J. C. Mogul. The case for persistent-connection HTTP. In Proceedings of ACM SIGCOMM '95, pages 299--313, October 1995.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....if the loss rate is higher. The reason is that higher RTTs make loss recovery more expensive since timeouts depend on the (estimated) RTT. E) Persistent Connections Next we explore how the response times change if multiple requests are permitted to use a single serial, persistent connection ([32]) for several requests. Figure 6(E) left, middle) shows the results for the setup in Table 2, row (E) where every connection is reused 5 times. Figure 6(E) right) shows the response time as a function of the number of requests per connection, ranging from 0 to 10. We see that using persistent ....

....FreeBSD. This agrees with our study. In [35] the authors study the effect of WAN conditions, and find that losses and delays can affect response times. They use a different workload from ours (Surge workload) but have similar findings. The benefits of persistent connections are evaluated by [32] and [10] in a LAN environment. There are also several papers which study real web servers in action, rather than a controlled lab setting, e.g. 33] and [41] 7 Conclusion This paper provides a detailed performance study of the effect of overload on an Apache web server running over Linux and ....

J. C. Mogul. The case for persistent-connection HTTP. In Proceedings of ACM SIGCOMM '95, pages 299--313, October 1995.

....0 100 200 300 0.E 00 1.E 06 2.E 06 3.E 06 4. E 06 MessageID(sortedbymsgsize) Messagesize(#bytes) Figure 4: Size distribution of notification messages (including duplicates) One suggestion for reducing the overhead of connection setup and teardown is to use persistent connections [13], i.e. reuse a TCP connection for multiple transfers. In our case, the servers sending the notification messages can maintain persistent connections with the gateways of the wireless ISPs and then send all messages on this connection. 4.2 Message Popularity Analysis and Its Implications ....

J. C. Mogul. The Case for Persistent-Connection HTTP. In Proceedings of ACM SIGCOMM 95, August 1995.

....is established with a 3 way handshake; and typically several additional round trip times (RTT) are needed for TCP to achieve appropriate transmission speed [34] Each connection establishment induces user perceived latency and processing overhead. Thus, persistent connections were proposed [31, 19, 32] and are now a default with the draft HTTP 1.1 standard [17, 18] HTTP 1.1 keeps open and reuses TCP connections to transmit sequences of request response messages; hence, reducing the number of connection establishments and resulting latency and processing overheads. Deployment of HTTP 1.1 ....

....when using the same average amount of open connections. Even more significant reductions, typically around 50 , are achieved on the clicks logs. The resulting performance improvement is considerable, since connection establishments induce user perceived latency and overhead [16, 31, 23] wheras large number of open connections is both detrimental to throughput [4] and is more likely to reach the server s hard set limits, causing it to refuse new connections. In Section 2 we discuss the interaction between HTTP and TCP at Web servers, and the nature of server logs data. We then ....

J. C. Mogul. The case for persistent-connection HTTP. Computer Communication Review, 25(4):299--313, October 1995. http://www.research.digital.com/wrl/techreports/abstracts/95.4.html.

....server must minimize latency, maximize throughput, and avoid utilizing the CPU(s) unnecessarily. # Programming simplicity The design of the server should simplify the use of efficient concurrency strategies; # Adaptability Integrating new or improved transport protocols (such as HTTP 1. 1 [3]) should incur minimal maintenance costs. A Web server can be implemented using several concurrency strategies, such as multiple synchronous threads and reactive synchronous event dispatching. But such conventional approaches have drawbacks as discussed in [1] The Proactor pattern provides a ....

J. C. Mogul, "The Case for Persistent-connection HTTP," in Proceedings of ACM SIGCOMM '95 Conference in Computer Communication Review, (Boston, MA), pp. 299--314, ACM Press, August 1995.

....Padmanabhan [2] Of particular relevance were persistent HTTP connections (P HTTP) and pipelining of multiple object requests over a single connection. Experimental data supported their recommendations and showed a significant reduction in overall latency for retrievals. A following paper by Mogul [3] demonstrated the significant reduction in latency that would be achieved by using P HTTP and pipelining; extensive simulation also demonstrated the reduction in use of server resources under TCP TIMEWAIT states of various durations. HTTP 1.1 [4] published in January 1997 incorporated persistent ....

