The optical disk system is a hierarchical system with primary storage on magnetic disk and secondary storage on the optical disks. New files are written to the write cache. Periodically the optical disk archiver copies the files from the write cache to the optical disk. This process also copies the files to the read cache and deletes them from the write cache. The PPPL utilities and libraries that access this data first check the magnetic cache. If the file is not found, the optical disk indices are searched; if the file is found, it is copied to the read cache. The data on the read cache is skimmed by access date.
The optical disks are stored in a jukebox. When a file is requested the platter which contains that file is mounted. This usually takes less than a minute, but may take longer if demand is heavy. Two platters can be mounted simultaneously and they remain mounted until a new platter is needed. Once the platter is mounted, it typically takes about 5 seconds to restore the file.
When you access a file that is on the optical disk, your process sends a request to a separate JUKE_SERVER process. Interactive jobs (but not batch jobs) normally display a message such as: display a message such as:
Notified JUKE_SERVER - you are first in queue. or Request queued - JUKE_SERVER is busy - Waiting for copy.If you get a message indicating some sort of error, contact the Help Desk, x2275. indicating some sort of error, call Computer Operations at 3444.