Networks are a shared resource connecting critical IT infrastructure, and the general practice is to always leave them on. Yet, meaningful energy savings can result from improving a network's ability to scale up and down, as traffic demands ebb and flow. We present ElasticTree, a network-wide power1 manager, which dynamically adjusts the set of active network elements -- links and switches--to satisfy changing data center traffic loads.We first compare multiple strategies for finding minimum-power network subsets across a range of traffic patterns. We implement and analyze ElasticTree on a prototype testbed built with production OpenFlow switches from three network vendors. Further, we examine the trade-offs between energy efficiency, performance and robustness, with real traces from a production e-commerce website. Our results demonstrate that for data center workloads, ElasticTree can save up to 50% of network energy, while maintaining the ability to handle traffic surges. Our fast heuristic for computing network subsets enables ElasticTree to scale to data centers containing thousands of nodes. We finish by showing how a network admin might configure ElasticTree to satisfy their needs for performance and fault tolerance, while minimizing their network power bill.

/pdf/elastictree-saving-energy-in-data-center-networks-2w53iv2yv5.pdf

ElasticTree: saving energy in data center networks

Client-side video players employ adaptive bitrate (ABR) algorithms to optimize user quality of experience (QoE). Despite the abundance of recently proposed schemes, state-of-the-art ABR algorithms suffer from a key limitation: they use fixed control rules based on simplified or inaccurate models of the deployment environment. As a result, existing schemes inevitably fail to achieve optimal performance across a broad set of network conditions and QoE objectives.We propose Pensieve, a system that generates ABR algorithms using reinforcement learning (RL). Pensieve trains a neural network model that selects bitrates for future video chunks based on observations collected by client video players. Pensieve does not rely on pre-programmed models or assumptions about the environment. Instead, it learns to make ABR decisions solely through observations of the resulting performance of past decisions. As a result, Pensieve automatically learns ABR algorithms that adapt to a wide range of environments and QoE metrics. We compare Pensieve to state-of-the-art ABR algorithms using trace-driven and real world experiments spanning a wide variety of network conditions, QoE metrics, and video properties. In all considered scenarios, Pensieve outperforms the best state-of-the-art scheme, with improvements in average QoE of 12%--25%. Pensieve also generalizes well, outperforming existing schemes even on networks for which it was not explicitly trained.

/pdf/neural-adaptive-video-streaming-with-pensieve-3y10qvjqj1.pdf

Neural Adaptive Video Streaming with Pensieve

Existing ABR algorithms face a significant challenge in estimating future capacity: capacity can vary widely over time, a phenomenon commonly observed in commercial services. In this work, we suggest an alternative approach: rather than presuming that capacity estimation is required, it is perhaps better to begin by using only the buffer, and then ask when capacity estimation is needed. We test the viability of this approach through a series of experiments spanning millions of real users in a commercial service. We start with a simple design which directly chooses the video rate based on the current buffer occupancy. Our own investigation reveals that capacity estimation is unnecessary in steady state; however using simple capacity estimation (based on immediate past throughput) is important during the startup phase, when the buffer itself is growing from empty. This approach allows us to reduce the rebuffer rate by 10-20% compared to Netflix's then-default ABR algorithm, while delivering a similar average video rate, and a higher video rate in steady state.

/pdf/a-buffer-based-approach-to-rate-adaptation-evidence-from-a-42j7jpd30s.pdf

A buffer-based approach to rate adaptation: evidence from a large video streaming service

Modern enterprises almost ubiquitously deploy middlebox processing services to improve security and performance in their networks. Despite this, we find that today's middlebox infrastructure is expensive, complex to manage, and creates new failure modes for the networks that use them. Given the promise of cloud computing to decrease costs, ease management, and provide elasticity and fault-tolerance, we argue that middlebox processing can benefit from outsourcing the cloud. Arriving at a feasible implementation, however, is challenging due to the need to achieve functional equivalence with traditional middlebox deployments without sacrificing performance or increasing network complexity.In this paper, we motivate, design, and implement APLOMB, a practical service for outsourcing enterprise middlebox processing to the cloud.Our discussion of APLOMB is data-driven, guided by a survey of 57 enterprise networks, the first large-scale academic study of middlebox deployment. We show that APLOMB solves real problems faced by network administrators, can outsource over 90% of middlebox hardware in a typical large enterprise network, and, in a case study of a real enterprise, imposes an average latency penalty of 1.1ms and median bandwidth inflation of 3.8%.

