6. Conclusion

method for obtaining digital signatures and public-key cryptosystems. then set sender = (j + 1) mod n. 3. When (i ? sender) mod n becomes less than S T , start the computation of p)B 2 for the set B 2 to be included in the next C-mcast, i.e., where B 2 includes the digests of all messages M 0 such that (i) there is no direct message M 00 such that D(M 00)!D(M 0), (ii) if D(M 0)! D(M 00), then M 00 is direct or delivered, and (iii) M 0 did not appear in B 2 in a previous execution of C-mcast. 4. When sender = i, dequeue the first message m in pending, and execute C-mcast(m; B 1 ; p)B 2), where p)B 2 is already prepared, and B 1 contains (unsigned) message digests of messages that currently satisfy requisites (i)–(iii) above. 5. If C-deliver(q; m) is executed, then execute R-deliver(q; m). Under normal (faultless) conditions, this reliable mul-ticast protocol can potentially achieve a delivery latency of T + S + d(2n + 1)=3e (U + T). This latency is derived as follows: The transmission of a message M takes T time. S is the time it takes for the process S T hops away from the sender to complete a signed acknowledgement for M, or equivalently for this process to become enabled to transmit. This is followed by d(2n + 1)=3e transmissions , each one being initiated as soon as the previous one is received and thus taking T time, and each one containing a signed acknowledgment for M (direct or indirect). Finally, the d(2n + 1)=3e signatures must be verified. This calculation assumes that each process is ready to transmit a message as soon as its turn arrives. In order for this to hold, it is assumed that n S T. Using the parameters above (with 300-bit RSA moduli and public exponents equal to 3), the delivery latency for 9 Sparc 20s is potentially between 30ms and 40ms. In this abstract we presented a high-throughput mul-ticast protocol that ensures that all members of the mul-ticast destination group receive the same multicast messages , despite the malicious collaboration of fewer than one-third of the group members. High throughput is achieved due to an acknowledgement chaining technique , whereby a single signature is used to indirectly acknowledge multiple messages. Our protocol also includes a flow control mechanism …