Better operating system features for faster network servers

Bangs, Gaurav; Druschel, Peter; Mogul, Jeffrey C.

doi:10.1145/306225.306234

Cited by 29 publications

(12 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Following this framework and using high-level asynchronous event primitives such as selectors (cf. (Banga et al, 1998;Lea, 2003;Sun Microsystems Inc., 2011)) for the event-based style, many studies compare these two programming styles, often focusing on performance of server architectures (see (Hu et al, 2010, § 6) for recent studies on event programming). These implementations implicitly or explicitly assume a transformation from a program written in the thread-based style, especially those which generate a new thread for each service request (as in some of the thread-based web servers), to an equivalent event-based program, which treats concurrent services using a single threaded event-loop (as in event-based web servers).…”

Section: Lauer-needham Transformmentioning

confidence: 99%

On asynchronous eventful session semantics

Kouzapas¹,

Yoshida²,

Hu³

et al. 2014

Math. Struct. Comp. Sci.

View full text Add to dashboard Cite

Event-driven programming is one of the major paradigms in concurrent and communication-based programming, where events are typically detected as the arrival of messages on asynchronous channels. Unfortunately, the flexibility and performance of traditional event-driven programming come at the cost of more complex programs: low-level APIs and the obfuscation of event-driven control flow make programs difficult to read, write and verify. This paper introduces a π-calculus with session types that models event-driven session programming (called ESP) and studies its behavioural theory. The main characteristics of the ESP model are asynchronous, order-preserving message passing, non-blocking detection of event/message arrivals, and dynamic inspection of session types. Session types offer formal safety guarantees, such as communication and event-handling safety, and programmatic benefits that overcome problems with existing event-driven programming languages and techniques. The new typed bisimulation theory developed for the ESP model is distinct from standard synchronous or asynchronous bisimulation, capturing the semantic nature of eventful session-based processes. The bisimilarity coincides with reduction-closed barbed congruence. We demonstrate the features and benefits of event-driven session programming and the behavioural theory through two key usecases. First, we examine an encoding and the semantic behaviour of the event selector, a central component of general event-driven systems, providing core results for verifying type-safe event-driven applications. Second, we examine the Lauer-Needham duality, building on the selector encoding and bisimulation theory to prove that a systematic transformation from multithreaded to event-driven session processes is type-and semantics-preserving.

show abstract

Section: Lauer-needham Transformmentioning

confidence: 99%

On asynchronous eventful session semantics

Kouzapas¹,

Yoshida²,

Hu³

et al. 2014

Math. Struct. Comp. Sci.

View full text Add to dashboard Cite

show abstract

“…Various researchers have also proposed modifications in operating system interfaces and mechanisms for efficient notification and delivery of network events to user-space servers [4,5,6,7] thus reducing the amount of data copied between the kernel and the user-space [8], reducing the number of kernel boundary crossings and a combination of the above [9]. Some of the researchers have tailored existing socket API's and implemented new APIs with newer web servers in mind [14] while others [15] have redesigned and improved the TCP/IP stack for Web server workloads to efficiently manage short-lived connections. To reduce the event notification overhead, Banga et al [16,17] optimized the implementation of select () and poll () system calls thereby improving the already existing interfaces and their implementation.…”

Section: Related Literaturementioning

confidence: 99%

Analysis of Delivery of Web Contents for Kernel-mode and User-mode Web Servers

Aaqib¹,

Sharma²

2011

IJCA

View full text Add to dashboard Cite

In this paper, the architecture of kernel-mode and user-mode webservers and the constraints that affect their performance are studied. A set of experiments have been performed to measure and analyze performance of kernel-mode and user-mode web servers on an open source Scientific Linux CERN platform. Web servers under study include kernel-mode TUX web server and user-mode Apache web server for varying static workload sizes. The results of the experiments revealed that the performance of kernel mode web servers greatly exceeds to that of user-space web servers.

show abstract

“…There have been many studies on the performance improvement of operating systems in the context of Internet servers [1], [2], [3], [4], [5], [6], [12]. Most of the studies optimized network or disk subsystems in the scope of individual subsystems.…”

Section: Previous Research On Internet Servermentioning

confidence: 99%

“…For performance improvement of Internet server systems, most studies have focused on either optimizing individual I/O subsystems [1], [2], [3], [4], [5], [6] or eliminating data copying among the subsystems [7]. There have been few studies about interoperation between two subsystems, considering each subsystem's specific mechanisms.…”

Section: Introductionmentioning

confidence: 99%

TPF: TCP Plugged File System for Efficient Data Delivery over TCP

Lim

Park

2007

IEEE Trans. Comput.

View full text Add to dashboard Cite

Abstract-Most Internet services, including Web, FTP, and streaming, have been realized on top of TCP, which is the de facto protocol for data delivery over the Internet. Therefore, in order to achieve high-performance data delivery over TCP, we thoroughly analyze TCP-based data delivery and identify three critical mismatches in a general file system design while supplying data to TCP. The first is the frequent sleeping/waking up of a server process that accompanies excessive context switching overheads due to processing TCP data and ACK segments in different contexts. The second is the inefficient uniform data prefetching for TCP connections, irrespective of their characteristics such as bandwidth, latency, and the status of a send buffer. The third is inefficient disk access due to the ignorance of abrupt changes in TCP connections. As a remedy to these mismatches, we newly design a TCPplugged file system (TPF) which is comprised of three novel mechanisms, each of which relieves the identified mismatches, integrated data sending routine, TCP aware data prefetching, and eager disk request cancellation. With these mechanisms, TPF becomes capable of supplying data managed by a file system to TCP connections timely and seamlessly and becomes reactive to abrupt changes in TCP connections. As a consequence, TPF provides minimal context switching overhead, high buffer utilization, and highly effective disk access utilization. We have implemented and tested the mechanisms in Linux 2.4. The experimental results show that the number of context switching is reduced by up to 40 percent and the overall system performance is improved by 3-34 percent.

show abstract

Better operating system features for faster network servers

Cited by 29 publications

References 19 publications

On asynchronous eventful session semantics

On asynchronous eventful session semantics

Analysis of Delivery of Web Contents for Kernel-mode and User-mode Web Servers

TPF: TCP Plugged File System for Efficient Data Delivery over TCP

Contact Info

Product

Resources

About