J. Almeida, M. Dabu, A. Manikutty, and P. Cao, Providing differentiated quality of service in Web hosting services, Proc. First Workshop on Internet Server Performance, 1998.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....Numerous mechanisms for service differentiation and performance isolation have been proposed in the literature. Such mechanisms for web servers include QoS aware extensions for admission control[6] SYN policing and request classification [14] accept queue scheduling [2] and CPU scheduling [3]. These mechanisms enable a web server to differentiate between requests from different classes and provide class specific guarantees on performance (for instance, by providing preferential treatment to users who are purchasing items at an e commerce site over users who are merely browsing) One ....

....claims about which resources become the bottleneck. For instance, one recent study has claimed that the (socket) accept queue is the bottleneck resource in web servers [2] while another has claimed that scheduling of requests on the CPU is the determining factor in web server performance [3]. Thus, it is not evident a priori as to which subset of QoS mechanisms should be employed by a web server and under what operating regions. The increasing complexity of the web server architecture, the dynamic nature of web workloads, and the interactions between various QoS mechanisms makes the ....

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. Providing Differentiated Quality-of-Service in Web Hosting Services. In Proceedings of the SIGMETRICS Workshop on Internet Server Performance, 1998.

....Numerous mechanisms for service differentiation and performance isolation have been proposed in the literature. Such mechanisms for web servers include QoSaware extensions for admission control[8] SYN policing and request classification[28] accept queue scheduling [2] and CPU scheduling [3]. These mechanisms enable a web server to differentiate between requests from different classes and provide class specific guarantees on performance (for instance, by providing preferential treatment to users who are purchasing items at an e commerce site over users who are merely browsing, or by ....

....claims about the utility and benefits of these mechanisms. For instance, one recent study has claimed that the (socket) accept queue is the bottleneck resource in web servers [2] while another has claimed that scheduling of requests on the CPU is the determining factor in web server performance [3]. Thus, it is not evident a priori as to which subset of QoS mechanisms should be employed by a web server and under what operating regions. The increasing complexity of the web server architecture, the dynamic nature of web workloads [10, 13] and the interactions between various QoS mechanisms ....

[Article contains additional citation context not shown here]

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. Providing Differentiated Quality-of-Service in Web Hosting Services. In Proceedings of the SIGMETRICS Workshop on Internet Server Performance, 1998.

....qualities of Web hosting service. Most of these are priority based in that they do not provide guaranteed QoS. In other words, these approaches allow one service class to receive qualitatively better service than the other, but does not provide a quantitative bound on by how much. Almeida et al. [1] used a special scheduler process that determines the number of concurrent Web server threads allowed for each service class, and further maps these threads to different priority levels of the kernel s process scheduler. Pandey et al. 22] used a dedicated QoS daemon that sits behind the Web ....

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. "Providing Differentiated Quality-of-Service in Web Hosting Services." In Proceedings of the Workshop on Internet Server Performance, Madison, WI, June 1998.

....scheduling of requests. These papers are generally interested in providing different levels of QOS to different customers, rather than using a size based scheme like our own. Various ideas have been tried to implement such prioritization schemes. We describe these below. Almeida et al. [2] use both a user level approach and a kernel level implementation to prioritizing HTTP requests at a Web server. The user level approach in [2] involves modifying the Apache Web server to include a Scheduler process which determines the order in which requests are fed to the Web server. This ....

....using a size based scheme like our own. Various ideas have been tried to implement such prioritization schemes. We describe these below. Almeida et al. 2] use both a user level approach and a kernel level implementation to prioritizing HTTP requests at a Web server. The user level approach in [2] involves modifying the Apache Web server to include a Scheduler process which determines the order in which requests are fed to the Web server. This modification is all in the application level and therefore does not have any control over what the o.s. does when servicing the requests. The ....

[Article contains additional citation context not shown here]

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. Providing differentiated quality-of-service in Web hosting services. In Proceedings of the First Workshop on Internet Server Performance, June 1998.

....on reducing delay at the server by using sizebased connection scheduling. In the remainder of this section we discuss only work on priority based or size based scheduling of requests. We first discuss related implementation work and then discuss relevant theoretical results. Almeida et al. [1] use both a user level approach and a kernel level implementation to prioritizing HTTP requests at a web server. In their experiments, the high priority requests only benefit by up to and the low priority requests suffer by up to . Another attempt at priority scheduling of HTTP ....

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. Providing differentiated quality-of-service in Web hosting services. In Proceedings of the First Workshop on Internet Server Performance, June 1998.

....on reducing delay at the server by using sizebased connection scheduling. In the remainder of this section we discuss only work on priority based or size based scheduling of requests. We first discuss related implementation work and then discuss relevant theoretical results. Almeida et al. [1] use both a user level approach and a kernel level implementation to prioritizing HTTP requests at a web server. In their experiments, the high priority requests only benefit by up to 20 and the low priority requests suffer by up to 200 . Another attempt at priority scheduling of HTTP requests ....