J. C. Mogul, "The case for persistent-connection http," in Proceedings of the Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication. ACM Press, 1995, pp. 299--313.

....and the server locates the file and returns it to the client requesting it. On the surface, therefore, Web servers appear to have few opportunities for optimization. This may lead to the conclusion that optimization efforts should be directed elsewhere (such as transport protocol optimizations [11], specialized hardware [5] and client side caching [21, 12] Empirical analysis reveals that the problem is more complex and the solution space is much richer. For instance, our experimental results show that a heavily accessed Apache Web server (the most popular server on the Web today [20] ....

....System Overview and Server Optimizations There are many levels at which research on Web server performance can be conducted. Figure 2 illustrate an architectural overview of a Web system and lists potential server side optimizations. Low level solutions, such as transport protocol optimizations [11] and specialized hardware [5] are beyond the scope of our work. While these solutions can improve the end to end performance of a Web system, they do not directly solve the problem of Web server efficiency. Therefore, our research will leverage off existing work in this field [2, 4, 5] and will ....

Jeffrey C. Mogul. The Case for Persistent-connection HTTP. In Proceedings of ACM SIGCOMM '95 Conference in Computer Communication Review, pages 299--314, Boston, MA, USA, August 1995. ACM Press.

....a 2 A which does not match either of these patterns can be reduced to one which does. These three properties are proven in Appendix A as Lemmas A.1, A.2, and A.3, respectively. 16 3. 2 HTTP Connection Management One early HTTP protocol optimization was the introduction of persistent connections [24]. This feature allows multiple transactions to be sequentially conducted over a single transport (TCP) connection, amortizing the cost of TCP connection setup and teardown. This feature is the cornerstone of the connection management features introduced in HTTP 1.1 [21] which standardizes the ....

Jeffrey C. Mogul. The case for persistent-connection HTTP. In Proceedings of ACM SIGCOMM '95, August 1995.

....a ##which does not match either of these patterns can be reduced to one which does. These three properties are proven in Appendix A as Lemmas A.1, A.2, and A.3, respectively. 16 3. 2 HTTP Connection Management One early HTTP protocol optimization was the introduction of persistent connections [24]. This feature allows multiple transactions to be sequentially conducted over a single transport (TCP) connection, amortizing the cost of TCP connection setup and teardown. This feature is the cornerstone of the connection management features introduced in HTTP 1.1 [21] which standardizes the ....

Jeffrey C. Mogul. The case for persistent-connection HTTP. In Proceedings of ACM SIGCOMM '95, August 1995.

.... that most request sizes are smaller than 500 bytes, and therefore t in a typical size TCP segment (about 500 bytes) On the other hand, the mean size of a reply (carrying one component of a page) is typically between 10,000 and 20,000 bytes, and the median ranges between 1,000 and 2,000 bytes [131]. This relatively early study found that most Web pages contain fewer than 5 in lined les, have an average size smaller than 32KB, and 90 of them are smaller than about 200KB [119] In a summary of Web studies [159] an average HTML le size of about 5KB, with a median of 2KB, and an average ....

....is the main requirement for Web applications. Human factors studies report that the performance rating is considered to be very good for download times below 5 seconds. Download times between 5 and 10 seconds may be acceptable, whereas times larger than 10 seconds give low performance ratings [31, 34, 131]. In addition, since users highly value predictable performance, the variance of the page downloads also needs to be small. It is possible to signi cantly improve user perceived performance of Web browsing by insuring that some form of early feedback for a transaction is received within a few ....

[Article contains additional citation context not shown here]

Mogul J., The Case for Persistent-Connection HTTP, in Proceedings of SIGCOMM'95, August 1995.

....without the need to synchronize the web server nodes or to prevent the TCP stream from draining during cloning. Figure 6 summarizes the workflow of a cluster based web server with Socket Cloning for a nonpipelined persistent HTTP connection. Some modern web browsers support pipelining of requests [23] in which a browser can send out requests before receiving a complete response is also supported. SC is also designed to handle this service. The length of the requested file is noted by the original node when it clones the socket. The Packet Router compares the acknowledged sequence number from ....

J. C. Mogul. The case for persistent-connection HTTP. In Proceedings of the SIGCOMM'95 conference, Cambridge, MA, Aug. 1995.

