A.C. Snoeren, D.G. Andersen, and H. Balakrishnan. Fine-grained failover using connection migration. In Proc. 3rd USENIX Symp. on Internet Technologies and Systems (USITS), pages 221--232, San Francisco, California, USA, March 2001. 40  Home/Search   Document Details and Download   Summary   Related Articles   Check

....that the data is manipulated on high performance machines that are part of the data Grid forming the back end instead of the mobile machines that are the front end. Providing mobile access to distributed computing resources and in particular web oriented resources is a well examined topic [2, 7, 8]. However, supporting mobile interfaces in the Grid context provides new challenges and opportunities. Following are some of the major research issues in this context: One of the main issues handled in Grid computing systems is resource heterogeneity. Introduction of mobile devices requires an ....

SNOEREN, A. C., ANDERSEN, D. G., AND BALAKR- ISHNAN, H. Fine-grained failover using connection migraton. In 3rd USENIX Symposium on Internet Technologies and Systems (Mar. 2001).

....paper also shows that an approach using TCP handoff of persistent connections among back end servers may be competitive with frontend server switching. However, the study does not explore sensitivity of this scheme to the connection migration cost. A more recent study of TCP connection migration [17] suggests that this cost is too high for request distribution in Web server clusters, although it is a useful approach for failover. In short, the relative merits of the various solutions to the persistent connection problem for fine grained request routing such as URL switching and other ....

Alex C. Snoeren, David G. Andersen, and Hari Balakrishnan. Fine-grained failover using connection migration. In Proc. of the Third Annual USENIX Symposium on Internet Technologies and Systems (USITS), 2001.

....Due to lack of space, we only provide a very brief overview of selected related projects. Two classes of projects are examined: projects that are related to the overall concept [BaS99, BeW97, Der99, FoK01, EsH99, KoC00, VaA98] and ones that are related to specific functionalities of the InviNet [ArO99, LaL99, SnB00, SnA01]. The 2K [KoC00] is a network operating system in which all entities (users, devices, etc) exist in the network and are represented by CORBA objects. One of the unique features of 2K is it reconfigures automatically such that at any given time only components absolutely needed by the applications ....

A. C. Snoeren, D. G. Andersen, and H. Balakrishnan, "Fine-grained failover using connection migraton," 3rd USITS, Mar. 2001.

....a failed server from a log. This requires storage of all communication between the client and the server for each logged connection. Furthermore, replay of this communication implies that the recovery process does not scale well. Another approach for migrating TCP connections is described in [21]. Unlike our technique, this approach is more suitable in cases where the backup server is located across a wide area network from the primary server. This approach, however, has a number of limitations. First, it requires a modified TCP stack which implements a protocol for connection migration. ....

Alex C. Snoeren, David G. Andersen, and Hari Balakrishnan. Fine-grained failover using connection migration. In Proc. of the Third Annual USENIX Symposium on Internet Technologies and Systems (USITS), March 2001. 12

....state, and limits the rollback and replay to restore only a small part of process state, specific to exactly one client. Providing high availability for Internet services through protocol support has been approached in several ways: using connection migration in application specific solutions [22, 31], fault tolerance for TCP [1] and new transport protocols [24] None of these approaches provides the OS support required for migrating and resuming complex, multi process service instances. A scheme that enables HTTP connection endpoints to migrate within a pool of support servers is described ....

....fault tolerance for TCP [1] and new transport protocols [24] None of these approaches provides the OS support required for migrating and resuming complex, multi process service instances. A scheme that enables HTTP connection endpoints to migrate within a pool of support servers is described in [22]. Migration is supported by broadcasting perconnection HTTP and TCP state within the server pool. The scheme adds an HTTP aware module at the transport layer that extracts information from the application data stream to be used for connection resumption. While this achieves transparency, it ....

[Article contains additional citation context not shown here]

A. C. Snoeren, D. G. Andersen, and H. Balakrishnan. FineGrained Failover Using Connection Migration. In Proc. 3rd USENIX Symp. on Internet Technologies and Systems (USITS), Mar. 2001.

....host. The protocol ensures that the destination server resumes the service while preserving the exactly once delivery semantics across migration, without freezing or otherwise disrupting the trac on the connection. Although ne grained connection migration solutions have been proposed for HTTP [24, 31] by exploiting details of the protocol, to our best knowledge M TCP is the rst solution that provides generic migration support through a TCP compatible transport protocol. The goal of this paper is to demonstrate that M TCP is a viable solution in building highly available services over the ....

....Finally, Section 8 concludes the paper. 2 Related Work High availability of Internet services through transport layer support has been approached in several ways: fault tolerance for TCP [4] new protocols like SCTP [27] and using connection migration in applicationspeci c solutions [24, 31]. Connection hando protocols have been used in several mobility extensions to TCP [6, 7, 25] but their relevance to highly available services is marginal. A fail over scheme that enables HTTP connection endpoints to migrate within a pool of support servers is described in [24] Migration takes ....

[Article contains additional citation context not shown here]

A. C. Snoeren, D. G. Andersen, and H. Balakrishnan. Fine-Grained Failover Using Connection Migration. In Proc. 3rd USENIX Symp. on Internet Technologies and Systems (USITS), Mar. 2001.

