Network measurement has long focused on the bits and bytes --- low-level network metrics such as latency and throughput, which have the advantage of being objective and directly characterizing the performance of the network. We argue that users provide a rich and largely untapped source of implicit as well as explicit signals that could complement and expand the coverage of traditional methods. Implicit feedback leverages user actions to indirectly infer the network performance and the resulting quality of user experience. Explicit feedback leverages user input, typically provided offline, to expand the reach of network measurement, especially for newer ones.

Resource management is important for conferencing services (such as Microsoft Teams, Zoom) to ensure good user experience while keeping the costs low. Key to this is the efficient provisioning and assignment of media processing (MP) servers, which do the heavy lifting of mixing and redistributing the media streams from and to the call participants.

Large services depend on correct configuration to run efficiently and seamlessly. Checking such configuration for correctness is important because services use a large and continuously increasing number of configuration files and parameters. Yet, very few such tools exist because the permissible values for a configuration parameter are seldom specified or documented, existing at best as tribal knowledge among a few domain experts.In this paper, we address the problem of configuration pattern mining: learning configuration rules from examples. Using program synthesis and a novel string profiling algorithm, we show that we can use file contents and histories of commits to learn patterns in configuration. We have built a tool called ConfMiner that implements configuration pattern mining and have evaluated it on four large repositories containing configuration for a large-scale enterprise service. Our evaluation shows that ConfMiner learns a large variety of configuration rules with high precision and is very useful in flagging anomalous configuration.

Large-scale services depend on Continuous Integration/Continuous Deployment (CI/CD) processes to maintain their agility and code-quality. Change-based testing plays an important role in finding bugs, but testing after every change is prohibitively expensive at a scale where thousands of changes are committed every hour. Test selection models deal with this issue by running a subset of tests for every change.

Analysts frequently require data from multiple sources for their tasks, but finding these sources is challenging in exabyte-scale data lakes. In this paper, we address this problem for our enterprise's data lake by using machine-learning to identify related data sources. Leveraging queries made to the data lake over a month, we build a relevance model that determines whether two columns across two data streams are related or not. We then use the model to find relations at scale across tens of millions of column-pairs and thereafter construct a data relationship graph in a scalable fashion, processing a data lake that has 4.5 Petabytes of data in approximately 80 minutes. Using manually labeled datasets as ground-truth, we show that our techniques show improvements of at least 23% when compared to state-of-the-art methods.

Today, we depend on numerous large-scale services for basic operations such as email. These services, built on the basis of Continuous Integration/Continuous Deployment (CI/CD) processes, are extremely dynamic: developers continuously commit code and introduce new features, functionality and fixes. Hundreds of commits may enter the code-base in a single day. Therefore one of the most time-critical, yet resource-intensive tasks towards ensuring code-quality is effectively testing such large code-bases. This paper presents FastLane, a system that performs data-driven test minimization. FastLane uses light-weight machine-learning models built upon a rich history of test and commit logs to predict test outcomes. Tests for which we predict outcomes need not be explicitly run, thereby saving us precious test-time and resources. Our evaluation on a large-scale email and collaboration platform service shows that our techniques can save 18.04%, i.e., almost a fifth of test-time while obtaining a test outcome accuracy of 99.99%.

Today's software development is distributed and involves continuous changes for new features and yet, their development cycle has to be fast and agile. An important component of enabling this agility is selecting the right reviewers for every code-change - the smallest unit of the development cycle. Modern tool-based code review is proven to be an effective way to achieve appropriate code review of software changes. However, the selection of reviewers in these code review systems is at best manual. As software and teams scale, this poses the challenge of selecting the right reviewers, which in turn determines software quality over time. While previous work has suggested automatic approaches to code reviewer recommendations, it has been limited to retrospective analysis. We not only deploy a reviewer suggestions algorithm - WhoDo - and evaluate its effect but also incorporate load balancing as part of it to address one of its major shortcomings: of recommending experienced developers very frequently. We evaluate the effect of this hybrid recommendation + load balancing system on five repositories within Microsoft. Our results are based around various aspects of a commit and how code review affects that. We attempt to quantitatively answer questions which are supposed to play a vital role in effective code review through our data and substantiate it through qualitative feedback of partner repositories.