....be pipelined via a single TCP connection. The persistent connection eliminates unnecessary TCP connection Object ID URL NO. Inter Local Data Access Network Communication Running lhread Figure 2. Workflow of GOS 16 setup time, resulting in improved user perceived latency and performance [51]. In our proposed system, because all the objects in the system are accessible via the Gas service, a request handling node can keep persistent connection with the client, while fetching object from other nodes via Gas. Without a Gas where objects can be shared on a global basis, it is very ....

Mogul J. The Case for Persistent-Connection HTTP. Proceedings of the ACM SIGCOMM '95 Symposium, Cambridge, USA, August 28-September 1, 1995.

....in combination with acquired knowledge about resource consumption of requests. An overload protection architecture also needs to deal with persistent connections. Persistent connections allow clients to send several requests on the same TCP connection to reduce client latency and server overhead [41]. Persistent connections represent a challenging problem for web server admission control, since the HTTP header of the first request does not reveal any information about the resource consumption of the requests that may follow on the same connection. This problem is addressed in Paper C. 1.2.2 ....

....literature. Web servers use admission control for overload protection. Some web servers base their admission decision on information found in the HTTP header. Persistent connections allow HTTP clients to send several requests on the same TCP connection to reduce client latency and server overhead [41]. Using the same TCP connection for several requests makes admission control more di# cult, since the admission control decision should be performed when the first request is received. However, the HTTP header of the first request does not reveal any information about the resource consumption of ....

J. C. Mogul. The case for persistent-connection HTTP. In SIGCOMM '95 Conference Proceedings, pages 299--313, Cambridge, MA, USA, August 1995. ACM SIGCOMM Computer Communication Review, 25(4).

....it was last used. Caches are also used in file systems, compilers, operating systems, and hardware. Extended Use predictions are useful for managing resources or connections. Applications that lease resources may predict the extension of their use. In the web, persistent HTTP connections [Mog95, FGM99] are the most cormnon example. Maintaining the connection to a web server avoids incurring additional setup delays for future requests to the server. Extending HTTP connections adds an additional management burden on both the client and the server machine that might be managed through ....

Mogul, J. C., "The Case for Persistent-Connection HTTP", in Proceedings of ACM SIGCOMM '95 Conference, Cambridge, Massachusetts, August 1995, pp. 299-314.

....load lasted 500 seconds. To more closely simulate the connection behavior of clients in the wide area, each client closes its TCP connection to the server after 5 HTTP requests, and reestablishes the connection before continuing. This value was chosen based on observations of HTTP traffic from [98]. Note that most Web servers are configured to use a much higher limit on the number of HTTP requests per connection, which is unrealistic but provides improved benchmark results. All measurements below were taken with the server running on a 4 way SMP 500 MHz 126 0 20 40 60 80 100 120 ....

J. C. Mogul. The case for persistent-connection HTTP. In Proceedings of ACM SIGCOMM '95, October 1995.

....of TCP allows it to operate in many network environments. Finally, the Hyper Text Transfer Protocol (HTTP) is a request response application layer protocol layered on top of TCP. HTTP is used to transfer Web documents between Web servers and Web clients (browsers) Currently, HTTP 1.0 and HTTP 1. 1 [21] are widely used on the Internet. Several interesting protocol interactions occur when TCP is used to transfer Web documents. For example, TCP s ow and congestion control algorithms are (arguably) designed to optimize throughput for long lived bulk data transfers. The TCP slow start phase is used ....

J. Mogul, "The Case for Persistent Connection HTTP", Proceedings of ACM SIGCOMM, Boston, MA, August 1995.

....the Department of Electrical Engineering, Stanford University, Stanford CA 94305. Email: noureddine, tobagi stanford.edu teractivity [14] or by using content replication and caching. In contrast, network delays, which form a significant part of total delay for Web transfers [4] 5] 24] [26], are usually outside the control of the provider or any other single organization, and are therefore not as easily reduced. In this paper, our focus is on network delays, and we assume that server performance has been properly addressed and server delays are therefore negligible. For a concrete ....

....reliability and congestion avoidance and control are typically the most significant. This is particularly the case for HTTP 1.0, where a TCP connection is opened for each component of a page, adding a non negligible connection establishment overhead to the total transaction delay. As discussed in [26], the use of one, persistent , TCP connection to transfer all Web requests and responses between a client and a server eliminates this overhead. This usage has been adopted in HTTP 1.1. We are interested here in the delays due to network congestion induced queuing and packet loss. TCP was ....

