G. LeLann. Distributed systems, towards a formal approach. In Information Processing 77, pages 155--160. North-Holland, New York, 1977.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....we formally define the problem and correctness criteria, elaborate the program data structures, and then state the program and argue that it satisfies the correctness criteria. 4. 1 The Correctness Criteria Region labeling is a two dimensional variant of the classical leader election problem [6, 13]. A region labeling program receives as input a digitized image. Each point in the image is called a pixel. The pixels are arranged in a rectangular grid of size N pixels in the x direction and M pixels in the y direction. An xy coordinate on the grid uniquely identifies each pixel. Also provided ....

G. LeLann. Distributed systems, towards a formal approach. In Information Processing 77, pages 155--160. North-Holland, New York, 1977.

....[14] the section presents a simple program to solve this problem. The GC program is then transformed, step by step, into a Swarm program. Key aspects of the shared dataspace model are introduced along the way. Region labeling is a two dimensional version of the classical leader election problem [11, 22]. Let us consider an image consisting of N rows and M columns. Each discrete point in the image is called a pixel. Each pixel is characterized by its coordinates in the image (i.e. row and column numbers) and intensity (i.e. brightness or color) Two pixels are called neighbors if they have ....

G. LeLann. Distributed systems, towards a formal approach. In Information Processing 77, pages 155--160. North-Holland, New York, 1977.

....time bounds [0, 3n 11)a] for the crit class and [0, 2a] for each remi class. Then D x U MD x U. 5. 5 Example: LeLann Chang Roberts Leader Election Algorithm Luchangco [18] gives a formal proof of an upper bound on time for the LeLannChang Roberts leader election algorithm for ring networks [4, 16]. The algorithm is simple: every processor in the network sends its processor identifier clockwise, and smaller identifiers that encounter larger identifiers are discarded. If a node receives its own identifier in a message, it elects itself as leader. An upper bound of c is assumed on the step ....

G. LeLann. Distributed systems, towards a formal approach. In IFIP Congress, pages 155-160, Toronto, 1977.

....behaviors, those in which time is unbounded, which correspond to the real behaviors of the system. We use these simulations, together with invariant assertions, to prove correctness and timing properties of two systems, a simple message passing protocol due to LeLann, Chang, and Roberts [LeL77, CR79] for leader election on a ring of processes, and Fischer s timingbased mutual exclusion algorithm using a single shared read write register [Lam87, Fis85] For both algorithms, we use intermediate specifications to obtain hierarchical proofs, and we extract general heuristics for finding ....

....= 1. Otherwise, u:last(report) s:last(report) if s:countdown = 0 s:last(decrement) s:countdown Delta c2 if s:countdown 0 ) s 0 :now 39 40 Chapter 4 LeLann Chang Roberts Election Algorithm We now consider a simple asynchronous algorithm by LeLann, Chang and Roberts [LeL77, CR79] which solves the election problem for a ring network in the message passing model. Although the algorithm is asynchronous, we assume bounds on the communication delay and the local step times in order to prove an upper bound on the time to election. This does not rule out any possible ....

[Article contains additional citation context not shown here]

G'erard LeLann. Distributed systems, towards a formal approach. In Information Processing 77: Proceedings of the Seventh IFIP World Congress, pages 155--160, Toronto, Ontario, 1977.

....a garbage chain passes through site R with an even lower id. In this case, object b will be migrated to S, while object d will be migrated to R. site S site T site R d f e d e g f g Figure 6.8: A chain reachable from a cycle may migrate to a different site. Unlike some leader election algorithms [LeL77] estimating the destination does not incorporate termination detection, so sites must guess when destination propagation has completed; we discuss how to make this guess in the next section. The advantage of our scheme is that it is simple and effective even in compound cycles. 6.2.2 When to ....

G. LeLann. Distributed systems, towards a formal approach. In IFIP Congress, pages 155--160, 1977.

....(n 1)a and (n 3)a for the milestones seize, stabilize , dropback and critical , respectively. Using I, we can provide two stylized simulations to show the needed time bound. LeLann Chang Roberts leader election algorithm. The LeLann Chang Roberts leader election algorithm for ring networks [10, 3] works as follows: LCR: Every process sends its process identifier clockwise. Smaller identifiers that encounter larger identifiers are discarded. If a node receives its own identifier in a message, it elects itself as leader. If we assume an upper bound of on the step time for each process, ....

G. LeLann. Distributed systems, towards a formal approach. In IFIP Congress, pages 155--160, Toronto, 1977.

....N 2 on the cycle proper. The cycle and the front of the chain (object u) will migrate to N 2 , while the trailing part of the chain (object w) would migrate to N 3 . The objects will not migrate further, however, so multiple migrations are still avoided. Unlike some leader election algorithms [LeL77], ours does not incorporate termination detection, so nodes must guess when destination propagation has completed; we discuss how to make this guess in the next section. The advantage of our scheme is that it is simple and effective even in compound cycles. 4.3 When to Migrate: the Second ....

G. LeLann. Distributed Systems, towards a formal approach. IFIP Congress, pages 155--160, Torornto, 1977.

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