E. Anderson, "The Magicrouter, an Application of Fast Packet Interposing," www.cs.berkeley.edu/ eanders /projects/magicrouter , May 1996.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....do not maintain state information from one client request to the next. Hence, most existing web server fault tolerance schemes simply detect failures and route future requests to backup servers. Examples of such fault tolerance techniques include the use of specialized routers and load balancers [4, 5, 12, 14] and data replication [6, 28] These methods are unable to recover in progress requests since, while the web server is stateless between transactions, it does maintain important state from the arrival of the first packet of a request to the transmission of the last packet of the reply. With the ....

....for moving the messages between the two levels of proxies and the server. V. RELATED WORK Early work in this field, such as Round Robin DNS [11] and DNS aliasing methods, focused on detecting a fault and routing future requests to available servers. Centralized schemes, such as the Magic Router [4] and Cisco Local Director [12] require request packets to travel through a central router where they are routed to the desired server. Typically the router detects server failures and does not route packets to servers that have failed. The central router is a single point of failure and a ....

E. Anderson, D. Patterson, and E. Brewer, "The Magicrouter, an Application of Fast Packet Interposing," Class Report, UC Berkeley - http://www.cs.berkeley.edu/eanders/projects/magicrouter/ (May 1996).

....NAT device) NAT can be implemented directly in the router and thus not require a separate dispatcher. However, this increases the processing load on the router and may introduce performance bottlenecks (in the router) Two examples of NAT redirection are Cisco Local Director [8] and Magicrouter [2]. There are two existing methods that do not require a dispatcher device, they are Distributed Packet Rewriting (DPR) 4] and ONE IP [11] In DPR each server machine in a cluster has knowledge of the load of all other server machines. IP addresses of all server machines are published, allowing ....

E. Anderson, D. Patterson, and E. Brewer, "The Magic Router, an Application of Fast Packet Interposing", Operating System Design and Implementation, 1996.

....in these distributed Web architectures. In the IBM TCP router [9] all HTTP requests reach the Web switch that distributes them by modifying the destination IP address of each incoming packet: the Web switch replaces its IP address with the private address of the selected Web server. Magicrouter [1], Distributed Packet Rewriting [5] and Cisco LocalDirector [7] are other Web cluster architectures relying on a Web switch that receives the totality of client requests. In particular, Magicrouter is a mechanism of fast packet interposing where a user level process acting as a switchboard ....

E. Anderson, D. Patterson, E. Brewer, "The Magicrouter, an application of fast packet interposing", unpublished Tech. Rep., Computer Science Department, University of Berkeley, May 1996.

....for the user, guarantees backward compatibility with present Web standards, and distributes all client requests to the back end servers. Multiprocessor or cluster architectures with Dispatcher have been adopted with different solutions in various academic and commercial Web farms, e.g. [1, 5, 8, 10, 12]. One of main goals for scalability of a parallel distributed system is the availability of a mechanism that optimally balances the load over the server nodes. Numerous scheduling algorithms were proposed for multi node architectures executing parallel or distributed applications. We want to ....

E. Anderson, D. Patterson, E. Brewer, "The Magicrouter, an application of fast packet interposing", unpublished Tech. Rep. Computer Science Department, University of Berkeley, May 1996.

....reservation processing, and shopping, where erroneous processing or outages are unacceptable. This has motivated the development of fault tolerance techniques for increasing the availability and reliability of Internet services. These techniques have included specialized routers and load balancers [2, 3, 10, 11, 20], data replication [5, 24] client aware server replication [14, 21] and transparent server replication [6, 12, 15, 23] Many Internet services are provided using a three tier architecture, consisting of: a client web browser, one or more replicated front end servers, and one or more back end ....

....of this method since after a failure the client may need to re issue the request several times before it is routed to a new server. Also in this scheme there is no support for recovering requests that were being processed when the failure occurred. Centralized schemes, such as the Magic Router [2] and Cisco Local Director [9, 10] require request packets to travel through a central router where they are routed to the desired server. Typically the router detects server failures and does not route packets to servers that have failed. The central router is a single point of failure and a ....

E. Anderson et al., "The Magicrouter, an Application of Fast Packet Interposing," Class Report - http://www.cs.berkeley.edu/eanders/projects/magicrouter/ (May 1996).

....Connection Routing: Rather than delegating to DNS the responsibility of distributing requests to individual servers in a cluster, several research groups have suggested the use of a local router to perform this function. For example, the NOW project at Berkeley has developed the MagicRouter [11], which is a packet filter based approach [13] to distributing network packets in a cluster. The MagicRouter acts as a switchboard that distributes requests for Web service to the individual nodes in the cluster. To do so requires that packets from a client be forwarded (or rewritten ) by the ....

