Background information about the HMI/AIA Joint Science Operations Center – JSOC

This page contains some background information that might help potential proposers responding to the NASA 2008 ROSES NRA for the SDO Science Center. Much of the information was collected in the fall 2007 in anticipation of the NRA and may be a bit dated. However it shows the overall structure of the JSOC-SDP system. The page assembled last fall is here. This page was made on 10 June 2008.

The official copy of the NRA is available via the NASA NSPIRES system at (while open) NNH08ZDA001N-SDOSC with a local copy here.

Proposers or potential proposers are welcome to contact the JSOC personnel at Stanford to learn more. We expect the SDO community to be learning about the system and are happy to describe it whether you are a proposer or not. General questions can go to jsoc_dev on our sun.stanford.edu mail system.

Some details that may be of interest are buried in the above docs. For instance the capability of the computing cluster that we have to support the end product of development supported by this AO. The cluster is a set of 64 nodes with each node consisting of a single board with dual quad-core Intel based x86-64 chips. Thus the 512 processors are essentially 64 8-processor systems. All processing on the cluster is via the Sun Grid Engine (SGE) queue management system. The cluster nodes are connected via Iinfiniband switches to the > 400TB disk array. The throughput to the disks is better than 2GBytes/sec. The tape system used by SUMS, also on the infiniband system, consists of 12 LTO-4 drives in a robotic library with 2200 tapes of 800GB each. So there is about 2PB nearline with access times of several minutes.

Most pipeline programming is done in C. We also have full support for FORTRAN and some key modules (e.g. vector field and time-distance helioseismology) are based on FORTRAN. Other languages, e.g. IDL, are not well supported. We have dropped earlier plans to fully support IDL. However the AIA team is planning on supporting additions to the SolarSoft IDL environment to allow operation with the DRMS/SUMS system.

We plan to allocate one cluster node (8 processor cores) to each of the two topics supported in this NRA without the need for prior detailed discussions. Computing resources beyond that would require detailed discussions with the JSOC team via Phil Scherrer.