....sound futuristic, the need for systems dynamic self modification is already well established, and existing systems already demonstrate some of these capabilities. In the embedded domain, some cellphones can save power by dynamically switching from cell modem to Bluetooth based communications [26]. The continuing merger of PDAs and cellphones causes the development of new techniques for the dynamic extension of such devices capabilities, exemplified by the ability to download plugins for phone resident browsers and more importantly, by the runtime deployment of image filters to cellphones ....

A. Snoeren, D. Anderson, and H. Balakrishnan. Finegrained failover using connection migration. In Proceedings of the Third Annual USENIX Symposium on Internet Technologies and Systems (USITS), March 2001.

....migration module to migrate the network connection to the new resource. The semantics of migrating the network connection from one resource to another depend on both the stateful ness of the service being accessed and the reliability guarantees offered by the underlying message transport protocol [13]. Migration of an unreliable network connection to a stateless service is accomplished by simply closing the old network connection and opening a fresh connection to the new resource. However, additional support is required for migrating reliable connections and for managing stateful services ....

....[13] Migration of an unreliable network connection to a stateless service is accomplished by simply closing the old network connection and opening a fresh connection to the new resource. However, additional support is required for migrating reliable connections and for managing stateful services [13]. Migration of a reliable connection requires support for preserving the sequence of messages across migration, while a connection to a stateful service can be migrated transparently across resources only when the state accessed at the old resource is also available at the new resource in the form ....

Alex C. Snoeren et al. Fine-Grained Failover Using Connection Migration, Proc. 3rd USENIX USITS, March 2001.

....is inapplicable to the already deployed applications. Several projects have explored the idea of socket level recovery, where the failure is hidden from the client by some lower layer that reestablishes connections when necessary and provides a reliable socket to the application. One such system [11] extends the TCP protocol with an option that enables migration of connections from one host to another. Among other things, this allows the service provider to ask the client TCP stack to migrate a failed connection to a backup. A similar approach was adopted by [12] but it requires the server ....

A. Snoeren, D. Andersen, and H. Balakrishnan. Finegrained failover using connection migration. In Proc. 3rd USENIX Symp. on Internet Technologies and Systems (USITS), pages 97--108, 2001.

....per request basis to the server where the request is best serviced. We obtain the coordination needed to allow the control pipe and the data pipe to reside on different nodes through a simple protocol, which we call split stack. Half pipe anchoring goes beyond existing connection handoff protocols [4, 5, 19, 24, 25, 29, 30, 31] in at least two respects. First, it supports efficient multiple handoffs of the same connection by allowing requests to be pipelined. Thus, a node processing request can hand off the connection to a new node for processing the next request before all TCP traffic related to has been ....

....this approach exposes the configuration details of the server cluster to the content authoring process (and hence the clients) and prevents the cluster from achieving finegrained load balancing across servers. The second end to end approach uses a transport level connection migration mechanism [29, 30] to migrate a transport connection between a client and a server to another server better suited to service the next request. The client application is transparent to this migration. This approach incurs large connection migration latencies because of WAN delays and hence is not suitable for ....

A. Snoeren, D. Andersen, and H. Balakrishnan. Finegrained failover using connection migration. In USITS, 2001.

....content can be moved to an alternative server within the HRS. For small data files, connections that were established prior to the move are allowed to terminate naturally. For large data files and streaming files, one can also consider the option of migrating the transmission to a new server [32]. 4.2 Replication in the P2P approach Users will also be able to control the extent to which their machine is used by the service. Again, we are interested in developing resource bounding techniques. To provide users with a high degree of control over their interaction with the service, the ....

....to maintain high performance levels. Previous work dealing with connection migration has addressed the challenge of seamlessly transferring an ongoing connection from one server to another without significantly deteriorating the performance of the transfer from the perspective of the client [32]. What has not been addressed is the issue of identifying an appropriate server to which the connection can be transfered. We are interested in performing two studies that relate to connection migration. The first is an experimental study that measures the potential performance gains that can be ....

A. Snoeren, D. Andersen, and H. Balakrishnan. Fine-Grained Failover Using Connection Migration. In Proceedings of the 3rd Usenix USITS'01, San Francisco, CA, March 2001.

....above, the client never receives complete replies to the in progress requests and has no way to determine whether or not a requested operation has been performed [1, 15, 16] see Figure 1) Some recent work does address the need for handling inprogress transactions. Client aware solutions such as [16, 23, 26] require modifications to the clients to achieve their goals. Since many versions of the client software, the Client Web Server Back end 1 2 3 4 Figure 1: If the web server fails before sending the client reply (step 4) the client can not determine whether the failure was before or after the ....

....be seen as an alternative implementation of the write once registers which also provides client transparency. Zhao et al. [29] describe a CORBA based infrastructure for replication in three tier systems which deal with the same issues, but again is not client transparent. The work by Snoeren et al. [26] is another example of a solution that is not transparent to the client. A transport layer protocol with connection migration capabilities, such as SCTP or TCP with proposed extensions, is used along with a session state synchronization mechanism between servers to achieve connection level ....

