15.01.2013 Views

U. Glaeser

U. Glaeser

U. Glaeser

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

advances that have been made in device fabrication and more importantly in the architecting of these<br />

individual components into powerful systems.<br />

In the 1980s, advances in computer architecture research were most pronounced on two fronts: on<br />

the one hand, new architectural techniques such as RISC made their appearance and revolutionized single<br />

processor design and allowed high performance for the single chip microprocessors which first came out<br />

as system components in the 1970s. At the same time, large-scale parallel processors became mature and<br />

could be used by researchers in many high-end, computationally intensive, scientific applications.<br />

In recent times, the appetite of Internet surfers has been fueling the design of architectures for powerful<br />

servers: in Section 5.1 “Server Computer Architecture,” Siamack Haghighi emphasizes the unique requirements<br />

of server design and identifies the characteristics of their applications.<br />

In Section 5.2, Binnl Matthews describes the VLIW (Very Long Instruction Word) processor model,<br />

compares it to more traditional approaches of Instruction Level Parallelism extraction, and demonstrates<br />

the future of VLIW processors, particularly in the context of multimedia applications.<br />

Similarly, multimedia applications have promoted a dual architectural approach. In Section 5.3 “Vector<br />

Processing,” Krste Asanovic traces the ancestry of vector processors to the supercomputers of the 1980s<br />

(Crays, Fujitsu, etc.) and describes the modern applications of this architecture model.<br />

Architectures cannot be evaluated independently of the underlying technology. Indeed, nowadays,<br />

while deep-submicron design rules for VLSI circuit are allowing increasing numbers of devices on the<br />

same chip, techniques of multiprocessing are seeing additional applications in different forms which<br />

range from Networks Of Workstations. Portability, all the way to multiprocessing on a chip, is the topic<br />

of Section 5.4 “Multiprocessing, Multiprocessing,” by Manoj Franklin.<br />

Taking concurrent processing to the next level, Donna Quammen surveys parallel systems in Section<br />

5.5 “Survey of Parallel Systems,” including large-scale tightly coupled parallel processors.<br />

Finally, in Section 5.6 “Virtual Memory Systems,” Bruce Jacob surveys the concepts underlying virtual<br />

memory systems and describes the tremendous advances this approach has undergone since first being<br />

proposed in the late 1960s.<br />

5.1 Server Computer Architecture<br />

Siamack Haghighi<br />

Introduction<br />

Widespread availability of inexpensive Internet access and powerful computers has resulted in considerable<br />

business productivity improvement. Electronic automation of business operations and Internet<br />

communication has resulted in many profitable electronic or “e” business models such as e-commerce.<br />

The cost-saving potential has required many modern companies to automate their traditional manual<br />

customer interface operations and processing through Web technologies. Businesses rely on enterprise<br />

information technology (IT) computing and communication infrastructure as a backbone for their<br />

operations. It is estimated that the current e-commerce revenues exceed hundreds of billions of dollars<br />

in the U.S. alone.<br />

Availability of robust, reliable, secure, and cost-effective IT infrastructure is one of the key drivers of the<br />

new Internet-based businesses. Customer usage models and applications also affect IT infrastructure performance,<br />

operation, and cost. IT infrastructure hardware and software requirements can be met cost<br />

effectively with client–server computing paradigm. Although not a new idea, availability of inexpensive high<br />

performance microprocessors, scalable computing and storage systems, and high bandwidth networks make<br />

client–server computing model an ideal fit. Shared data storage and backup, infrastructure cost, reliability,<br />

serviceability, and availability are some of the other reasons for adoption of client–server computing.<br />

Powerful central processing units (CPU), reliable and available memory system, scalable input/output (I/O),<br />

advanced software, and services and packaging are some of the elements of servers. This chapter will<br />

highlight some of the architecture, design, deployment, usage model, and application of servers.<br />

© 2002 by CRC Press LLC

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!