Resource allocation is an integral, evolving part of many data center management problems such as virtual machine placement in data centers, network virtualization, and multi-path network routing. Since the problems are inherently NP-Hard, most existing systems use custom-designed heuristics to find a suitable solution. However, such heuristics are often rigid, making it difficult to extend them as requirements change.

Many networked applications that run in the background on a mobile device incur significant energy drains when using the cellular radio interface for communication. This is mainly due to the radio-tail, where the cellular radio remaining in a high energy state for up to 20s after each communication spurt. In order to cut down energy consumption, many recent devices employ fast dormancy, a feature that forces the client radio to quickly go into a low energy state after a fixed short idle period. However, aggressive idle timer values for fast dormancy can increase signaling overhead due to frequent state transitions, which negatively impacts the network. In this work, we have designed and implemented RadioJockey, a system that uses program execution traces to predict the end of communication spurts, thereby accurately invoking fast dormancy without increasing network signaling load. We evaluate RadioJockey on a broad range of background applications and show that it achieves 20-40\% energy savings with negligible increase in signaling overhead compared to fixed idle timer-based approaches.

We ask the question: can network problems experienced by applications be identified based on symptoms contained in a network packet trace? An answer in the affirmative would open the doors to many opportunities, including non-intrusive monitoring of such problems on the network and matching a problem with past instances of the same problem.

We introduce Zen, a new resource allocation framework that assigns application components to node clusters to achieve high availability for partial-fault tolerant (PFT) applications. These applications have the characteristic that under partial failures, they can still produce useful output though the output quality may be reduced. Thus, the primary goal of resource allocation for PFT applications is to prevent, delay, or minimize the impact of failures on the application output quality. This paper is the first to approach this resource allocation problem from a theoretical perspective, and obtains a series of results regarding component assignments that provide the highest service availability under the constraints imposed by the application data flow graph and the hosting clusters. We show that (1) even simple versions of this resource allocation problem are NP-Hard, (2) a 2-approximate polynomial-time algorithm works for tree topologies, and (3) a simple greedy component placement performs well in practice for general application topologies. We implement a system prototype to study the application availability achieved by Zen compared to failure-oblivious placement, replication, and Zen+replication. Our experimental results show that three PFT applications achieve significant data output quality and availability benefits using Zen.

We present a new input-queued switch architecture designed to support deadline-ordered scheduling at extremely high-speeds. In particular, deadline-ordered scheduling is enabled through a combination of hardware-based sorted priority queues called P-heaps and a round-robin crossbar scheduler. The priority queues are implemented using a novel scalable pipelined heap-based architecture. Using a 0.35 micron CMOS standard-cell technology, we demonstrate a 32-port switch capable of sustaining 10 Gb/s line rates.

In this paper, we present a fast and scalable pipelined priority queue architecture for use in high-performance switches with support for fine grained quality of service (QoS) guarantees. Priority queues are used to implement highest-priority-first scheduling policies. Our hardware architecture is based on a new data structure called a pipelined heap, or P-heap for short. This data structure enables the pipelining of the enqueue and dequeue operations, thereby allowing these operations to execute in essentially constant time. In addition to being very fast, the architecture also scales very well to a large number of priority levels and to large queue sizes. We give a detailed description of this new data structure, the associated algorithms and the corresponding hardware implementation. We have implemented this new architecture using a 0.35 micron CMOS technology. Our current implementation can support 10 Gb/s connections with over 4 billion priority levels.

The complexity and dynamic nature of the Internet (and the emerging Computational Grid) demand that middleware and applications adapt to the changes in configuration and availability of resources. However, to the best of our knowledge there are no simulation tools which support systematic exploration of dynamic Grid software (or Grid resource) behavior. We describe our vision and initial efforts to build tools to meet these needs. Our MicroGrid simulation tools enable Globus applications to be run in arbitrary virtual grid resource environments, enabling broad experimentation. We describe the design of these tools, and their validation on micro- benchmarks, the NA parallel benchmarks, and an entire Grid application. These validation experiments show that the MicroGrid can match actual experiments within a few percent (2% to 4%).