remains a serious problem for software distributed shared memory systems. These results suggest that future work on software DSMs should concentrate on reducing the amount ofsynchronization or its effect.

In our former studies, it was found that even if there was relatively long overlapping between two successive sounds with long decay, the overlapping wasn't detected. We related this phenomenon to self-masking, that is, masking of the former portion of a sound to the latter portion of the same sound. At the same time, this phenomenon might be caused by cognitive factor like whether two sounds are grasped as single sound event. In this study, the localization of each sound was controlled by varying interaural level difference of each sound and it was investigated whether this phenomenon was caused by perceptual factor like self-masking or cognitive factor like cognition of sound event. A piano sound has a short attack and a fairly long decay (referred to as decaying sounds in this paper). Our former study revealed that a short melody played by decaying sounds needed fairly long overlapping in order to be perceived as being marginally connected (Kuwano, Namba, Yamasaki, and Nishiyama, 1994). This result suggested that such a long overlapping was not perceived in the melody. Another study showed that such a long overlapping made the sounds smooth and beautiful. While the long overlapping like this was not perceived as described above, the sounds were perceived as impure in case that the former portion of the first sound was cut off and two sounds after the onset of the second sound were presented (see figure 1-a). If the appropriate part of former portion of the first sound was added to the sounds as shown in figure 1-b, the impression of the sounds changed to pure one (see figure 1-b)

We investigate the preemptive scheduling of periodic, real-time task systems on one processor. First, we show that when all parameters to the system are integers, we may assume without loss of generality that all preemptions occur at integer time values. We then assume, for the remainder of the paper, that all parameters are indeed integers. We then give as our main lemma both necessary and sufficient conditions for a task system to be feasible on one processor. Although these conditions cannot, in general, be tested efficiently (unless P = NP), they do allow us to give efficient algorithms for deciding feasibility on one processor for certain types of periodic task systems. For example, we give a pseudo-polynomial time algorithm for synchronous systems whose densities are bounded by a fixed constant less than 1. This algorithm represents an exponential improvement over the previous best algorithm. We also give a polynomial-time algorithm for systems having a fixed number of distinct typ...

remains a serious problem for software distributed shared memory systems. These results suggest that future work on software DSMs should concentrate on reducing the amount ofsynchronization or its e ect. 1

In a security system that allows people to choose their own passwords, those people tend to choose passwords that can be easily guessed. This weakness exists in practically all widely used systems. Instead of forcing users to choose well-chosen secrets, which are likely to be di cult to remember, we propose solutions that maintain both user convenience and a high level of security at the same time. The basic idea is to ensure that data available to the attacker is sufficiently unpredictable to prevent an off-line verification of whether a guess is successful or not. We examine common forms of guessing attacks, develop examples of cryptographic protocols that are immune to such attacks, and suggest a systematic way to examine protocols to detect vulnerabilities to such attacks.

Phenomenal improvements in the computational performance of multiprocessors have not been matched by comparable gains in I/O system performance. This imbalance has resulted in I/O becoming a significant bottleneck for many scientific applications. One key to overcoming this bottleneck is improving the performance of parallel file systems. The design of a high-performance parallel file system requires a comprehensive understanding of the expected workload. Unfortunately, until recently, no general workload studies of parallel file systems have been conducted. The goal of the CHARISMA project was to remedy this problem by characterizing the behavior of several production workloads, on different machines, at the level of individual reads and writes. The first set of results from the CHARISMA project describe the workloads observed on an Intel iPSC/860 and a Thinking Machines CM-5. This paper is intended to compare and contrast these two workloads for an understanding of their essential similarities and differences, isolating common trends and platform-dependent variances. Using this comparison, we are able to gain more insight into the general principles that should guide parallel file-system design.

Bulk data transmission is now finding more and more application in various fields. The major performance concern of a bulk data transfer protocol is high throughput. Theoretically, a packet switched network allows any single user an unlimited share of the netWOrk

Speculator provides Linux kernel support for speculative execution. It allows multiple processes to share speculative state by tracking causal dependencies propagated through interprocess communication. It guarantees correct execution by preventing speculative processes from externalizing output, e.g., sending a network message or writing to the screen, until the speculations on which that output depends have proven to be correct. Speculator improves the performance of distributed file systems by masking I/O latency and increasing I/O throughput. Rather than block during a remote operation, a file system predicts the operation’s result, then uses Speculator to checkpoint the state of the calling process and speculatively continue its execution based on the predicted result. If the prediction is correct, the checkpoint is discarded; if it is incorrect, the calling process is restored to the checkpoint, and the operation is retried. We have modified the client, server, and network protocol of two distributed file systems to use Speculator. For PostMark and Andrew-style benchmarks, speculative execution results in a factor of 2 performance improvement for NFS over local-area networks and an order of magnitude improvement over wide-area networks. For the same benchmarks, Speculator enables the Blue File System to provide the consistency of single-copy file semantics and the safety of synchronous I/O, yet still outperform current distributed file systems with weaker consistency and safety.

We describe a new fault-tolerant algorithm for solving a variant of Lamport's clock synchronization problem. The algorithm is designed for a system of distributed processes that communicate by sending messages. Each process has its own read-only physical clock whose drift rate from real time is very small. By adding a value to its physical clock time, the process obtains its local time. The algorithm solves the problem of .maintaining closely synchronized local times, assuming that processes' local times are clo. sely synchronized initially. The algorithm is able to tolerate the failure of just under a third of the participating processes. It maintains synchronization to within a small constant, whose magnitude depends upon the rate of clock drift, the message delivery time, and the initial closeness of synchronization. We also give a characterization of how far the clocks drift from real time. Reintegration of a repaired process can be accomplished using a slight modification of the 'basic algorithm. A similar style algorithm can also be used to achieve synchronization initially.

ABSTRACT In interphase Chinese hamster ovary (CHO) cells, the centrosome is attached to the nucleus very firmly. This nuclear-centrosome complex is isolated as a coherent structure by lysis and extraction of cells with Triton X-100 in a low ionic strength medium. Under these conditions, the ultrastructure of the centrioles attached to the nucleus can be discerned by electron microscopy of whole-mount preparations. The structural changes of the centrioles as a function of the cell cycle were monitored by this technique. Specifically, centriolar profiles were placed into six categories according to their orientation and the length ratio of daughter and parent centrioles. The proportion of centrioles in each category was plotted as a frequency histogram. The morphological changes in the centriole cycle were characterized by three distinguishable events: nucleation, elongation, and disorientation. The progress of centrioles through these stages was determined in synchronous populations of cells starting from S or M phase, in cells inhibited in DNA synthesis by addition of thymidine, and in cytoplasts. The results provide a quantitative description of the events of the centriole cycle. They also show that, in complete cells, nucleation, elongation, and disorientation are not dependent upon DNA synthesis. However, in cytoplasts, although elongation and disorientation occur as in