Mogul J., The Case for Persistent-Connection HTTP, in Proceedings of SIGCOMM '95, Aug. 1995.

....required 3300 lines of C code (excluding any code for XDFS, compression, and the Berkeley DB) HTTP version 1.1 provides certain features over 1.0 that can increase overall performance and reduce both CPU and network overhead. The current version of xProxy uses persistent 20 connections [19] so that multiple objects can be requested and received over the same network connection. With HTTP 1.0 at most one object can be requested on each connection. xProxy also supports pipelining , whereby multiple requests are sent over a single persistent connection without waiting for each ....

Mogul, J. "The Case for Persistent-Connection HTTP", Western Research Laboratory Research Report 95/4, Digital Equipment Corporation, May 1995.

....avoiding per transaction re connection and TCP slow start restart overheads. Transaction TCP provides transaction oriented service over TCP via extensions to the TCP protocol. HTTP performance can be improved by using persistent connections, in which a single connection can access multiple file [14]. Touch et al. [16] suggest that the persistent connection optimizations do not substantially affect Web access for the vast majority of users. Most users see end to end latencies of about 250 ms and use modem lines. Bandwidths over 200 Kbps are required to provide user noticeable performance ....

Mogul, J. "The Case for Persistent-Connection HTTP," Western Research Laboratory Research Report 95/4, http://www.research.digital.com/wrl/publications/abstracts/ 95.4.html, Digital Equipment Corporation, May 1995. -19-

....and [21] We examine our data set for similar characteristics to determine how the workload changes in a cable modem environment. Other studies have examined the workloads of various components of the Web, including clients ( 7] 13] servers ( 4] 12] 22] and the HTTP protocol ( 6] 16] [23]) The main performance benefit of Web proxies is the file caching that they perform. There are two general approaches to file cache management. One approach attempts to use as few resources as possible by making good replacement decisions when the cache is full (we call this the elegant ....

J. Mogul, "The Case for Persistent-Connection HTTP", Proceedings of ACM SIGCOMM `95, Cambridge, MA, pp. 299-313, 1995.

....for small data transfers (like the Web s requests and responses) Nagle algorithm, which tries to avoid the overhead of small packets can introduce delays to Web requests . Multiple connections are often used for physically co located transfers These and other problems have been discussed in [10, 13, 12, 3, 6, 9, 4]. Many of the problems mentioned above are inherent to the HTTP 1.0 protocol and attempts have been made to fix them in HTTP 1.1. All of those changes concentrated on the interactions between HTTP and TCP protocols, trying to optimize how TCP is used within the HTTP protocol. There is also an ....

J. Mogul, "The Case for Persistent-Connection HTTP", Western Research Laboratory Research Report 95/4, Digital Equipment Corporation, May 1995.

....or accept connection requests based on information provided in the HTTP header [5, 9] In this context, persistent connections represent a challenging problem. Persistent connections allow HTTP clients to send several requests on the same TCP connection to reduce client latency and server overhead [8]. When the same TCP connection is used for several requests, the HTTP header of the rst request does not reveal any information about the resource consumption of the following requests on the same connection. In [9] we presented kernel based mechanisms that provide admission control and service ....

J. C. Mogul. The case for persistent-connection http. In Proc. of SIGCOMM, 1995.

....systems today. HTTP 1.0 requires a separate TCP connection for transfer of each file associated with a Web object. Since busy servers can serve millions of requests per day, efficiency in setting up and tearing down connections on Web servers is important to good server performance using HTTP 1. 0 [84]. However there are many inherent drawbacks to this approach. In [67] Liu and Edwards show that over one quarter of the transaction time in HTTP 1.0 is spent in TCP handshaking before any data flows between Web server and client. Studies by Mogul [84] and Padmanabhan [99] address the issue of how ....

....to good server performance using HTTP 1.0 [84] However there are many inherent drawbacks to this approach. In [67] Liu and Edwards show that over one quarter of the transaction time in HTTP 1.0 is spent in TCP handshaking before any data flows between Web server and client. Studies by Mogul [84] and Padmanabhan [99] address the issue of how network conditions affect user perceived latency during Web transactions. Congestion in the Internet is a well known cause of delay in TCP transfers [104] and when network delays are assumed to be the primary source of user perceived latency, a ....