Eric Anderson, David Patterson, and Eric Brewer. "The MagicRouter: An application of fast packet interposing." Available from http://HTTP.CS.Berkeley.EDU/~eanders/ projects/magicrouter/osdi96-mr-submission.ps, May 1996.

....an inherent fault tolerance flaw in that the location of the service is hard coded into its name, thereby making it difficult to move the service to a new location when a fault occurs. A number of techniques have been developed to address this problem. DNS round robin [13] and specialized routers [16, 5] can be used to hide a number of replicated and or distributed servers behind a single logical name. Smart clients [50] can assist the service in transparent service location in the presence of faults by relying on browser support through Java and JavaScript. Services that rely on the IP ....

....result of the fault. We rely on two main principles in our handling of failures: component restart, and timeouts. We present a mode of fault tolerance that we call starfish fault tolerance. 3.4. 1 Starfish Fault Tolerance Most traditional systems have relied on primary secondary fault tolerance [5, 10] to ensure availability of the system. This involves replicating each component: one of the clones acts as the primary process and participates in the system, while the other is the secondary and simply monitors the primary to mirror its state. If the primary crashes or fails in any other way, the ....

[Article contains additional citation context not shown here]

Anderson, E. The Magicrouter: An Application of Fast Packet Interposing. Class report, UC Berkeley, Dec. 1995.

....real IP addresses. It is intended to be a general purpose solution, capable of handling other services in addition to the web. No discussion is given in the white paper as to what extent the LocalDirector is a bottleneck of the system. Fast Packet Interposing is a user level technique developed in [2] and is used by the MagicRouter to distribute load. The MagicRouter is used to make a cluster of servers appear to have a single IP address without modifications to any servers. Fast Packet Interposing is used to modify network addresses within the data packets that pass through the MagicRouter. ....

Eric Anderson, Dave Patterson, and Eric Brewer. The MagicRouter: An application of fast packet interposing. Submitted for publication.

....To achieve load balance through file caching, requests have to be (re)directed to cache copies whose locations must therefore be determined. A common approach that has been adopted for load balancing over local area networks, performs this redirection by modifying the naming [21] or the routing [1] mechanisms to maintain an explicit directory of cache copies, from which to select suitable destinations for client requests. Frequent and pronounced changes in request patterns can force the locations, and number, of cache copies to be highly transient. The resulting high update rate of the ....

....improvements resulting from load sharing. We address protocol design questions (e.g. we rule out the possibility of clients or servers consulting any cache directory) while Bestavros concentrates on evaluating resource allocation policies. Through minor kernel modification Anderson and Patterson [1] implement user level packet filters to redirect requests to lightly loaded servers. Other approaches [21, 24] use the round robin feature of Domain Name Servers (DNS) to distribute load among a number of Web servers. Our work differs in that it addresses a significantly larger scale of load ....

[Article contains additional citation context not shown here]

E. Anderson and D. Patterson. The Magicrouter: An application of fast packet interposing. Technical report, Univ. of California, Berkeley, EECS, Computer Science Division, May 1996. Available as http://www.cs.berkeley.edu/ eanders/magicrouter/osdi96mr -submission.ps.

.... and computation in the server and a study on the necessity of re balancing after DNS rotation is in [4] Another approach for implementing re assignment is socket forwarding, which avoids the overhead of re connection, but requires significant changes in the OS kernel or network interface drivers [2]. We have not used it for compatibility reasons. 3.1 A cost model for processing task chains For each request, we predict the processing time and assign this request to an appropriate processor. Our cost model for a request is t s = t redirection t data t server t net t client : 1) t ....

E. Anderson, D. Patterson, E. Brewer, "The Magicrouter, an Application of Fast Packet Interposing", submitted to the Second Symposium on Operating System Design and Implementation (OSDI96), 1996. Also available at http://HTTP.CS.Berkeley.EDU/¸eanders/262/262paper.ps.

....(IP SVA) Response, IP SVA) dispatcher Address Document request URL URL Server 1 (Response, IP 1) IP N) Response, IP N) IP 1) Figure 2: Packet double rewriting by the dispatcher. Two architectures that are based on this approach with a server fault detection mechanism are the Magicrouter [1] and the LocalDirector [4] Magicrouter uses a mechanism of fast packet interposing, where a user level process intercepts network packets and modifies them by changing the addresses and checksum fields. Three algorithms can be used to select the Web servers: round robin, random and incremental ....

E. Anderson, D. Patterson, E. Brewer, "The Magicrouter, an application of fast packet interposing ", University of California, Berkeley, May 1996.

....round robin DNS schemes is that as DNS servers know nothing about network topology, server availability or server capacity (powerful multiprocessor or a low end PC) users might be connected to a distant, unavailable, or overloaded server. ffl Server Based IP Level Solutions: Both the Magic router [7] and Local Director [8] fall under this category. In these schemes, services are replicated across a cluster of machines on a single subnet with a modified router. Only one logical IP address is advertised for the site and the router maps the incoming 6 MAGIC ROUTER OR TCP ROUTER SERVER SERVER ....

