-
[show abstract]
[hide abstract]
ABSTRACT: Leader election is an important problem in distributed computing systems. In this problem, when the leader is crashed, other nodes must elect another leader. Garcia-Molinapsilas Bully Algorithm is a classic solution to cope with this problem. In this paper we introduce a new method based on electing a leader and alternatives. If the leader crash, the alternative takes, care of the leader's responsibilities. Our results revealed that often, after a leader crash, leader alternative elect as a leader and continue to work. increases. We show that our proposed algorithm is more efficient than the Garcia Molinapsilas one in term of time, message complexity and in the fewer stages.
Communication Theory, Reliability, and Quality of Service, 2009. CTRQ '09. Second International Conference on; 08/2009
-
[show abstract]
[hide abstract]
ABSTRACT: In this paper, we present a new low-latency asynchronous pipeline control circuit. The control circuit has only two gate delays in its critical path, which is faster compared to other works reported in the literature. Two applications of the design are shown to demonstrate its efficiency. The first is a 16-bit FIFOs and the second is a 4times4 multipliers which are designed both using LLA and GasP approaches. The designs are then simulated using HSPICE and 0.18 mum CMOS technology parameters. The results show a latency improvement of about 38% in FIFOs and 18% in multipliers without any considerable degradation in other parameters such as power and throughput. The energy-delay curves also show better overall performance of LLA.
Signals, Circuits and Systems, 2009. ISSCS 2009. International Symposium on; 08/2009
-
[show abstract]
[hide abstract]
ABSTRACT: In this paper, an efficient yet accurate model for extracting analytical expressions for throughput and latency of asynchronous pipeline styles is presented. The model is then applied to ten styles including GasP, MOUSETRAP, IPCMOS, LP<sub>SR</sub>2/1, LP<sub>HC</sub>, STFB, LDA, LP2/1, RSPCFB and NCL. The first five designs are based on the bundled-data (BD) category while the last five are based on the data-driven (DD) category. The performance of the styles are compared in terms of throughput, latency, power dissipation and transistor count. A figure of merit for the energy consumption and the performance is used to determine the better styles. The results of the analytical model are compared with (pre-layout) HSPICE simulation with a 0.18 μm CMOS technology revealing a very good accuracy for the proposed model.
Circuits and Systems, 2005. ISCAS 2005. IEEE International Symposium on; 06/2005
-
[show abstract]
[hide abstract]
ABSTRACT: In distributed systems which use RPC, if a failed process sent a request before failing, the receiver of this request becomes an orphan process and must roll back to undo the effects of receiving the message. There are two types of orphan: one of them mostly called "crash- orphan" in which client crashes. Another which causes orphan process is called abort-orphan in which parent's process is aborted (Baumann and Rothermel). Orphans are undesirable because they waste system resources and can make inconsistent data (Herlihy and Mckendry, 1990). In this paper initially we present two novel methods for orphan detection. Finally we compare our new methods with the older ones.
Advanced Information Networking and Applications, 2005. AINA 2005. 19th International Conference on; 04/2005
-
[show abstract]
[hide abstract]
ABSTRACT: In this paper, ten asynchronous pipeline styles are studied and compared. These include GasP, MOUSETRAP, IPCMOS, LP<sub>SR</sub>2/1, LP<sub>HC</sub>, STFB, LDA, LP2/1, RSPCFB and NCL. The first five designs are based on bundled-data (BD) category while the last five are based on data-driven (DD) category. Analytical expressions for the throughput and the latency of the asynchronous styles are presented. A 4-bit 4-stage FIFO circuit based on each style is designed and simulated utilizing HSPICE with a 0.18 μm CMOS technology. The simulation results are then used to compare the FIFOs in terms of throughput, latency, power dissipation and transistor count.
Microelectronics, 2004. ICM 2004 Proceedings. The 16th International Conference on; 01/2005
-
Proceedings of the IASTED International Conference on Parallel and Distributed Computing and Networks, part of the 23rd Multi-Conference on Applied Informatics, Innsbruck, Austria, February 15-17, 2005; 01/2005
-
[show abstract]
[hide abstract]
ABSTRACT: In this paper, an efficient micropipeline FIFO for low-power applications is introduced. In this FIFO, instead of using a latch in each stage, an inverter gate with a weak keeper is utilized. This leads to a significant reduction in the number of transistors and the power consumed by the circuit. To demonstrate the efficiency of the design, a 16-bit 4-stage FIFO with four-phase micropipeline control is designed using this approach and a traditional style which uses latches in each stage of the FIFO. The number of transistors is reduced more than 43 percent. The designs are then simulated using HSPICE with 0.6μm CMOS parameters. The results of the power consumption show a power reduction of more than 72 percent.
Circuits and Systems, 2002. MWSCAS-2002. The 2002 45th Midwest Symposium on; 09/2002
-
[show abstract]
[hide abstract]
ABSTRACT: In distributed operating systems orphan detection is a key subject. There exist some methods for cope with orphans. One of them is DSG method which overcomes to previous method. In this paper initially we introduce DSG method and then we present its advantages. After it we intended to prevent burst traffic in a special scenario that can be accrued in DSG environment.
-
[show abstract]
[hide abstract]
ABSTRACT: Orphan detection methods demonstrate different performance, memory consumption in different scenarios. Dedicated Server Group (DSG) method is one of the most proper one. In this paper, we overviewed DSG method and analyzed its advantages and disadvantages. Based on the analytical results, we improved the DSG method in both process overhead and communication/traffic overhead. The dependability of improved method is evaluated by Markov chain modeling using SHARPE package and the Availability, Reliability, and Mean-Time to Failure are calculated.
-
[show abstract]
[hide abstract]
ABSTRACT: Leader election is an important problem in distributed computing, and it is applied in many scientific fields such as communication network [1,2,3,4,5], centralized mutual exclusion algorithm [6,7], centralized control IPC, Berkeley algorithm, etc. Synchronization between processes often requires one process acting as a coordinator. The coordinator might not remain the same, because it might get crashed. Bully election algorithm is one of the classic methods which is used to determine the process with highest priority number as the coordinator. In this paper, we will discuss the drawbacks of Garcia_Molina's Bully algorithm and then we will present an optimized method for the Bully algorithm called modified bully algorithm. Our analytical simulation shows that, our algorithm is more efficient rather than the Bully algorithm with fewer message passing and fewer stages.
-
[show abstract]
[hide abstract]
ABSTRACT: Different orphan detection methods prefer different tradeoffs between performance, storage overhead, and simplicity of recovery. There exist some methods for orphan detection. One of them called DSG method that has been presented in [1]. In this paper initially we introduce DSG method and then we present its advantages. Finally we improve this method in either process overhead and communication/traffic overhead.
