SEARCH · Searcharxiv

Results for “cs.OS”

Search indexed arXiv papers on artificial intelligence, large language models, computer vision and robotics. Read source abstracts and follow links to arXiv.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 433 records · Page 24Linked to original sources

arXiv2020

Nefele: Process Orchestration for the Cloud

Virtualization, either at OS- or hardware level, plays an important role in cloud computing. It enables easier automation and faster deployment in distributed environments. While virtualized infrastructures provide a level of management flexibility, they lack practical abstraction of the distributed resources. A developer in such an environment still needs to deal with all the complications of building a distributed software system. Different orchestration systems are built to provide that abstraction; however, they do not solve the inherent challenges of distributed systems, such as synchronization issues or resilience to failures. This paper introduces Nefele, a decentralized process orchestration system that automatically deploys and manages individual processes, rather than containers/VMs, within a cluster. Nefele is inspired by the Single System Image (SSI) vision of mitigating the intricacies of remote execution, yet it maintains the flexibility and performance of virtualized infrastructures. Nefele offers a set of APIs for building cloud-native applications that lets the developer easily build, deploy, and scale applications in a cloud environment. We have implemented and deployed Nefele on a cluster in our datacenter and evaluated its performance. Our evaluations show that Nefele can effectively deploy, scale, and monitor processes across a distributed environment, while it incorporates essential primitives to build a distributed software system.

cs.DC↗

arXiv2020

FastDrain: Removing Page Victimization Overheads in NVMe Storage Stack

Host-side page victimizations can easily overflow the SSD internal buffer, which interferes I/O services of diverse user applications thereby degrading user-level experiences. To address this, we propose FastDrain, a co-design of OS kernel and flash firmware to avoid the buffer overflow, caused by page victimizations. Specifically, FastDrain can detect a triggering point where a near-future page victimization introduces an overflow of the SSD internal buffer. Our new flash firmware then speculatively scrubs the buffer space to accommodate the requests caused by the page victimization. In parallel, our new OS kernel design controls the traffic of page victimizations by considering the target device buffer status, which can further reduce the risk of buffer overflow. To secure more buffer spaces, we also design a latency-aware FTL, which dumps the dirty data only to the fast flash pages. Our evaluation results reveal that FastDrain reduces the 99th response time of user applications by 84%, compared to a conventional system.

cs.OS↗

arXiv2020

Optimizing Placement of Heap Memory Objects in Energy-Constrained Hybrid Memory Systems

Main memory (DRAM) significantly impacts the power and energy utilization of the overall server system. Non-Volatile Memory (NVM) devices, such as Phase Change Memory and Spin-Transfer Torque RAM, are suitable candidates for main memory to reduce energy consumption. But unlike DRAM, NVMs access latencies are higher than DRAM and NVM writes are more energy sensitive than DRAM write operations. Thus, Hybrid Main Memory Systems (HMMS) employing DRAM and NVM have been proposed to reduce the overall energy depletion of main memory while optimizing the performance of NVM. This paper proposes eMap, an optimal heap memory object placement planner in HMMS. eMap considers the object-level access patterns and energy consumption at the application level and provides an ideal placement strategy for each object to augment performance and energy utilization. eMap is equipped with two modules, eMPlan and eMDyn. Specifically, eMPlan is a static placement planner which provides one time placement policies for memory object to meet the energy budget while eMDyn is a runtime placement planner to consider the change in energy limiting constraint during the runtime and shuffles the memory objects by taking into account the access patterns as well as the migration cost in terms of energy and performance. The evaluation shows that our proposed solution satisfies both the energy limiting constraint and the performance. We compare our methodology with the state-of-the-art memory object classification and allocation (MOCA) framework. Our extensive evaluation shows that our proposed solution, eMPlan meets the energy constraint with 4.17 times less costly and reducing the energy consumption up to 14% with the same performance. eMDyn also satisfies the performance and energy requirement while considering the migration cost in terms of time and energy.

cs.AR↗

arXiv2020

DPCP-p: A Distributed Locking Protocol for Parallel Real-Time Tasks

Real-time scheduling and locking protocols are fundamental facilities to construct time-critical systems. For parallel real-time tasks, predictable locking protocols are required when concurrent sub-jobs mutually exclusive access to shared resources. This paper for the first time studies the distributed synchronization framework of parallel real-time tasks, where both tasks and global resources are partitioned to designated processors, and requests to each global resource are conducted on the processor on which the resource is partitioned. We extend the Distributed Priority Ceiling Protocol (DPCP) for parallel tasks under federated scheduling, with which we proved that a request can be blocked by at most one lower-priority request. We develop task and resource partitioning heuristics and propose analysis techniques to safely bound the task response times. Numerical evaluation (with heavy tasks on 8-, 16-, and 32-core processors) indicates that the proposed methods improve the schedulability significantly compared to the state-of-the-art locking protocols under federated scheduling.

