Dan Mosedale, William Foss, and Rob McCool. Administering very high volume internet services. In LISA IX: System administration conference, pages 95--102, 1995.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....improve the netnews performance. Web server load balancing also attempts to adapt around slow nodes in a number of ways. Prior to the existence of load balancing routers, Netscape used a cluster of heterogeneous machines to serve web pages and performed asymmetric load balancing in the clients [MFM95] In fact load balancing routers [Cis00] and the follow on improvements [PAB 98] distribute web accesses across a set of servers, potentially incorporating server load information. Web server load balancing could be extended to include the DQ idea of getting all the related parts of a query to ....

Dan Mosedale, William Foss, and Rob McCool. Administering very high volume internet services. In LISA IX: System administration conference, pages 95--102, 1995.

....1.2 1.4 1.5 Table 4: Bursty MRPS for processing wavelets. 5.3 Load balancing with hot spots Hot spots is a typical problem with DNS rotation, where a single server exhibits a higher load than its peers. Various authors have noted that DNS rotation seems to inevitably lead to load imbalances [18, 19]. We examine how our system deals with hot spots by sending a fixed number of requests to a subset of nodes in our server cluster, giving a wide range of load disparities. Without our scheduler, the selected nodes would have to process all of those requests. The scheduler can effectively deal with ....

....we have theoretical MRPS predictions in Section 4. We ran experiments to determine the actual MRPS by testing for a period of 120 seconds and choosing the highest RPS such that the server response times are reasonable and no requests are dropped. We chose the period of 120 seconds based on [13, 19], which indicate most long bursts on the Internet are actually relatively short. Thus the sustained RPS required in practice are for a period shorter than 1. For this experiment, the clients are simulated within the Meiko machine. The aggregated bandwidth (B s ) of the 6 node server to the ....

D. Mosedale, W. Foss, R. McCool, "Administering Very High Volume Internet Services", 1995 LISA IX, Monterey, CA, 1995.

....we have theoretical MRPS predictions in Section 4. We ran experiments to determine the actual MRPS by testing for a period of 120 seconds and choosing the highest RPS such that the server response times are reasonable and no requests are dropped. We chose the period of 120 seconds based on [11, 16], which indicates most long bursts on the Internet are actually relatively short. Thus the sustained RPS required in practice are for a period shorter than 1. Notice that for this experiment, the clients are simulated within the Meiko machine. The aggregated bandwidth (Bs) of the 6 node server ....

D. Mosedale, W. Foss, R. McCool, "Administering Very High Volume Internet Services", 1995 LISA IX, Monterey, CA, 1995.

....1.4 1.5 Table 3: Bursty MRPS for processing wavelets on Meiko. Load balancing with hot spots . Hot spots is a typical problem with DNS rotation, where a single server exhibits a higher load than its peers. Various authors have noted that DNS rotation seems to inevitably lead to load imbalances [12, 13]. We examine how our system deals with hot spots by sending a fixed number of requests to a subset of nodes in our server cluster, giving a wide range of load disparities. Without our scheduler, the selected nodes would have to process all of those requests. The scheduler can effectively deal with ....

D. Mosedale, W. Foss, R. McCool, "Administering Very High Volume Internet Services", 1995 LISA IX, Monterey, CA, 1995.

....degree of name caching which occurs. DNS caching enables a local DNS system to cache the name to IP address mapping, so that most recently accessed hosts can quickly be mapped. The downside is that all requests for a period of time from a DNS server s domain will go to a particular IP address [MFM95]. It should be noted that WWW applications only represent a special class of Internet information systems. There are other information servers dealing with huge file sizes and large numbers of users, for example multi media servers [HL95 ] Our situation has several differences. First, our current ....

....maximum sustained 17.8 RPS for 1.5M files on Sparc Elan, consistent with the 16 RPS achieved in practice. It should be noted that in practice requests come in periodic bursts. Thus we use a 30 second test period in the rest of experiments, representing a non trivial but limited burst of requests [MFM95]. Response time and drop rate. In Table 2, we report the response time (the time from after the client sends a request until the completion of this request) when we vary the number of server nodes. The system starts to drop requests if the server reaches its RPS limit. In general, the performance ....

D. Mosedale, W. Foss, R. McCool, "Administering Very High Volume Internet Services", Proc. of 1995 LISA IX, Monterey, CA, September, 1995.

....16 10 10 h alpha = 1.5 and = 0.8 Slowdown factor Waiting time factor Figure 7: Tradeoff Regions for SITA V. evolved from a 6 processor tightly coupled multiprocessor serving about 5 million hits per day, to a distributed system with more than 20 hosts serving more than 100 million hits per day [12, 25, 26]. Recent performance measurements of the Web and Web servers show that Web file sizes exhibit heavy tailed distributions, leading to very high variability [2, 6] As a result, practical resource allocation methods for Web servers must address the basic problems that arise from highly variable file ....

Dan Mosedale, William Foss, and Rob McCool. Administering very high volume internet services. Available at http://www.keynote.com/techrpts/nspaper.html.

....0.5 1 1.5 2 p=6 142 164 192 679 4029 p=5 96 183 191 1838 7893 Table 1: Response time improvement ratio with and without client resource. Load balancing with hotspots. Hotspots are a typical problem with DNS rotation, where a single server node exhibits a higher load than its peers [13, 14]. We examine how our system deals with hotspots. We directed a fixed number of requests to a subset of nodes in our server cluster, giving a wide range of load disparities. Without our scheduler, those nodes would have to process all of those requests. Our scheduling algorithm can effectively deal ....

D. Mosedale, W. Foss, R. McCool, "Administering Very High Volume Internet Services", 1995 LISA IX, Monterey, CA, September, 1995.

....maximum RPS with and without client resources for wavelets. 6.3 Load balancing with hot spots Hot spots is a typical problem with DNS rotation, where a single server exhibits a higher load than its peers. Various authors have noted that DNS rotation seems to inevitably lead to load imbalances [22, 24]. We examine how our system deals with hot spots by sending a fixed number of requests to a subset of nodes in our server cluster, giving a wide range of load disparities. Without our scheduler, the selected nodes would have to process all of those requests. The SWEB scheduler can effectively ....

D. Mosedale, W. Foss, R. McCool, "Administering Very High Volume Internet Services", 1995 LISA IX, Monterey, CA, September, 1995.

No context found.

Dan Mosedale, William Foss, and Rob McCool. Administering very high volume internet services. In Proceedings of LISA IX, pages 95--102, September 1995.

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