DataPark
Historically, nearly all data analysis has been done on NCAR
supercomputers because these were the only machines that had
high-speed access to the NCAR Mass Storage System and that
had the requisite disk capacity. With the advent of high speed,
shared memory microprocessor systems, it is technically and
economically feasible to provide a facility that is tailored to
analysis. We call this facility the DataPark, and a prototype of it
has been deployed during FY97. Our objectives for the DataPark
include:
- Provide users with an amount of fast, random access storage
space that is at least an order of magnitude larger than typically
available on their local workstation or departmental system.
- Where end-to-end network bandwidth is sufficiently high,
enable users to work in a distributed environment wherein data
are exported to user systems using industry-standard protocols,
e.g. NFS and DFS.
- Where end-to-end bandwidth is insufficient or a user's
local computing system does not have adequate computing power,
provide users with access to a compute server and associated data
analysis environment within the DataPark.
As shown in the "Prototype DataPark" figure, objectives 2) and 3)
are implemented on separate machines. During FY97, SCD established
a prototype DataPark system. It is comprised of two SGI Power
Challenge computers; one is used as a fileserver (middlepark)
and the other a compute server (winterpark). SCD will be adding
another 130 Gbytes of local disk storage to winterpark soon,
which will enable special projects to access dedicated blocks
of high-speed storage for long periods of time. The first two
projects will be the Data Support Section's Reanalysis Project
and the joint Climate and Global Dynamics Division-SCD CSM
Post-Processor Project.
As of FY97 end, this is our conceptual diagram for a DataPark
that meets the objectives stated above:
| Next page |
Top of this section |
Table of contents |
| NCAR |
UCAR |
NSF |
NCAR FY97 ASR |