RCI services are designed to strengthen both the technical and budgetary elements of proposals requiring research cyberinfrastructure. Below are examples of brief summary language describing current RCI production services. The RCI team have approved these examples for direct use in your proposals.
Please click to open the drawers below, then copy and paste the text for quick and accurate descriptions of RCI services using accepted proposal language.
UC San Diego's data communications network is a high-speed network with a redundant 10GE backbone serving building switches at 10Gbps and 1Gbps. The UCSD network connects more than 200 buildings and includes more than 900 edge switches and 65,000 ports. All desktop ports are 1Gbps capable, with 10Gbps connections (dedicated if necessary) available to research applications on request. UCSD's network is IPv6-ready, with IPv6 available on request for individual VLANs and routed to the Internet.
The campus also provides ubiquitous 802.11n wireless service with over 3500 wireless access points, secured with WPA2-Enterprise using 802.1x for access control.
UCSD is redundantly connected to the California Research and Education network (CalREN) via multiple 10Gbps connections to the Corporation for Education Networking in California (CENIC), providing the campus with 30Gbps of connectivity to NLR, Internet2 and other research networks as well as 10Gbps connectivity to the commodity Internet. In addition, 10Gbps connectivity to ESnet is available at the San Diego Supercomputer Center (SDSC).
Colocation services provided by the San Diego Supercomputer Center are designed to be a cost-effective hosting solution for researcher purchased computer and storage equipment leveraging historical UC investments and economies-of-scale. The environmentally controlled data centers span 19,000 square feet and have a total power capacity of 13 Megawatts. Interior and exterior security systems include two-factor biometric access to both the host building and data centers as well as a 120-camera digital security system. Operations staff is available 24/7/365 to provide remote hands assistance, monitor critical and customer systems, facility oversight, and quick response to any data center event. Emergency power systems, Uninterruptible Power Supply (UPS) and diesel generation are both available on site for research with uptime requirements or to maintain production environments. Seismic isolation systems rated for earthquakes of 7.0 or higher magnitude are installed on every rack to further protect equipment and data. The colocation facility is strategically positioned on the UCSD network to maximize diverse and robust connections to many research networks in addition to the commodity Internet.
The TSCC is a medium-scale, high performance, parallel cluster using the latest processor and interconnect (networking) technologies. Currently offered configurations comprise nodes with the latest generation Intel server class processors (Xeon) with 16 computing cores and either 64 gigabytes or 128 gigabytes (optional) of main memory. Two interconnect technologies are available: Infiniband for low latency parallel computing and 10 Gigabit Ethernet (GbE) as a more cost-effective alternative for computing workloads less sensitive to latency. Computing nodes with Graphics Processing Unit (GPU) accelerators are also available. A wide suite of research software is offered; researchers may also install and run software tools of their choosing. All researchers using TSCC have access to a high capacity parallel file system and external, high bandwidth research networks such as XSEDE and Internet2. Vendor contracts are negotiated to provide for annual technology insertion as newer processors and other components become available at competitive prices.
The Triton Shared Computing Cluster (TSCC) is a high performance research computing system operated under the auspices of UCSD’s Research Cyberinfrastructure (RCI) program, which provides for high performance computing, scientific data storage and management, energy-efficient equipment co-location, and high bandwidth networking to support UCSD faculty and researchers.
TSCC is operated under a hybrid business model, which includes researcher-contributed (condo) computing nodes and pre-purchased computing time on a shared (hotel) portion of the system.
In the condo portion of the system, researchers purchase computing nodes using funds from grants or other sources and contribute the nodes to the cluster. In exchange for an annual operating fee, the researcher-owned nodes are located in an energy-efficient data center at the San Diego Supercomputer Center (SDSC) and maintained by professional system administrators. Researchers may compute exclusively on a number of nodes equal to that purchased, or may use the entire cluster as a shared resource.
On the hotel portion of the cluster, scientists not desiring or able to participate in the condo, or requiring a small amount or short duration of computing, may purchase time on a shared partition of TSCC at a measured rate (per processor core per hour).
In order to facilitate use of the TSCC, UCSD subsidizes the program on the order of $300,000 per year, resulting in an effective computing rate for condo participants of less than 2 cents per core-hour and for hotel participants of approximately 2.5 cents per core-hour, highly competitive rates for state-of-the-art research computing systems.