J-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. In Proceedings of the International Conference on Dependable Systems and Networks, pages 374--383, June 2002.  Home/Search   Document Details and Download   Summary   Related Articles   Check

No context found.

J-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. In Proceedings of the International Conference on Dependable Systems and Networks, pages 374--383, June 2002.

No context found.

J-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. In Proc. of the Intl. Conference on Dependable Systems and Networks (DSN), pages 374--383, June 2002.

No context found.

J-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. In Proceedings of the International Conference on Dependable Systems and Networks, pages 374--383, June 2002.

No context found.

J-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. In Proceedings of the International Conference on Dependable Systems and Networks, pages 374--383, June 2002.

No context found.

J-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. In Proc. of the Intl. Conference on Dependable Systems and Networks (DSN), pages 374--383, June 2002.

No context found.

J-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. In Proc. of the Intl. Conference on Dependable Systems and Networks (DSN), pages 374--383, June 2002.

....are guaranteed to end eventually. Section 4 shows that as a result, our protocol cannot be adapted to unreliable networks. Instead, we adapt the original protocols of Malkhi and Reiter to this more general model. Because of space constraints, we refer the reader to our extended technical report [22] for the correctness proof of the SBQ protocols. 4. Network models Both the MR and the SBQ protocols assume a reliable asynchronous network, that is for any pair of correct machines A and B, if A sends a message, then B is guaranteed to eventually receive it. In some systems, reliable ....

....data) Because the choice of the value of t is left to the implementor, S SBQ can either use as few servers as Bazzi s protocol or always be self timing like SBQ. More interestingly, its performance can be adjusted to any intermediate scenario. Due to space constraints, we refer the reader to [22] for the complete description of the S SBQ protocol. Note that even though the discussion of the previous paragraph was limited to the threshold case, S SBQ uses a more general failure model that includes not only a fail prone system but also a new delay prone system to describe the conditions ....

[Article contains additional citation context not shown here]

J.-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. Technical report, University of Texas at Austin, Department of Computer Sciences, December 2001.

....that the resulting protocol is non confirmable. The bottom two lines of Table i indicate that the SBQ L protocol can be modified to be non confirmable. In that configuration, it can provide regular semantics for generic data using only 2f 1 servers instead of the 3f 1 required in prior work [15]. We again show that our protocol is optimal by proving that 2f i servers are required to provide even safe semantics for non confirmable writes. The existence of the SBQ L protocol shows that this bound is tight. This result shows that the distinction between confirmable and non confirmable ....

....protocols is fundamental. Table 1. Required number of servers and semantics for various protocols for Byzantine distributed shared memory. New results and improvements over previous protocols are shown in bold Required Semantics Existing Protocols SBQ L Safe Semantics af l, safe [11, 12] [15] conf. generic 4f l, partial atomic [17] 2 3f kl, atomic 2 3f ] I 3f 1, regular [11] 15] servers conf. self verifying 3f 1, atomic [12] 5] 3f 1, atomic non conf. generic 3f 1, safe [15] 2fq 1, regular 2fq 1 non conf. self verifying 2f 1, regular [15] 2f 1, regular z servers ....

[Article contains additional citation context not shown here]

J-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. In Proceedings of the International Conference on Dependable Systems and Networks, pages 374-383, June 2002.

....that the resulting protocol is non confirmable. The bottom two lines of Table i indicate that the SBQ L protocol can be modified to be non confirmable. In that configuration, it can provide regular semantics for generic data using only 2f i servers instead of the 3f i required in prior work [14]. We again show that our protocol is optimal by proving that 2f i servers are required to provide even safe semantics for non confirmable writes. The existence of the SBQ L protocol shows that this bound is tight. This result shows that the distinction between confirmable and non confirmable ....

....provide even safe semantics for non confirmable writes. The existence of the SBQ L protocol shows that this bound is tight. This result shows that the distinction between confirmable and non confirmable protocols is fundamental. Existing Protocols SBQ L Bound on server count 4f 1, safe [11, 1212,[14] confirmable, generic 4f l, partial atomic [16] 2 3fq 1, atomic 2 3f ] I for safe 3f 1, regular [11] 14] or stronger semantics confirmable, self verifying 3f 1, atomic [12] 5] 2 3f 1, atomic 2 non confirmable, generic 3f 1, safe [14] 2f I l, regular 2f ] I for safe non confirmable, ....

[Article contains additional citation context not shown here]

J-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. In Proceedings of the International Conference on Dependable Systems and Networks, pages 374-383, June 2002.

....are guaranteed to end eventually. Section 4 shows that as a result, our protocol cannot be adapted to unreliable networks. Instead, we adapt the original protocols of Malkhi and Reiter to this more general model. Because of space constraints, we refer the reader to our extended technical report [22] for the correctness proof of the SBQ protocols. 4. Network models Both the MR and the SBQ protocols assume a reliable asynchronous network, that is for any pair of correct machines and , if sends a message, then is guaranteed to eventually receive it. In some systems, reliable ....