cs.OS↗

arXiv2021

IOCA: High-Speed I/O-Aware LLC Management for Network-Centric Multi-Tenant Platform

In modern server CPUs, last-level cache (LLC) is a critical hardware resource that exerts significant influence on the performance of the workloads, and how to manage LLC is a key to the performance isolation and QoS in the cloud with multi-tenancy. In this paper, we argue that besides CPU cores, high-speed network I/O is also important for LLC management. This is because of an Intel architectural innovation -- Data Direct I/O (DDIO) -- that directly injects the inbound I/O traffic to (part of) the LLC instead of the main memory. We summarize two problems caused by DDIO and show that (1) the default DDIO configuration may not always achieve optimal performance, (2) DDIO can decrease the performance of non-I/O workloads which share LLC with it by as high as 32%. We then present IOCA, the first LLC management mechanism for network-centric platforms that treats the I/O as the first-class citizen. IOCA monitors and analyzes the performance of the cores, LLC, and DDIO using CPU's hardware performance counters, and adaptively adjusts the number of LLC ways for DDIO or the tenants that demand more LLC capacity. In addition, IOCA dynamically chooses the tenants that share its LLC resource with DDIO, to minimize the performance interference by both the tenants and the I/O. Our experiments with multiple microbenchmarks and real-world applications in two major end-host network models demonstrate that IOCA can effectively reduce the performance degradation caused by DDIO, with minimal overhead.

cs.AR↗

arXiv2020

LINTS^RT: A Learning-driven Testbed for Intelligent Scheduling in Embedded Systems

Due to the increasing complexity seen in both workloads and hardware resources in state-of-the-art embedded systems, developing efficient real-time schedulers and the corresponding schedulability tests becomes rather challenging. Although close to optimal schedulability performance can be achieved for supporting simple system models in practice, adding any small complexity element into the problem context such as non-preemption or resource heterogeneity would cause significant pessimism, which may not be eliminated by any existing scheduling technique. In this paper, we present LINTS^RT, a learning-based testbed for intelligent real-time scheduling, which has the potential to handle various complexities seen in practice. The design of LINTS^RT is fundamentally motivated by AlphaGo Zero for playing the board game Go, and specifically addresses several critical challenges due to the real-time scheduling context. We first present a clean design of LINTS^RT for supporting the basic case: scheduling sporadic workloads on a homogeneous multiprocessor, and then demonstrate how to easily extend the framework to handle further complexities such as non-preemption and resource heterogeneity. Both application and OS-level implementation and evaluation demonstrate that LINTS^RT is able to achieve significantly higher runtime schedulability under different settings compared to perhaps the most commonly applied schedulers, global EDF, and RM. To our knowledge, this work is the first attempt to design and implement an extensible learning-based testbed for autonomously making real-time scheduling decisions.

cs.OS↗

arXiv2021

Analyzing and Mitigating Data Stalls in DNN Training

Training Deep Neural Networks (DNNs) is resource-intensive and time-consuming. While prior research has explored many different ways of reducing DNN training time, the impact of input data pipeline, i.e., fetching raw data items from storage and performing data pre-processing in memory, has been relatively unexplored. This paper makes the following contributions: (1) We present the first comprehensive analysis of how the input data pipeline affects the training time of widely-used computer vision and audio Deep Neural Networks (DNNs), that typically involve complex data preprocessing. We analyze nine different models across three tasks and four datasets while varying factors such as the amount of memory, number of CPU threads, storage device, GPU generation etc on servers that are a part of a large production cluster at Microsoft. We find that in many cases, DNN training time is dominated by data stall time: time spent waiting for data to be fetched and preprocessed. (2) We build a tool, DS-Analyzer to precisely measure data stalls using a differential technique, and perform predictive what-if analysis on data stalls. (3) Finally, based on the insights from our analysis, we design and implement three simple but effective techniques in a data-loading library, CoorDL, to mitigate data stalls. Our experiments on a range of DNN tasks, models, datasets, and hardware configs show that when PyTorch uses CoorDL instead of the state-of-the-art DALI data loading library, DNN training time is reduced significantly (by as much as 5x on a single server).