[Article contains additional citation context not shown here]

J. Mogul. The case for persistent-connection HTTP. Technical Report WRL 95/4, DEC Western Research Laboratory, Palo Alto, CA, 1995.

....accept connection requests based on information provided in the HTTP header [10, 21] In this context, persistent connections represent a challenging problem. Persistent connections allow HTTP clients to send several requests on the same TCP connection to reduce client latency and server overhead [18]. When the same TCP connection is used for several requests, the HTTP header of the rst request does not reveal any information about the resource consumption of the following requests on the same connection. In [21] we presented kernel based mechanisms that provide admission control and service ....

J. C. Mogul. The case for persistent-connection http. In Proc. of SIGCOMM, 1995.

....HTTP protocols are having on the characteristics of web traffic in the Internet. For example, measurements of TCP connection usage for early versions of the HTTP protocols pointed to clear inefficiencies in design, notably the creation of a different TCP connection for each web object reference [23]. Recent revisions to the HTTP protocol, notably version 1.1 [24] have introduced the concepts of persistent connections and pipelining. Persistent connections are provided to enable the reuse of a single TCP connection for multiple object references at the same IP address (typically embedded ....

J. Mogul, The Case for Persistent-Connection HTTP, Proceedings of ACM SIGCOMM '95, pp. 299-313.

....accept connection requests based on information provided in the HTTP header [10, 21] In this context, persistent connections represent a challenging problem. Persistent connections allow HTTP clients to send several requests on the same TCP connection to reduce client latency and server overhead [18]. When the same TCP connection is used for several requests, the HTTP header of the rst request does not reveal any information about the resource consumption of the following requests on the same connection. In [21] we presented kernel based mechanisms that provide admission control and service ....

J. C. Mogul. The case for persistent-connection http. In Proc. of SIGCOMM, 1995.

....before requesting the next page. To more closely simulate the connection behavior of clients in the wide area, each client closes the TCP connection after 5 HTTP requests, and reestablishes the connection before continuing. This value was chosen based on observations of HTTP traffic from [39]. 5 All benchmarks were run with warm filesystem and Web page caches. Note that the file set size of 3.31 GB is much larger than physical memory, and the static page cache for Haboob and Flash was set to only 200 MB; therefore, these measurements include a large amount of disk I O. 5.1.3 ....

J. C. Mogul. The case for persistent-connection HTTP. In Proc. ACM SIGCOMM'95, October 1995.

....Improvement: Table V gives an overview of several general schemes that have been proposed. The normal HTTP protocol (first row) makes n separate connections for each document (web page plus n 1 inline images) n separate requests per document, and one DNS lookup per server. Persistent HTTP [14] (incorporated into HTTP 1.1) reduces the number of connections to one per server (multiple documents at the same server can be accessed over the same connection) Client Prefetching (third row) does not reduce the number of connections, requests or DNS lookups. Client prefetching can, however, ....

J. C. Mogul, "The case for persistent connection http," in Proceedings of the ACM SIGCOMM '95 Symposium, September 1995, pp. 299--313.

....Response latency is also a ected when a new TCP connection is created for each request and because each transfer independently goes through the TCP slow start phase. Early on in the development of the Web this ineciency of the simple HTTP model was realized and addressed by a number of proposals [23, 9, 17] advocating the use of a single TCP connection for multiple request reply exchanges. Two new HTTP methods were also suggested, GETALL and GETLIST, which could be used to get al..l objects from a Web page in a single retrieval, and to get an arbitrary list of objects from a server in a single request ....

J. C. Mogul. The case for persistent-connection HTTP. In Proceedings of the ACM SIGCOMM '95 Conference.ACM, Aug. 1995.

....latency is also affected when a new TCP connection is created for each request and because each transfer independently goes through the TCP slow start phase. Early on in the development of the Web this inefficiency of the simple HTTP model was realized and addressed by a number of proposals [23,9,17] advocating the use of a single TCP connection for multiple request reply exchanges. Two new HTTP methods were also suggested, GETALL and GETLIST, which could be used to get al..l objects from a Web page in a single retrieval, and to get an arbitrary list of objects from a server in a single request ....