....data) Because the choice of the value of is left to the implementor, S SBQ can either use as few servers as Bazzi s protocol or always be self timing like SBQ. More interestingly, its performance can be adjusted to any intermediate scenario. Due to space constraints, we refer the reader to [22] for the complete description of the S SBQ protocol. Note that even though the discussion of the previous paragraph was limited to the threshold case, S SBQ uses a more general failure model that includes not only a fail prone system but also a new delay prone system to describe the conditions ....

[Article contains additional citation context not shown here]

J.-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. Technical report, University of Texas at Austin, Department of Computer Sciences, December 2001.

....that the resulting protocol is non con rmable. The bottom two lines of Table 1 indicate that the SBQ L protocol can be modi ed to be non con rmable. In that con guration, it can provide regular semantics for generic data using only 2f 1 servers instead of the 3f 1 required in prior work [14]. We again show that our protocol is optimal by proving that 2f 1 servers are required to provide even safe semantics for non con rmable writes. The existence of the SBQ L protocol shows that this bound is tight. This result shows that the distinction between con rmable and non con rmable ....

....even safe semantics for non con rmable writes. The existence of the SBQ L protocol shows that this bound is tight. This result shows that the distinction between con rmable and non con rmable protocols is fundamental. Existing Protocols SBQ L Bound on server count 4f 1, safe [11, 12] [14] con rmable, generic 4f 1, partial atomic [16] 3f 1, regular [11] 14] or stronger semantics con rmable, self verifying 3f 1, atomic [12] 5] 1;2 non con rmable, generic 3f 1, safe [14] 2f 1, regular non con rmable, self verifying 2f 1, regular [14] 2f 1, regular or ....

[Article contains additional citation context not shown here]

J-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. In Proceedings of the International Conference on Dependable Systems and Networks, pages 374-383, June 2002.

....at least 2f 1 servers to provide safe semantics. Third, non blocking protocols with fewer than 3f 1 servers cannot provide atomic semantics. Our second contribution is a new protocol, SBQ L, in two variants: blocking and non blocking. As shown Existing Protocols SBQ L 3f 1, regular [10] [14] ; blocking, self verifying 3f 1, atomic [11] 5] 1;2 4f 1, safe [10, 11] 14] blocking, generic 4f 1, partial atomic [15] non blocking, self verifying 2f 1, regular [14] 2f 1, regular (tight) non blocking, generic 3f 1, safe [14] 2f 1, regular (tight) 1) Does not require ....

....than 3f 1 servers cannot provide atomic semantics. Our second contribution is a new protocol, SBQ L, in two variants: blocking and non blocking. As shown Existing Protocols SBQ L 3f 1, regular [10] 14] blocking, self verifying 3f 1, atomic [11] 5] 1;2 4f 1, safe [10, 11] [14] blocking, generic 4f 1, partial atomic [15] non blocking, self verifying 2f 1, regular [14] 2f 1, regular (tight) non blocking, generic 3f 1, safe [14] 2f 1, regular (tight) 1) Does not require reliable channels. 2) Tolerates faulty clients. Table 1: Required number of servers and ....

[Article contains additional citation context not shown here]

J-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. Technical Report TR-02-01, University of Texas at Austin, Department of Computer Sciences, January 2002.

....quorum systems The following lemmas, that we give without proof, establish that the SBQ protocol for a dissemination quorum systems given in Figure 2 implements a regular variable. The proofs are very similar to those given above for a masking quorum systems and safe variables and can be found in [27]. recently written value. Lemma 5 A read operation that is concurrent with one or more write operations returns either the most recently written value, or one of the values being written by the concurrent write operations. 10 4. Network models Both the MR and the SBQ protocols assume a ....

....of this protocol can bound network retransmission buffer consumption. This protocol is a straightforward extension of Malkhi and Reiter s protocol for asynchronous reliable networks [24] Due to space constraints, we summarize the protocol and its properties in this section. We refer the reader to [27] for a full statement of the protocol as well as proofs for the theorems and lemmas stated in this section. Although the model used by Malkhi and Reiter s original protocol ensures that all correct servers receive all transmitted messages, the protocol itself only relies on a quorum of servers ....

[Article contains additional citation context not shown here]

J-P. Martin, L. Alvisi, and M. Dahlin. Small byzantine quorum systems. Technical report, University of Texas at Austin, Department of Computer Sciences, December 2001.

No context found.

J.-P. Martin, L. Alvisi, and M. Dahlin. Small Byzantine quorum systems. In Dependable Systems and Networks, DSN 01, 2001.

No context found.

L. Alvisi, J.P. Martin and M. Dahlin. Small Byzantine Quorum Systems. In Proceedings of the International Conference on Dependable Systems and Networks (DSN 2002 and FTCS 32), DCC Symposium, Washington, DC, June 2002, pp. 374-383.

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