the latency of current DRL systems cannot handle flow-level TO at the scale of current datacenters, because short flows (which constitute the majority of traffic) are usually gone before decisions can be made. Leveraging the long-tail distribution of datacenter traffic, we develop a two-level DRL system, AuTO, mimicking the Peripheral & Central Nervous Systems in animals, to solve the scalability problem. Peripheral Systems (PS) reside on end-hosts, collect flow information, and make TO decisions locally with minimal delay for short flows. PS's decisions are informed by a Central System (CS), where global traffic information is aggregated and processed. CS further makes individual TO decisions for long flows. With CS&PS, AuTO is an end-to-end automatic TO system that can collect network information, learn from past decisions, and perform actions to achieve operator-defined goals. We implement AuTO with popular machine learning frameworks and commodity servers, and deploy it on a 32-server testbed. Compared to existing approaches, AuTO reduces the TO turn-around time from weeks to ~100 milliseconds while achieving superior performance. For example, it demonstrates up to 48.14% reduction in average flow completion time (FCT) over existing solutions.

AuTO: scaling deep reinforcement learning for datacenter-scale automatic traffic optimization

For provisioning large-scale online applications such as web search, social networks and advertisement systems, data centers face extreme challenges in providing low latency for short flows (that result from end-user actions) and high throughput for background flows (that are needed to maintain data consistency and structure across massively distributed systems). We propose L2DCT, a practical data center transport protocol that targets a reduction in flow completion times for short flows by approximating the Least Attained Service (LAS) scheduling discipline, without requiring any changes in application software or router hardware, and without adversely affecting the long flows. L2DCT can co-exist with TCP and works by adapting flow rates to the extent of network congestion inferred via Explicit Congestion Notification (ECN) marking, a feature widely supported by the installed router base. Though L2DCT is deadline unaware, our results indicate that, for typical data center traffic patterns and deadlines and over a wide range of traffic load, its deadline miss rate is consistently smaller compared to existing deadline-driven data center transport protocols. L2DCT reduces the mean flow completion time by up to 50% over DCTCP and by up to 95% over TCP. In addition, it reduces the completion for 99th percentile flows by 37% over DCTCP. We present the design and analysis of L2DCT, evaluate its performance, and discuss an implementation built upon standard Linux protocol stack.

/pdf/minimizing-flow-completion-times-in-data-centers-352oxedd6f.pdf

Minimizing flow completion times in data centers

Many data center transports have been proposed in recent times (e.g., DCTCP, PDQ, pFabric, etc). Contrary to the common perception that they are competitors (i.e., protocol A vs. protocol B), we claim that the underlying strategies used in these protocols are, in fact, complementary. Based on this insight, we design PASE, a transport framework that synthesizes existing transport strategies, namely, self-adjusting endpoints (used in TCP style protocols), innetwork prioritization (used in pFabric), and arbitration (used in PDQ). PASE is deployment friendly: it does not require any changes to the network fabric; yet, its performance is comparable to, or better than, the state-of-the-art protocols that require changes to network elements (e.g., pFabric). We evaluate PASE using simulations and testbed experiments. Our results show that PASE performs well for a wide range of application workloads and network settings.

/pdf/friends-not-foes-synthesizing-existing-transport-strategies-3hp1kij2zl.pdf

Friends, not foes: synthesizing existing transport strategies for data center networks

Load factor based congestion control schemes have shown to enhance network performance, in terms of utilization, packet loss and delay. In these schemes, using more accurate representation of network load levels is likely to lead to a more efficient way of communicating congestion information to hosts. Increasing the amount of congestion information, however, may end up adversely affecting the performance of the network. This paper focuses on this trade-off and addresses two important and challenging questions: (i) How many congestion levels should be represented by the feedback signal to provide near-optimal performance? and (ii) What window adjustment policies must be in place to ensure robustness in the face of congestion and achieve efficient and fair bandwidth allocations in high bandwidth-delay product (BDP) networks, while keeping low queues and negligible packet drop rates? Based on theoretical analysis and simulations, our results show that 3-bit feedback is sufficient for achieving near-optimal rate convergence to an efficient bandwidth allocation. While the performance gap between 2-bit and 3-bit schemes is large, gains follow the law of diminishing returns when more than 3 bits are used. Further, we show that using multiple levels for the multiplicative decrease policy enables the protocol to adjust its rate of convergence to fairness, rate variations and responsiveness to congestion based on the degree of congestion at the bottleneck. Based on these fundamental insights, we design multi-level feedback congestion control protocol (MLCP). In addition to being efficient, MLCP converges to a fair bandwidth allocation in the presence of diverse RTT flows while maintaining near-zero packet drop rate and low persistent queue length. These features coupled with MLCP's smooth rate variations make it a viable choice for many real-time applications. Using extensive packet- level simulations we show that the protocol is stable across a diverse range of network scenarios. A fluid model for the protocol shows that MLCP remains globally stable for the case of a single bottleneck link shared by identical round-trip time flows.

/pdf/on-the-design-of-load-factor-based-congestion-control-1bhma2bomq.pdf

On the Design of Load Factor based Congestion Control Protocols for Next-Generation Networks

This paper proposes a framework for congestion control, called Binary Marking Congestion Control (BMCC) for high bandwidth-delay product networks. The basic components of BMCC are i) a packet marking scheme for obtaining high resolution congestion estimates using the existing bits available in the IP header for Explicit Congestion Notification (ECN) and ii) a set of load-dependent control laws that use these congestion estimates to achieve efficient and fair bandwidth allocations on high bandwidth-delay product networks, while maintaining a low persistent queue length and negligible packet loss rate. We present analytical models that predict and provide insights into the convergence properties of the protocol. Using extensive packet-level simulations, we assess the efficacy of BMCC and perform comparisons with several proposed schemes. BMCC outperforms VCP, MLCP, XCP, SACK+RED/ECN and in some cases RCP, in terms of average flow completion times for typical Internet flow sizes.

http://netlab.caltech.edu/publications/BMCC_infocom09.pdf

Congestion Control using Efficient Explicit Feedback

Mobile devices are becoming increasingly dominant in the developing world. However, there is little insight into the characteristics of devices being used in such regions. Using a dataset of ~0.5 million subscribers from one of the largest cellular operators in Pakistan, we analyze the characteristics of cell phones based on different features (e.g., CPU, memory, and cellular interface). We identify potential device-level bottlenecks for Internet access and analyze the security implications of the phones being used. To aid the analysis of cell phones, we propose abstractions (e.g., connectivity, capacity, and device security) and cluster phones based on these abstractions. Our analysis reveals interesting insights for improving mobile web performance.

Ihsan Ayyub Qazi

Papers

Minimizing flow completion times in data centers

Friends, not foes: synthesizing existing transport strategies for data center networks

On the Design of Load Factor based Congestion Control Protocols for Next-Generation Networks

Congestion Control using Efficient Explicit Feedback

A View from the Other Side: Understanding Mobile Phone Characteristics in the Developing World