NOWNetwork of Workstations Sun ULTRA 5 with 32

NOW(Network of Workstations) Sun ULTRA 5 with 32 GB disk １ 6 Units 6 Sun ULTRA 10 Eathernet Switch 32 GB disk 100 M One unit

The Grid 7

MIMD系� 分� sequent Encor 23 超级计算机 RP 3 LAN 作站 超立方 Transputer

Intel 单核CPU芯片 28

Intel多核结构 l 29 Replicate multiple processor cores on a single die.

2 Processors vs. 2 Cores 2 Processors • Two separate chips connected externally by a system bus • Requires explicit software support • Much greater power consumption 2 Cores l Two cores on same die connected directly l Multiple threads and processes automatically processed in parallel l Only a small increase in power 31 consumption

Pentium D with 975 X Chipset DDR 2 Memory Core 0 L 2 Cache (for Core 0) Core 1 L 2 Cache (for Core 1) Inter-Core Bus Interface 1066 / 800 MHz FSB Memory Controller Hub PCI Express x 16 or 2 PCI Express x 8 DMI (2 GB/s) 6 USB 2. 0 High-Definition Audio I/O Controller Hub 6 PCI 4 Serial ATA Ports 6 PCI Express x 1 32 Intel Pro 1000 LAN BIOS Support Intel Matrix Storage

软件相关概念 System Description Main Goal DOS Tightly-coupled operating system for multiprocessors and homogeneous multicomputers Hide and manage hardware resources NOS Loosely-coupled operating system for heterogeneous multicomputers (LAN and WAN) Offer local services to remote clients Middleware Additional layer atop of NOS implementing generalpurpose services Provide distribution transparency l l 35 l l An overview of DOS (Distributed Operating Systems) NOS (Network Operating Systems) Middleware

多� 算机操作系� Synchronization point Send buffer Reliable comm. guaranteed? Block sender until buffer not full Yes Not necessary Block sender until message sent No Not necessary Block sender until message received No Necessary Block sender until message delivered No Necessary l 44 Relation between blocking, buffering, and reliable communications.

分布式共享内存系� 45 a) Pages of address space distributed among four machines b) Situation after CPU 1 references page 10 c) Situation if page 10 is read only and replication is used

分布式共享内存系� l 46 False sharing of a page between two independent processes.

中� 件和开放性 1. 23 l 48 In an open middleware-based distributed system, the protocols used by each middleware layer should be the same, as well as the interfaces they offer to applications.

可伸� 性（scalability） 避免： • 集中式硬件 • 集中式算法 • 集中式的数据结构 55 Concept Example Centralized services A single server for all users Centralized data A single on-line telephone book Centralized algorithms Doing routing based on complete information