cs.DC↗

arXiv2020

Scheduling of Real-Time Tasks with Multiple Critical Sections in Multiprocessor Systems

The performance of multiprocessor synchronization and locking protocols is a key factor to utilize the computation power of multiprocessor systems under real-time constraints. While multiple protocols have been developed in the past decades, their performance highly depends on the task partition and prioritization. The recently proposed Dependency Graph Approach showed its advantages and attracted a lot of interest. It is, however, restricted to task sets where each task has at most one critical section. In this paper, we remove this restriction and demonstrate how to utilize algorithms for the classical job shop scheduling problem to construct a dependency graph for tasks with multiple critical sections. To show the applicability, we discuss the implementation in Litmus^{RT} and report the overheads. Moreover, we provide extensive numerical evaluations under different configurations, which in many situations show significant improvement compared to the state-of-the-art.

cs.OS↗

arXiv2020

DBOS: A Proposal for a Data-Centric Operating System

Current operating systems are complex systems that were designed before today's computing environments. This makes it difficult for them to meet the scalability, heterogeneity, availability, and security challenges in current cloud and parallel computing environments. To address these problems, we propose a radically new OS design based on data-centric architecture: all operating system state should be represented uniformly as database tables, and operations on this state should be made via queries from otherwise stateless tasks. This design makes it easy to scale and evolve the OS without whole-system refactoring, inspect and debug system state, upgrade components without downtime, manage decisions using machine learning, and implement sophisticated security features. We discuss how a database OS (DBOS) can improve the programmability and performance of many of today's most important applications and propose a plan for the development of a DBOS proof of concept.

cs.OS↗

arXiv2020

HeRTA: Heaviside Real-Time Analysis

We investigate the mathematical properties of event bound functions as they are used in the worst-case response time analysis and utilization tests. We figure out the differences and similarities between the two approaches. Based on this analysis, we derive a more general form do describe events and event bounds. This new unified approach gives clear new insights in the investigation of real-time systems, simplifies the models and will support algebraic proofs in future work. In the end, we present a unified analysis which allows the algebraic definition of any scheduler. Introducing such functions to the real-time scheduling theory will lead two a more systematic way to integrate new concepts and applications to the theory. Last but not least, we show how the response time analysis in dynamic scheduling can be improved.

cs.OS↗

arXiv2020

Interprocess Communication in FreeBSD 11: Performance Analysis

Interprocess communication, IPC, is one of the most fundamental functions of a modern operating system, playing an essential role in the fabric of contemporary applications. This report conducts an investigation in FreeBSD of the real world performance considerations behind two of the most common IPC mechanisms; pipes and sockets. A simple benchmark provides a fair sense of effective bandwidth for each, and analysis using DTrace, hardware performance counters and the operating system's source code is presented. We note that pipes outperform sockets by 63% on average across all configurations, and further that the size of userspace transmission buffers has a profound effect on performance - larger buffers are beneficial up to a point (~32-64 KiB) after which performance collapses as a result of devastating cache exhaustion. A deep scrutiny of the probe effects at play is also presented, justifying the validity of conclusions drawn from these experiments.

cs.OS↗

arXiv2020

eXpOS: A Simple Pedagogical Operating System for Undergraduate Instruction

An operating system project suitable for undergraduate computing/electrical sciences students is presented. The project can be used as a course project in a one semester course, or as a self-study project for motivated students. The course is organized such that a student with a basic background in programming and computer organization can follow the implementation road map available online, and build the OS from scratch on her personal machine/laptop, with minimal instructional supervision. The student is provided with a simulated abstract machine, an application interface specification, specification and design of the OS, and a step by step project implementation road map. The functionalities of the OS include multitasking, virtual memory, semaphores, shared memory, an elementary file system, interrupt driven disk and console I/O, and a limited multi-user support. The final stage of the project involves porting the OS to a two-core machine. An independent one semester compiler design project, where the student builds a compiler for a tiny object oriented programming language that generates target code that can be loaded and executed by the OS is also briefly discussed.

cs.OS↗

arXiv2020

Consideration for effectively handling parallel workloads on public cloud system

We retrieved and analyzed parallel storage workloads of the FUJITSU K5 cloud service to clarify how to build cost-effective hybrid storage systems. A hybrid storage system consists of fast but low-capacity tier (first tier) and slow but high-capacity tier (second tier). And, it typically consists of either SSDs and HDDs or NVMs and SSDs. As a result, we found that 1) regions for first tier should be assigned only if a workload includes large number of IO accesses for a whole day, 2) the regions that include a large number of IO accesses should be dynamically chosen and moved from second tier to first tier for a short interval, and 3) if a cache hit ratio is regularly low, use of the cache for the workload should be cancelled, and the whole workload region should be assigned to the region for first tier. These workloads already have been released from the SNIA web site.

