Memory-Based Communication Facilities and Asymmetric Distributed Shared Memory
نویسندگان
چکیده
In general-purpose parallel and distributed systems, performance of the protected and virtualized user-level communications and synchronizations is the most crucial issue to realize efficient execution environments. We proposed a novel high-speed user-level communication and synchronization scheme “Memory-Based Communication Facilities (MBCF)” for a general-purpose system with an off-the-shelf communication-hardware. The MBCF realizes the direct remote-memory-accesses in user-task-space and offers programmers and compilers a large shared-memory space. This paper describes outlines and characteristics of the MBCF, and evaluates basic performance of the MBCF/Ether which is the first sample implementation with 100BASE-TX interfaces. The evaluation tells that its peak bandwidth at half-duplex mode is 11.2Mbyte/sec and its round-trip latency is 49 sec. Finally, we introduce a brand-new remote cache scheme “Asymmetric Distributed Shared Memory (ADSM)”, which is suitable for the MBCF and optimizing compilers, and show the effects of optimization methods for the ADSM.
منابع مشابه
Compiler-Assisted Distributed Shared Memory Schemes Using Memory-Based Communication Facilities
To execute shared-memory-based parallel programs efficiently, we introduce two compiler-assisted software cache schemes which are well-suited to automatic optimizations of remote communications. One scheme is a full user-level software cache (User-level Distributed Shared Memory: UDSM) and another is a page-based cache (Asymmetric Distributed Shared Memory: ADSM) which exploits TLB/MMU only in ...
متن کاملDIPC: A System Software Solution for Distributed Programming
Distributed Inter-Process Communication (DIPC) provides the programmers of the Linux operating system with distributed programming facilities, including Distributed Shared Memory (DSM). It works by making UNIX System V IPC mechanisms (shared memory, message queues and semaphores) network transparent, thus integrating neatly with the rest of the system. The underlying network protocol used is TC...
متن کاملParallel Computing in a World of Workstations
Todays computer systems consist of powerful workstations with fast communication links. However, until now network-based parallel computers which employ interconnected workstations as processing elements are not widely used. Poor programmability of those systems as a whole seems to be one of the main reasons. New language constructs, programming paradigms and techniques for managing memory in s...
متن کاملStructuring Concurrent Applications Using Active Expressions
The paper presents the concept of Active Expressions and describes a model, called A e , for structuring concurrent applications. Relying solely on the features of an object oriented language, A e allows deenition of common communication and synchronization patterns. These patterns are then combined with the user provided application code to instantiate complete concurrent applications. The app...
متن کاملRealizing Consistent Event Ordering in Distributed Shared Memory Systems
A large number of tasks in distributed systems can be traced down to the fundamental problem of attaining a consistent global view on a distributed computation. Based on our previous theoretical work concerning consistent event ordering in systems featuring both message passing and distributed shared memory facilities, in the paper at hand we discuss the more practical issues, solutions, and re...
متن کامل