A. C. Snoeren, D. G. Andersen, and H. Balakrishnan, "Fine-Grained Failover Using Connection Migration," Proceedings of the 3rd USENIX Symposium on Internet Technologies and Systems, San Francisco, California (March 2001).

....on a per request basis to the server where the request is best serviced. We obtain the coordination needed to allow control pipe and data pipe to reside on different nodes through a simple protocol, which we call split stack. Half pipe anchoring goes beyond existing connection handoff protocols [16, 17, 18, 19, 20, 21, 22] in at least two respects. First, it supports efficient multiple handoffs of the same connection by allowing requests to be pipelined a node processing request r 1 can hand off the connection to a new node for the processing of the next request r 2 before all TCP traffic related to r 1 has been ....

....this approach exposes the configuration details of the server cluster to the content authoring process (and hence the clients) and prevents the cluster from achieving fine grained load balancing across servers. The second end to end approach uses a transport level connection migration mechanism [20, 21] to migrate a transport connection between a client and a server to another server better suited to service the next request. To realize this, a server informs the transport level connection migration mechanism at the client of the address of the new server; in response, the mechanism migrates ....

Alex C. Snoeren, David G. Andersen, and Hari Balakrishnan, "Fine-grained failover using connection migration," in Proc. of 3rd USENIX Symposium on Internet Technologies and Systems (USITS), 2001.

No context found.

SNOEREN,A.C.,ANDERSEN,D.G.,AND BALAKRISH- NAN, H. Fine-grained failover using connection migration. In Proc. 3rd USITS (Mar. 2001), pp. 221--232.

....with only one attachment point. Under the traditional IP addressing model, all of these end points share the same IP address; hence, network layer mobility schemes would require these end points to move in concert. Increasing support for application migration [38, 63, 83] and service redirection [85, 119] ensures that the end points may in fact move independently, however. This fine grained mobility requires each end point to have its own EID or home address, resulting in an increase in the number of addresses that must be managed by the IP routing infrastructure and severely stressing the scaling ....

....establishment between the new attachment points. We present a TCP connection migration scheme in Chapter 4 that remedies these deficiencies as well as several others. After its original development as an approach to handle host mobility [120] and subsequent extension to support service fail over [119], other researchers adapted (e.g. MTCP [129] and extended (e.g. support for concurrent migration [136] our approach. 2.3.3 New transport protocols A number of proposed transport protocols have begun to incorporate rudimentary support for mobile or even multi homed end points but have not yet ....

[Article contains additional citation context not shown here]

Alex C. Snoeren, David G. Andersen, and Hari Balakrishnan. Fine-grained failover using connection migration. In Proc. 3rd USENIX Symposium on Internet Technologies and Systems, pages 221--232, San Francisco, California, March 2001.

No context found.

A.C. Snoeren, D.G. Andersen, and H. Balakrishnan. Fine-grained failover using connection migration. In Proc. 3rd USENIX Symp. on Internet Technologies and Systems (USITS), pages 221--232, San Francisco, California, USA, March 2001. 40

No context found.

A.C. Snoeren, D.G. Andersen, and H. Balakrishnan. Fine-grained failover using connection migration. In Proc. USENIX USITS: Symp. Internet Technol. Syst., 2001.

No context found.

A. C. Snoeren, D. G. Andersen, and H. Balakrishnan. Fine-grained failover using connection migration. In Proc. USENIX USITS: Symp. Internet Techn. Syst., 2001.

No context found.

A.C. Snoeren, D.G. Andersen, and H. Balakrishnan. Finegrained failover using connection migration. In Proc. USENIX USITS: Symp. Internet Technol. Syst., 2001.

No context found.

A. C. Snoeren, D. G. Andersen, and H. Balakrishnan, "Fine-grained failover using connection migration," in Proc. of 3rd USENIX Symposium on Internet Technologies and Systems (USITS), 2001.

No context found.

A. Snoeren, D. Andersen, and H. Balakrishnan. Fine-grained failover using connection migration. In Proceedings of USENIX USITS, San Francisco, CA, USA, Mar. 2001.

No context found.

A. C. Snoeren, D. G. Anderson, and H. Balakrishnan. Fine-grained failover using connection migration. In Proc. 3rd USENIX Symp, on Internet Technologies and Systems, pages 221--232, San Francisco, CA, 2001.

No context found.

A. C. Snoeren et al., "Fine-Grained Failover Using Connection Migration," 3rd USENIX Symposium on Internet Technologies and Systems, San Francisco, California, pp. 221-232 (March 2001).

No context found.

Alex C. Snoeren, David G. Andersen, and Hari Balakrishnan, "Fine-grained failover using connection migration," in Proc. of 3rd USENIX Symposium on Internet Technologies and Systems (USITS), 2001.

No context found.

A. C. Snoeren, D. G. Andersen, and H. Balakrishnan. FineGrained Failover Using Connection Migration. In Proc. 3rd USENIX Symp. on Internet Technologies and Systems (USITS), Mar. 2001.

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