cs.DC↗

arXiv2020

Making Distributed Mobile Applications SAFE: Enforcing User Privacy Policies on Untrusted Applications with Secure Application Flow Enforcement

Today's mobile devices sense, collect, and store huge amounts of personal information, which users share with family and friends through a wide range of applications. Once users give applications access to their data, they must implicitly trust that the apps correctly maintain data privacy. As we know from both experience and all-too-frequent press articles, that trust is often misplaced. While users do not trust applications, they do trust their mobile devices and operating systems. Unfortunately, sharing applications are not limited to mobile clients but must also run on cloud services to share data between users. In this paper, we leverage the trust that users have in their mobile OSes to vet cloud services. To do so, we define a new Secure Application Flow Enforcement (SAFE) framework, which requires cloud services to attest to a system stack that will enforce policies provided by the mobile OS for user data. We implement a mobile OS that enforces SAFE policies on unmodified mobile apps and two systems for enforcing policies on untrusted cloud services. Using these prototypes, we demonstrate that it is possible to enforce existing user privacy policies on unmodified applications.

cs.CR↗

arXiv2021

BumbleBee: Application-aware adaptation for container orchestration

Modern applications have embraced separation of concerns as a first-order organizing principle through the use of containers, container orchestration, and service meshes. However, adaptation to unexpected network variation has not followed suit. We present BumbleBee, a lightweight extension to the container ecosystem that supports application-aware adaptation. BumbleBee provides a simple abstraction for making decisions about network data using application semantics. Because this abstraction is placed within the communications framework of a modern service mesh, it is closer to the point at which changes are detected, providing more responsive and effective adaptation than possible at endpoints.

cs.DC↗

arXiv2020

Analysis of Interference between RDMA and Local Access on Hybrid Memory System

We can use a hybrid memory system consisting of DRAM and Intel Optane DC Persistent Memory (We call it DCPM in this paper) as DCPM is now commercially available since April 2019. Even if the latency for DCPM is several times higher than that for DRAM, the capacity for DCPM is several times higher than that for DRAM and the cost of DCPM is also several times lower than that for DRAM. In addition, DCPM is non-volatile. A Server with this hybrid memory system could improve the performance for in-memory database systems and virtual machine (VM) systems because these systems often consume a large amount of memory. Moreover, a high-speed shared storage system can be implemented by accessing DCPM via remote direct memory access (RDMA). I assume that some of the DCPM is often assigned as a shared area among other remote servers because applications executed on a server with a hybrid memory system often cannot use the entire capacity of DCPM. This paper evaluates the interference between local memory access and RDMA from a remote server. As a result, I indicate that the interference on this hybrid memory system is significantly different from that on a conventional DRAM-only memory system. I also believe that some kind of throttling implementation is needed when this interference occures.

cs.PF↗

arXiv2020

toki: A Build- and Test-Platform for Prototyping and Evaluating Operating System Concepts in Real-Time Environments

Typically, even low-level operating system concepts, such as resource sharing strategies and predictability measures, are evaluated with Linux on PC hardware. This leaves a large gap to real industrial applications. Hence, the direct transfer of the results might be difficult. As a solution, we present toki, a prototyping and evaluation platform based on FreeRTOS and several open-source libraries. toki comes with a unified build- and test-environment based on Yocto and Qemu, which makes it well suited for rapid prototyping. With its architecture chosen similar to production industrial systems, toki provides the ground work to implement early prototypes of real-time systems research results, up to technology readiness level 7, with little effort.

cs.OS↗

arXiv2020

Quantifying the Latency and Possible Throughput of External Interrupts on Cyber-Physical Systems

An important characteristic of cyber-physical systems is their capability to respond, in-time, to events from their physical environment. However, to the best of our knowledge there exists no benchmark for assessing and comparing the interrupt handling performance of different software stacks. Hence, we present a flexible evaluation method for measuring the interrupt latency and throughput on ARMv8-A based platforms. We define and validate seven test-cases that stress individual parts of the overall process and combine them to three benchmark functions that provoke the minimal and maximal interrupt latency, and maximal interrupt throughput.

cs.OS↗