/pdf/making-middleboxes-someone-else-s-problem-network-processing-2dz3orsdcj.pdf

Making middleboxes someone else's problem: network processing as a cloud service

User-perceived quality-of-experience (QoE) is critical in Internet video applications as it impacts revenues for content providers and delivery systems. Given that there is little support in the network for optimizing such measures, bottlenecks could occur anywhere in the delivery system. Consequently, a robust bitrate adaptation algorithm in client-side players is critical to ensure good user experience. Previous studies have shown key limitations of state-of-art commercial solutions and proposed a range of heuristic fixes. Despite the emergence of several proposals, there is still a distinct lack of consensus on: (1) How best to design this client-side bitrate adaptation logic (e.g., use rate estimates vs. buffer occupancy); (2) How well specific classes of approaches will perform under diverse operating regimes (e.g., high throughput variability); or (3) How do they actually balance different QoE objectives (e.g., startup delay vs. rebuffering). To this end, this paper makes three key technical contributions. First, to bring some rigor to this space, we develop a principled control-theoretic model to reason about a broad spectrum of strategies. Second, we propose a novel model predictive control algorithm that can optimally combine throughput and buffer occupancy information to outperform traditional approaches. Third, we present a practical implementation in a reference video player to validate our approach using realistic trace-driven emulations.

/pdf/a-control-theoretic-approach-for-dynamic-adaptive-video-1mmix9lrz0.pdf

A Control-Theoretic Approach for Dynamic Adaptive Video Streaming over HTTP

As the distribution of the video over the Internet becomes main- stream and its consumption moves from the computer to the TV screen, user expectation for high quality is constantly increasing. In this context, it is crucial for content providers to understand if and how video quality affects user engagement and how to best invest their resources to optimize video quality. This paper is a first step towards addressing these questions. We use a unique dataset that spans different content types, including short video on demand (VoD), long VoD, and live content from popular video con- tent providers. Using client-side instrumentation, we measure quality metrics such as the join time, buffering ratio, average bitrate, rendering quality, and rate of buffering events.We quantify user engagement both at a per-video (or view) level and a per-user (or viewer) level. In particular, we find that the percentage of time spent in buffering (buffering ratio) has the largest impact on the user engagement across all types of content. However, the magnitude of this impact depends on the content type, with live content being the most impacted. For example, a 1% increase in buffering ratio can reduce user engagement by more than three minutes for a 90-minute live video event. We also see that the average bitrate plays a significantly more important role in the case of live content than VoD content.

/pdf/understanding-the-impact-of-video-quality-on-user-engagement-dbkb3mbmmp.pdf

Understanding the impact of video quality on user engagement

Data centers deploy a variety of middleboxes (e.g., firewalls, load balancers and SSL offloaders) to protect, manage and improve the performance of applications and services they run. Since existing networks provide limited support for middleboxes, administrators typically overload path selection mechanisms to coerce traffic through the desired sequences of middleboxes placed on the network path. These ad-hoc practices result in a data center network that is hard to configure and maintain, wastes middlebox resources, and cannot guarantee middlebox traversal under network churn.To address these issues, we propose the policy-aware switching layer or PLayer, a new layer-2 for data centers consisting of inter-connected policy-aware switches or pswitches. Unmodified middleboxes are placed off the network path by plugging them into pswitches. Based on policies specified by administrators, pswitches explicitly forward different types of traffic through different sequences of middleboxes. Experiments using our prototype software pswitches suggest that the PLayer is flexible, uses middleboxes efficiently, and guarantees correct middlebox traversal under churn.

/pdf/a-policy-aware-switching-layer-for-data-centers-376vwt6hkh.pdf

A policy-aware switching layer for data centers

The lack of a concise and standard language to describe diverse middlebox functionality and deployment configurations adversely affects current middlebox deployment, as well as middlebox-related research. To alleviate this problem, we present a simple middlebox model that succinctly describes how different middleboxes process packets and illustrate it by representing four common middleboxes. We set up a pilot online repository of middlebox models and prototyped model inference and validation tools.

Modeling middleboxes

In order for overlays and new network architectures to gain real user acceptance, users should be able to leverage overlay functionality without any modifications to their applications and operating systems. We present our design, implementation, and experience with OCALA, an overlay convergence architecture that achieves this goal. OCALA interposes an overlay convergence layer below the transport layer. This layer is composed of an overlay independent sub-layer that interfaces with legacy applications, and an overlay dependent sub-layer that delivers packets to the overlay. Unlike previous efforts, OCALA enables: (a) simultaneous access to multiple overlays (b) communication between hosts in different overlays (c) communication between overlay hosts and legacy hosts (d) extensibility, allowing researchers to incorporate their overlays into OCALA. We currently support five overlays, i3, RON, HIP, DOA and OverDoSe on Linux, Windows XP/2000 and Mac OS X. We (and a few other research groups and end-users) have used OCALA for over a year with many legacy applications ranging from web browsers to remote desktop applications.

/pdf/ocala-an-architecture-for-supporting-legacy-applications-45b75g5te1.pdf

Dilip Antony Joseph

Papers

Understanding the impact of video quality on user engagement

A policy-aware switching layer for data centers

Modeling middleboxes

Understanding the impact of video quality on user engagement

OCALA: an architecture for supporting legacy applications over overlays