....the Redirector needs to maintain bindings only for non redirected clients which requires much less memory compared to schemes where the server needs to maintain the mapping for each client accessing the site. ffl This scheme eliminates dog leg routing , a problem both with the Magic Router [7] and Local Director [8] As mentioned in chapter II, the term refers to 25 the scenario where all packets between the client and the actual server are routed through the specialized router in spite of there being a direct, shorter path between the two. Packets flow directly between the client ....

[Article contains additional citation context not shown here]

E. Anderson, D. Patterson, and E. Brewer, "The magicrouter, an application of fast packet interposing." http://www.cs.berkeley.edu/ ¸ eanders/magicrouter/ osdi96-mr-submission.ps, accessed on July 21, 1997. 64

....assignment decisions based on this information, and routes requests to appropriate processors. We did not take this approach mainly because the single central distributor becomes a single point of failure, making the entire system more vulnerable, unless significant low level precautions are taken [APB96]. The current version of SWEB uses a distributed scheduler. The user requests are first evenly routed to SWEB processors via DNS rotation, as indicated in Figure 2. The rotation on available workstation network IDs is in a round robin fashion. This functionality is available in current DNS ....

....another processor. Two approaches, URL redirection or request forwarding, could be used to achieve reassignment and we use the former. Request forwarding, which would involve transferring the TCP IP connection to another server without the client s knowledge, is difficult to implement within HTTP [APB96]. URL redirection gives us excellent compatibility with current browsers and near invisibility to users. Any HTTP request is not allowed to be redirected more than once to avoid the ping pong effect. The functional structure of the scheduler at each processor is depicted in Fig. 3. It contains a ....

E. Anderson, D. Patterson, E. Brewer, "The Magicrouter, an Application of Fast Packet Interposing", submitted to the Second Symposium on Operating System Design and Implementation (OSDI96), 1996. Also available at http://HTTP.CS.Berkeley.EDU/¸eanders/262/262paper.ps.

....round robin DNS schemes is that as DNS servers know nothing about network topology, server availability or server capacity (powerful multiprocessor or a low end PC) users might be connected to a distant, unavailable, or overloaded server. Server Based IP Level Solutions: Both the Magic router [7] and Local Director [8] fall under this category. In these schemes, services are replicated across a cluster of machines on a single subnet with a modified router. Only one logical IP address is advertised for the site and the router maps the incoming client requests to one of the hidden servers ....

....each client maintains its own list of redirection bindings, the Redirector needs to maintain bindings only for non redirected clients which requires much less memory compared to schemes where the server needs to maintain the mapping for each client accessing the site. Unlike the Magic Router [7], this scheme does not place any topological restriction on the servers to be geographically localized. Since redirection is performed for each client, this scheme eliminates the problem of cached bindings being used by multiple clients causing skewed loads on the servers, as is the case in DNS ....

E. Anderson, D. Patterson, and E. Brewer, "The magicrouter, an application of fast packet interposing." http://www.cs.berkeley.edu/ ¸ eanders/magicrouter/osdi96-mr-submission.ps, accessed on July 21, 1997.

.... [Sar96] presents preliminary concepts of a Java based system like Javelin, identifying fundamental issues related to incentives (for hosts) efficiency, and security (for clients) Our work has benefited from the more mature networked computer projects, such as Network of Workstations(NOW) AP96] PVM [Sun90] MPI [MPI94] WebOS [VDA] Nexus [Tri89] Linda [WL88] Piranha [Sch92] and the Legion project [GWF 94] These earlier projects can be distinguished from December 17, 1996 01 : 55 DRAFT 14 Javelin since they either require the user to have login access to all machines ....

E. Anderson and D. A. Petterson. The Magic Router: An Application of Fast Packet Interposing. submitted to OSDI '96, 1996.

....locate the service depends largely on the service itself. Web based services use Uniform Resource Locators (URLs) as the primary service location mechanism. Although this has fault tolerance implications, techniques such as DNS round robin (Brisco 1995) and specialized routers (Cisco Systems 1996, Anderson 1995) can be used to hide a number of replicated and or distributed servers behind a single logical name. Smart clients (Yoshikawa, Chun, Eastham, Anderson Culler 1997) can assist the service in transparent service location in the presence of faults by relying on browser support through Java and ....

....independent of the others. We rely on two main principles in our handling of failures: component restart and timeouts used by a mode of fault tolerance that we call starfish fault tolerance. 3.4. 1 Starfish Fault Tolerance Most traditional systems have relied on primary secondary fault tolerance (Anderson 1995, Bartlett 1981) to ensure availability of the system. This involves replicating each component: one of the clones acts as the primary process and participates in the system, while the other is the secondary and simply monitors the primary to mirror its state. If the primary crashes or fails in ....

