Hitachi 9900 Series User Manual

Page of 80
At the heart of the Lightning 9900
Series’ enormous capacity for throughput lies
the Hi-Star
architecture. This network of interconnecting high-speed links to system
components provides an unequaled internal aggregate bandwidth of 6.4GB/sec
(gigabytes/sec) in the 9960. This bandwidth is divided equally into 3.2GB/sec for data
transfer and 3.2GB/sec for separate internal control information transfer. 
The main components that make up the Lightning 9900
Series are the Array Control
Processor modules (ACPs), Host Adapter modules, called Client Host Interface Processors
(CHIPs). These elements enable the unsurpassed bandwidth of the Hi-Star
network by
bringing their own network paths (and bandwidth) to the Cache Hierarchical Star Network
(C-HSN) and the Control Memory Hierarchical Star Network (CM-HSN). This means
that the full 6.4GB/sec internal bandwidth is available with a fully configured 9960 with all
four ACP pairs (eight modules total) and all four CHIP pairs (eight modules total). This is
a truly high-performance and cost-effective configuration. With any Lightning 9900
Series
configuration, growing or scaling can be easily completed by adding more CHIP pairs for
connectivity or ACP pairs for more capacity or performance. The internal bandwidth
increases every time components are added. The bandwidths stated are the rated
bandwidths for the paths; the effective rates will vary based on application workload and
bandwidth consumption.
Figure 2 – The Hitachi
Lightning 9960
Series’
Hi-Star
internal
switched-fabric
architecture.
3
Introduction