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. Providing differentiated quality-of-service in Web hosting services. In Proceedings of the First Workshop on Internet Server Performance, June 1998.

....us cache certain tables in preference to other tables based on the priority and hence provide some minimal quality of service. This would be 4 part of future work. A similar approach to provide quality of service schemes in a similar environment (though not for relational tuples) can be found in [7]. 5 Details of Implementation ChangeMgr Api We have implemented a change manager which handles the requests of clients for providing delta tables, current version of tables and previous versions of tables. The change manager has three threads a client thread, a server thread and an event ....

J.Almeida, M.Dabu, A.Manikuti and P.Cao. Providing Differentiated Quality of Service in Web Hosting Services

....qulities of Web hosting service. Most of them are priority based in that they do not provide quaranteed QoS. These appraches allow one service class to receive qualitatively better service than another service class, but does not provide a quantitative bound on by how much. Almeida et al. [1] attempted to provide differentiated QoS by introducing priority based request scheduling at both user and kernel levels. The user level approach uses a special scheduler process to limit the number of concurrent Web server processes allowed for each service class. The kernel approach mapps ....

....show later on that this simplicity does not lead to any compromise in the QoS guarantee that Gage can provide under various workloads. As a result of this simplicity, Gage can be readily ported to other Internet services without any changes to the service programs themselves. The projects [1, 7, 12, 22] attempted to provide differentiated qulities of Web hosting service. Most of them are priority based in that they do not provide quaranteed QoS. These appraches allow one service class to receive qualitatively better service than another service class, but does not provide a quantitative ....

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. "Providing Differentiated Quality-of-Service in Web Hosting Services." In Proceedings of the Workshop on Internet Server Performance, Madison, WI, June 1998.

....scheduling of requests. These papers are generally interested in providing different levels of service to different customers, rather than using a size based scheme like our own. Various ideas have been tried to implement such prioritization schemes. We describe these below. Almeida et al. [1] use both a user level approach and a kernel level implementation to prioritizing HTTP requests at a Web server. The user level approach in [1] involves modifying the Apache Web server to include a Scheduler process which determines the order in which requests are fed to the Web server. This ....

....using a size based scheme like our own. Various ideas have been tried to implement such prioritization schemes. We describe these below. Almeida et al. 1] use both a user level approach and a kernel level implementation to prioritizing HTTP requests at a Web server. The user level approach in [1] involves modifying the Apache Web server to include a Scheduler process which determines the order in which requests are fed to the Web server. This modification is all in the application level and therefore does not have any control over what the o.s. does when servicing the requests. The ....

[Article contains additional citation context not shown here]

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. Providing differentiated quality-of-service in Web hosting services. In Proceedings of the First Workshop on Internet Server Performance, June 1998.

....is also real time oriented and reservations are expressed in terms of processor requirements per time period. Their work addresses the priority inversion problem between different reservations. They identify the problem of processor co dependency and deal with resource composability. The work in [1] shows that by adding priorities to service requests on a web server, the server can provide differentiated QoS. Rather that building QoS into the kernel, they map requests to two different numerical priorities delivering them accordingly in order, to web processes. The research in [2] introduces ....

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. "Providing Differentiated Quality of Service in Web Hosting Services". In Proceedings of the Workshop on Internet Server Performance, June 1998. 14

....scheduling of requests. These papers are generally interested in providing different levels of QOS to different customers, rather than using a size based scheme like our own. Various ideas have been tried to implement such prioritization schemes. We describe these below. Almeida et al. [2] use both a user level approach and a kernel level implementation to prioritizing HTTP requests at a Web server. The user level approach in [2] involves modifying the Apache Web server to include a Scheduler process which determines the order in which requests are fed to the Web server. 3 This ....

....using a size based scheme like our own. Various ideas have been tried to implement such prioritization schemes. We describe these below. Almeida et al. 2] use both a user level approach and a kernel level implementation to prioritizing HTTP requests at a Web server. The user level approach in [2] involves modifying the Apache Web server to include a Scheduler process which determines the order in which requests are fed to the Web server. 3 This modification is all in the application level and therefore does not have any control over what the o.s. does when servicing the requests. The ....

[Article contains additional citation context not shown here]

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. Providing differentiated quality-of-service in Web hosting services. In Proceedings of the First Workshop on Internet Server Performance, June 1998.

....support largescale Internet middleware. It suggests that specialized operating systems can better meet the needs of these services than can their general purpose counterparts. It concludes with the measured performance and scalability of this proxy cache service. 1. INTRODUCTION Previous studies [2,5,25] have shown general purpose operating systems lack key abstractions necessary for large scale Internet server applications. According to one study, The performance of Internet server applications on a general purpose operating system is often dismayingly lower than what one would expect from ....

