M. Livingston and Q. F. Stout. Fault tolerance of allocation schemes in massively parallel computers. Proc. 2nd Symp. Massively Parallel Computation (1988) 491-494.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....by allocating useful system resources such as cpu, memory, and disks. A careful allocation of processes can minimize the overhead of interprocess interactions. In a system with replicated processes to tolerate a fault, process allocation affects the performance and reliability of the system [1, 2, 3, 4, 5]. Not only does the system s performance degrade because of the increment in the number of running processes, but the system s reliability also depends on the placement of replicated processes in the system. Process allocation in fault tolerant multi computer systems has been studied by several ....

M. Livingston and Q. F. Stout, "Fault tolerance of allocation schemes in massively parallel computers, " in Proc. of 2nd Massively Parallel Computers Conference, pp. 491--494, 1989.

No context found.

M. Livingston and Q. F. Stout. Fault tolerance of allocation schemes in massively parallel computers. Proc. 2nd Symp. Massively Parallel Computation (1988) 491-494.

....location that improves upon the buddy strategy, and proposed a multiple gray coded scheme using Gamma d bd=2c Delta gray codes to effect complete subcube recognition. Al Dhelaan and Bose [2] introduced a modified buddy strategy which improves upon the single gray coded buddy scheme. In [16], we examined several subcube location schemes including multiple buddy and gray coded buddy schemes, and introduced a new family of location schemes which generalized and improved upon the earlier schemes. We make the distinction between the problems of subcube location and allocation, viewing ....

....indicates, the cyclic buddy system does no better than the single or double buddy or gray coded buddy system. The values of Kw (d; q) are not known for all d and q, but we include a few of the known values appropriate for comparison. Theorem 3. 1 For d q, i ) Cw (d; q) 2 d Gammaq , ii ) [16], Bw (d; q) Gw (d; q) DBw (d; q) DGw (d; q) 2 d Gammaq , iii ) 6] Kw (d; d Gamma 2) lg d Theta(lg lg d) iv ) 6] Kw (d; q) 2 d Gammaq Gamma1 lg(q 3) Gamma (d Gamma q) Delta lg(d Gamma q) Gamma 2 lg lg d. 2 Exact analytic expressions for expected fault tolerance ....

[Article contains additional citation context not shown here]

M.L. Livingston and Q.F. Stout, "Fault tolerance of allocation schemes in massively parallel computers", Proc. 2nd Symp. Frontiers of Massively Parallel Computation (1988), pp. 491-494.

....less fault tolerant as d increases. Further, if we consider the expected case behavior, where B e (n; m) and K e (n; m) denote the corresponding numbers for the situation in which the faulty (or busy) processors are distributed independently and uniformly throughout the hypercube, it is shown in [LiSt] through simulation that B e (2 20 ; 2 18 ) 8:1 and K e (2 20 ; 2 18 ) 24:6. Thus, in the worst case, we suffer a 50 decrease and, in the expected case, a decrease of 67 in the fault tolerant allocation ability of the system. To increase the number of subsystems which can be ....

....q, all subcubes of dimension d Gamma q in Theta(d) time. Complete allocation in hypercubes may not be the method of choice when d, q, and d Gamma q are all large, or when allocation of all sizes is necessary. For such cases, we recommend the use of the k cube buddy system first introduced in [LiSt]. For fixed k, this system allocates q dimensional subcubes in which the last d Gamma k bits are arbitrary and the first d Gamma q k bits are the nodes of a k subcube of a d Gamma q k dimensional cube. In Section 3.5 we give a Theta(d) algorithm for the allocation of all subcubes allowed ....

[Article contains additional citation context not shown here]

M. Livingston and Q.F. Stout, "Fault tolerance of allocation schemes in massively parallel computers ", Proc. 2nd Symp. on the Frontiers of Massively Parallel Computation (1988), (to appear) .

....: ng, and it is easy to check that (B; n; m) B; n; m) 2 n Gammam for n m 1. While the buddy system is the only allocation scheme used on hypercube computers thus far, we see it is not particularly fault tolerant. For some specific allocation schemes of interest, Livingston and Stout [29] determined (A; n; m) For arbitrary allocation scheme A, Becker and Simon [3] showed that the problem of determining (A; n; n Gamma 2) is equivalent to a graph coloring problem. The general problem of determining (A; n; m) and (A; n; m) is open. The fault tolerance questions considered here can ....

....an (n Gamma 1) cube. In contrast to (n; n Gamma 1) 2 we find that its expected value, denoted E (n; n Gamma 1) is Theta(log n) Some of the properties of E (n; m) and E (n; m) for arbitrary n and m, and of E (A; n; m) and E (A; n; m) for certain allocation schemes A, are studied in [29, 30]. A related but somewhat different situation arises if we are only concerned that, with high probability, G fails to have property P . What is the expected number of copies of H that must be removed in this case Becker and Simon [3] considered an instance of this question in which G is Q n , H ....

M. Livingston and Q. F. Stout. Fault tolerance of allocation schemes in massively parallel computers. Proc. 2nd Symp. Massively Parallel Computation (1988) 491-494.

No context found.

M. Livingston and Q.F. Stout, "Fault tolerance of allocation schemes in massivelypyN4### comp###N2S Proc. 2nd Massively ParallelComplN44 Conference, pnference, 1989.

No context found.

M. Livingston and Q. F. Stout, "Fault tolerance of allocation schemes in massively parallel computers," in Proc. of 2nd Massively Parallel Computers Conference, pp. 491-- 494, 1989.

No context found.

M. Livingston and Q. F. Stout, "Fault tolerance of allocation schemes in massively parallel computers," in Proc. of 2nd Massively Parallel Computers Conference, pp. 491-- 494, 1989.

No context found.

M. Livingston and Q. F. Stout, "Fault tolerance of allocation schemes in massively parallel computers, " in Proc. of 2nd Massively Parallel Computers Conference, pp. 491--494, 1989.

No context found.

M. Livingston and Q. F. Stout, "Fault tolerance of allocation schemes in massively parallel computers," in Proc. of 2nd Massively Parallel Computers Conference, pp. 491-- 494, 1989.

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