J. C. Mogul. The case for persistent-connection HTTP. In Proceedings of the ACM SIGCOMM '95 Conference. ACM, Aug. 1995.

....allow faster page transmission and avoid head of line blocking. HTTP 1.0 [10] will start a separate, dedicated TCP connection for every web transaction. Thus, during page retrieval, a bundle of concurrent TCP connections is active between client and server. Persistent connection HTTP (P HTTP) [24], which has been included in HTTP 1.1 [9] since being proposed, allows multiple HTTP transactions to take place sequentially over one TCP connection. However, most clients still open more than one persistent connection to a server to speed up initial page rendering. Due to the nature of the TCP ....

J. Mogul. "The case for persistent-connection HTTP." In ACM Computer Communication Review, vol. 25, pp. 299-313 October 1995.

....TCP connection resulting in higher throughput and smaller response time. Since a connection can not be maintained indefinitely, a decision on when a persistent connection is to be closed needs to be made. A fixed timeout period after which a server automatically closes a connection is suggested [1, 2]. A quasi dynamic solution, where two fixed timers are used, is proposed in [2] Nielsen et al. 3] suggest a connection to be closed after servicing a fixed number of requests. Version 1.3.1 of the Apache Web server uses a combination of the solutions proposed in [2, 3] None of these solutions ....

....a connection can not be maintained indefinitely, a decision on when a persistent connection is to be closed needs to be made. A fixed timeout period after which a server automatically closes a connection is suggested [1, 2] A quasi dynamic solution, where two fixed timers are used, is proposed in [2]. Nielsen et al. 3] suggest a connection to be closed after servicing a fixed number of requests. Version 1.3.1 of the Apache Web server uses a combination of the solutions proposed in [2, 3] None of these solutions accounts for the variability of the server load during different hours of the ....

[Article contains additional citation context not shown here]

J. C. Mogul, "The case for persistentconnection HTTP," in ACM SIGCOMM conference, pp. 299--313, Oct. 1995.

....terminating. The Thread per Session model is not appropriate for HTTP 1.0 since protocol establishes a new connection for each request. Thus, Thread per Session is equivalent to Thread per Request in HTTP 1.0. This model is applicable in HTTP 1. 1, however, since it supports persistent connections [33, 34]. Figure 17 illustrates the Thread per Session model. Thread per Session provides good support for prioritization of client requests. For instance, higher priority clients can be associated with higher priority threads. Thus, request from higher priority clients will be served ahead of requests ....

J. C. Mogul, "The Case for Persistent-connection HTTP," in Proceedings of ACM SIGCOMM '95 Conference in Computer Communication Review, (Boston, MA, USA), pp. 299--314, ACM Press, August 1995.

....HTTP protocols are having on the characteristics of web traffic in the Internet. For example, measurements of TCP connection usage for early versions of the HTTP protocols pointed to clear inefficiencies in design, notably the creation of a different TCP connection for each web object reference [26]. Recent revisions to the HTTP protocol, notably version 1.1 [28] have introduced the concepts of persistent connections and pipelining. Persistent connections are provided to enable the reuse of a single TCP connection for multiple object references at the same IP address (typically embedded ....

J. Mogul, The Case for Persistent-Connection HTTP, Proceedings of ACM SIGCOMM '95, pp. 299-313.

....Our early experiments suggested that P HTTP performance was ten times slower than the corresponding HTTP transactions in a simple page retrieval benchmark. This result is surprising since P HTTP is intended to improve performance by amortizing costs of connection creation across multiple requests [16, 13]. We found several interactions between P HTTP and TCP which explain the exceedingly poor P HTTP performance. These performance problems are not caused by specific errors in our server (Apache, beta version 1.1b4) or in our TCP implementation (SunOS 4.1.3) but they instead result from ....

Jeffrey C. Mogul. The case for persistentconnection HTTP. In Proceedings of the SIGCOMM '95, pages 299--313. ACM, August 1995.