....abstractions necessary for large scale Internet server applications. According to one study, The performance of Internet server applications on a general purpose operating system is often dismayingly lower than what one would expect from the underlying hardware [5] Several of these studies [2,5] have modeled and implemented operating system extensions designed to alleviate these deficiencies. Our microkernel [21] provides the execution semantics and interfaces necessary to reduce the observed impedance mismatches between the design assumptions of general purpose operating systems and ....

Almeida, J., Dabu, M., Manikutty, A.,and Cao, P. Providing Differentiated Quality-of-Service in Web Hosting Services. Proceedings of the1st Workshop on Internet Server Performance, June 1998. Available from http://www.cs.wisc.edu/~cao/WISP98/final-versions/ jussara.ps.

....of connections. This is because most of the work performed in network processing is done in the kernel, and cannot be controlled by the application. This lack of control over kernel resources makes it difficult to build a Web server to provide differentiated quality of service (QoS) to its clients [1]. For example, consider a Web server which would like to provide clients with differentiated service depending on a variety of factors, such as the difference in access fees paid by corporate and home user clients, or the difference in fees paid by the owners of various content provided by the ....

....threads of Mach [14] and AlphaOS [11] and the shuttles of Spring [16] The chief differences are related to the more general nature of resource containers, and the fact that none of the above abstractions addresses the problem of controlling resource usage of kernel I O processing. Almeida et al. [1] attempted to implement QoS support in a modified Apache [4] Web server, running on the Linux operating system. QoS requirements were mapped onto process priorities. The authors were able to provide differentiated HTTP service to different QoS classes. However, the effectiveness of this technique ....

J. Almeida et al. Providing Differentiated Quality-of-Service in Web Hosting Services. In Proc. of the Workshop on Internet Server Performance, Madison, WI, June 1998.

....may still be vulnerable to denial of service attacks, in which a malicious client manages to consume all of the server s resources. Also, service providers want to exert explicit control over resource consumption policies, in order to provide differentiated quality of service (QoS) to clients [1] or to control resource usage by guest servers in a Rent A Server host [45] Existing APIs do not allow applications to directly control resource consumption throughout the host system. The root of this problem is the model for resource management in current general purpose operating systems. In ....

....can therefore directly control the resource consumption for all of its network and file I O. It seems feasible to implement the resource container abstraction as a feature of an exokernel library operating system, since the exokernel delegates most resource management to user code. Almeida et al. [1] attempted to implement QoS support in a modified Apache [3] Web server, running on a general purpose monolithic operating system. Apache uses a process for each connection, and so they mapped QoS requirements onto numeric process priorities, experimenting both with a fully user level ....

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. Providing Differentiated Quality-of-Service in Web Hosting Services. In Proc. Workshop on Internet Server Performance, June 1998.

....Digital UNIX) or by a per process thread (as in LRP) Moreover, the consumption of other resources, such as buffer space, is also uncontrolled. This lack of control over kernel resources makes it difficult to build a Web server to provide differentiated quality of service (QoS) to its clients [1]. For example, consider a Web server which would like to provide clients with differentiated service depending on a variety of factors, such as the difference in access fees paid by corporate and home user clients, or the difference in fees paid by the owners of various content provided by the ....

....reserves and shuttles are micro kernel specific solutions which do not address the problem of controlling the resource consumption of kernel I O processing. Almeida et al. attacked the problem of providing QoS support in a Web server running on a widely available general purpose operating system [1]. They mapped QoS requirements onto scheduling priorities, experimenting both with a user level implementation, and with a slightly modified Linux kernel scheduler. They used the Apache server [4] and so followed the process per connection model, although their approach could probably be extended ....

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. Providing Differentiated Quality-of-Service in Web Hosting Services. In Proceedings of the Workshop on Internet Server Performance, June 1998.

....the connection requiring service. This presents two problems: first, kernel space resource allocation is not under the control of the application; and second, resource allocation is difficult to perform on a per connection basis. These two problems have been have been noted as well in other work [1, 3]. In particular [3] has proposed an abstraction called resource containers to allow precise allocation of system resources, suitable for use at the granularity of individual connections (as would be required for shortest connection first scheduling) Mechanisms proposed in other experimental ....

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. Providing differentiated quality-of-service in Web hosting services. In Proceedings of the First Workshop on Internet Server Performance, June 1998.

No context found.

J. Almeida, M. Dabu, A. Manikutty, and P. Cao, Providing differentiated quality of service in Web hosting services, Proc. First Workshop on Internet Server Performance, 1998.

No context found.

ALMEIDA, J., DABU, M., MANIKUTTY, A., AND CAO, P. Providing differentiated quality-of-service in web hosting services. In Proc. 1st WISP (1998).

No context found.

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. Providing Differentiated Quality-of-Service in Web Hosting Services. In Proceedings of the SIGMETRICS Workshop on Internet Server Performance, 1998.

No context found.

J. Almeida, M. Dabu, A. Manikutty, and P. Cao. Providing Differentiated Quality-of-Service in Web Hosting Services. In Proceedings of the SIGMETRICS Workshop on Internet Server Performance, 1998.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC