Shared-memory multiprocessors cover a wide range of prices and features, from commodity SMPs to large high-performance cc-NUMA machines.
In these designs, accessing main memory has nonuniform access costs to a processor and in this way, architectures of this type are often called cache-coherent, nonuniform memory access or cc-NUMA architectures.
There are several factors limiting the scalability of cc-NUMA designs.
Figure 2 presents the normalized average miss latency obtained when running several applications on top of a simulated 64-node cc-NUMA multiprocessor using RSIM .
In this paper, we present a revision of the proposals that have recently appeared facing these two important issues in cc-NUMA multiprocessors.
5 is a key milestone in our overall UNIX strategy, punctuating a history of accomplishment that has culminated in unrivaled CC-NUMA
scalability within a complete 32- or 64-bit environment," said Richard E.
With the size of the system software for modern computers in the millions of lines of code, the changes for CC-NUMA machines represent a significant development cost.
Computer hardware vendors that use "commodity" operating systems such as Microsoft's Windows NT [Custer 1993] face an even greater problem in obtaining operating system support for their CC-NUMA multiprocessors.
A large CC-NUMA multiprocessor can be configured with multiple virtual machines each running a commodity operating system such as Microsoft's Windows NT or some variant of UNIX.
The use of commodity software leverages the significant engineering effort invested in these operating systems and allows CC-NUMA machines to support their large application base.
This will be the fourth SPP product based on a hardware-based CC-NUMA architecture called Global Shared Memory, which is far ahead of any competitive platform," Bemis said.
With the SPP architecture, we are far ahead of any competitive platform, and scalable parallel processing based on CC-NUMA is rapidly becoming the industry standard," Wallach said.