....is only one of the necessary steps for understanding the changes occurring in Web traffic. Research efforts on Web client workloads (e.g. 5] Web proxy workloads (e.g. 2] 7] 8] 15] 17] 18] 22] 27] 33] network traffic characterizations ( e.g. 34] as well as HTTP analyses (e.g. 4] 21] 23][30]) are all required in order better understand the Web. The remainder of the paper is organized as follows. Section 2 provides background information on the 1998 World Cup, focusing on the structure of the tournament. Section 3 introduces the World Cup Web site and describes the technology that it ....

....on the Web. By reducing the number of TCP connections persistent connections reduce user latency by eliminating unnecessary round trips for the establishment of TCP connections. Persistent connections are also able to avoid latency associated with TCP slow start under certain conditions [4][30]. One disadvantage of persistent connections is the need for the server to Workload Characterization User Session Analyses Arlitt and Jin Page 47 of 90 maintain a much larger number of open TCP connections. We estimate this effect by monitoring the number of active sessions at the World Cup Web ....

[Article contains additional citation context not shown here]

J. Mogul, "The Case for Persistent-Connection HTTP", Proceedings of ACM SIGCOMM `95, Cambridge, MA, pp. 299-313, 1995.

.... Points ffl General Web Performance Studies [14, 40] ffl Web Client Behavior Studies [9, 16, 20, 28, 32, 34, 52, 22] ffl Web Server Behavior Studies [3, 4, 5, 6, 7, 12, 13, 26, 29, 45] ffl Web Proxy and Caching Studies [2, 15, 23, 25, 36, 38, 51, 56, 57] ffl Network Effects of Web Traffic [11, 17, 19, 37, 44, 43, 42, 24, 27] ffl Load Generators [1, 8, 10, 18, 39, 46, 54, 55] # Web Server Performance Analysis Paul Barford Page 7 Key Questions for Performance Analysis ffl What performance metrics are important Functional correctness Sizing, tuning and capacity planning ffl How will the test ....

Jeffrey Mogul. The case for persistent-connection HTTP. Technical Report WRL 95/4, DEC Western Research Laboratory, Palo Alto, CA, 1995.

No context found.

J. C. Mogul. The case for persistent-connection HTTP. In Proceedings of ACM SIGCOMM '95, pages 299--313, October 1995.

No context found.

J. C. Mogul. The case for persistent-connection HTTP. In Proceedings of the SIGCOMM'95 conference, Cambridge, MA, Aug. 1995.

No context found.

J. C. Mogul, "The Case for Persistent-Connection HTTP," in Proceedings of the ACM SIGCOMM '95, 1995, pp. 299--313.

No context found.

Jeffrey C. Mogul. The case for persistent-connection HTTP. In Proceedings of the ACM SIGCOMM '95 Conference. ACM, August 1995.

No context found.

Je#rey C. Mogul. The Case for Persistent-Connection HTTP. In Proceedings of the Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication, pages 299--313.

No context found.

J. Mogul. The Case for Persistent-Connection HTTP. ACM Proceeding Sigcomm '95, pages 299--313, August 1995. http://www.research.digital.com/wrl/techreports/abstracts/95.4.html.

No context found.

Jeffrey C. Mogul. The case for persistent-connection HTTP. In Proceedings of the SIGCOMM '95, pages 299-- 313. ACM, August 1995.

No context found.

J. C. Mogul. The case for persistent-connection HTTP. In Proceedings of ACM SIGCOMM '95, pages 299--313, October 1995.

No context found.

J. C. Mogul. The case for persistent-connection HTTP. In Proceedings of ACM SIGCOMM '95, pages 299--313, October 1995.

No context found.

Mogul, J. The case for persistent-connection HTTP. In Computer Communication Review, pages 299-313, October 1995.

No context found.

Jeffrey C. Mogul (1995). "The Case for Persistent-Connection HTTP." Proceedings of the SIGCOMM '95 Conference on Communications Architectures 213 and Protocols: 299-313. http://www.research.compaq.com/wrl/techreports/abstracts/95.4.html

No context found.

Jeffrey C. Mogul, "The case for persistent-connection HTTP," in SIGCOMM, 1995, pp. 299--313.

No context found.

J. Mogul, "The Case for Persistent-Connection HTTP", Proceedings of 1995 ACM SIGCOMM Conference, Cambridge, MA, pp. 299-313, August 28 - September 1, 1995.

No context found.

Mogul, J. The case for persistent connection HTTP. In ACM SIGCOMM, 1995.

First 50 documents  Next 50

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