Site Architecture & Equipment

Figure 1. Typical DICE Site Architecture. Download full-page diagram.

The typical architecture at each of the DICE sites includes the following:

• Compute (HPC) – A modestly configured and stable 64-bit capable cluster. The compute clusters vary at each site to evaluate diverse technologies.  
• Scratch File System – A high performance file system operating on a storage device to provide elevated levels of I/O throughput to support cluster applications and testing.
• Cluster Interconnect –Configured with a high-bandwidth, low-latency interconnects.
• Cluster System Memory – Configured with a minimum of 1 gigabyte (GB) of memory per processor.
• HSM Software – Utilizing the SUN SAM-QFS application as a baseline HSM Software application.  
• Storage Technologies – Utilizing various disk, tape and library technologies to represent different tiers of storage in a data center environment.  
• Network – A diverse network vendor topology with support of 1 Gigabit (Gb) and 10 Gb speeds for internal connections and a 1 Gb Ethernet external connectivity for user access and connectivity between sites.  The network provides support for the IPv4 and the new IPv6 network protocol.
• Security – Kerberos and key distribution center (KDC) Server for providing two-factor authentication to the environment.
• Infrastructure Services – Additional hardware and software to support domain name services (DNS), mail, network time protocol (NTP) and Gateway services.
• Project Servers – Additional servers and hardware are provided to supplement projects in the environment for testing non-cluster related technologies.