[Article contains additional citation context not shown here]

Anderson, E. (1995), The Magicrouter: An Application of Fast Packet Interposing. Class report, UC Berkeley.

....packet in such a way that all packets stemming from one request go to the same server host, but successive requests are mapped to successive hosts in a round robin fashion; this allows equality in work assignment at a very fine time scale. TCP routers have been employed in experimental Web servers [1, 10, 22]. Such routers may introduce delays because they must modify each packet that flows through them. However, because they achieve better load balance, they are commonly proposed for high performance distributed Web servers [10, 22] and a number of commercial products have appeared that implement ....

E. Anderson, D. Patterson, and E. Brewer. The magicrouter: An application of fast packet interposing. Technical report, UC Berkeley, 1996.

....a number of IP addresses for a single domain name. Similarily, there are many ways to perform the second mapping (from initial IP address to actual host) For example, this mapping could be done at the application level, using the HTTP redirection approach [1] or using a dispatcher at the server [2, 17]. While initial attempts to construct scalable Web servers focussed on using the mapping from domain names to IP addresses [10] recent attempts have focussed on the second kind of mapping (IP addresses to hosts) because of the potential for finer control of load distribution. One common feature ....

....to hosts) because of the potential for finer control of load distribution. One common feature of all of these attempts (whether proposed or implemented) is that a centralized mechanism is employed to perform the mapping from IP addresses to hosts. Examples include the Berkeley MagicRouter [2], the Cisco Local Director [17] and IBM s TCP Router [6] and Network Dispatcher [4] In contrast, DPR is a technique that allows the mapping between IP address and host to be implemented in a distributed, efficient, and scalable fashion. In particular, DPR can be viewed as a distributed method of ....

[Article contains additional citation context not shown here]

Eric Anderson, David Patterson, and Eric Brewer. The MagicRouter: An application of fast packet interposing. submitted to OSDI 1996, May 1996.

....in section 7. 2 Alternative Solutions In this section, we will describe approaches for adding transparent fault tolerance and load balancing to Web services. Popular approaches currently include DNS Aliasing [Brisco 1995, Katz et al. 1994] HTTP redirect [Berners Lee 1995] Magic Routers [Anderson et al. 1996], fail safe TCP [Goldstein Dale 1995] and Active Networks [Wetherall Tennenhouse 1995] Figure 1 describes how Smart Clients fits in the space of existing solutions. We will describe each of the existing solutions in turn leading to our description of the Smart Client architecture in Section ....

E. Anderson, D. Patterson, and E. Brewer. "The Magicrouter, an Application of Fast Packet Interposing". May 1996. Submitted For Publication. Also see http://HTTP.CS.Berkeley.EDU/~eanders/magicrouter/.

....5 evaluates our requirements above in the context of the Smart Client architecture. Section 6 describes related work, and Section 7 concludes. 2 Alternative Solutions Existing architectures include DNS Aliasing [Brisco 1995, Katz et al. 1994] HTTP redirect [Berners Lee 1995] Magic Routers [Anderson et al. 1996], failsafe TCP [Goldstein Dale 1995] and Active Networks [Wetherall Tennenhouse 1995] Figure 1 describes how Smart Clients fits in the space of existing solutions. We will describe each of the existing solutions in turn leading to a description of the Smart Client architecture. A number of ....

E. Anderson, D. Patterson, and E. Brewer. "The Magicrouter, an Application of Fast Packet Interposing". May 1996. Submitted For Publication. Also see http://HTTP.CS.Berkeley.EDU/eanders- /magicrouter/.

No context found.

E. Anderson, "The Magicrouter, an Application of Fast Packet Interposing," www.cs.berkeley.edu/ eanders /projects/magicrouter , May 1996.

No context found.

E. Anderson, D. Patterson and E. Brewer (1996). "The Magicrouter: An Application of Fast Packet Interposing." Second Symposium on Operating Systems Design and Implementation. http://www.cs.berkeley.edu/~eanders/projects/magicrouter/

No context found.

Anderson, Eric, The Magic Router: An application of fast packet interposing, available from http://HHTP.CS.BERKELY.edu/~eanders/projects/magicrouter/osdi96-mrsubmission. ps, May 1996.

No context found.

E. Anderson, D. Patterson, and E. Brewer, The MagicRouter: an Application of Fast Packet Interposing". In Proceedings of OSDI, October 1996. 88

No context found.

Eric Anderson, David Patterson, and Eric Brewer. "The MagicRouter: An application of fast packet interposing." Available from http://HTTP.CS.Berkeley.EDU/~eanders/projects/magicrouter/osdi96-mrsubmission. ps, May 1996.

First 50 documents

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