Implementation of a Software Layer for the Control of Data Transfers to/from Magnetic Tape at GridKa

Description

Magnetic tape is used as a cost effective storage medium for data which is infrequently accessed. GridKa users currently store 54PB of data on tape. Since tape is a linear medium, access to data on tape has to be controlled in order to optimize the optimum throughput of the system. If data is to be recalled from tape, requests have to be gathered and ordered. A new system will be set up where the GridKa storage middleware dCache [0] transmits the requests via a java plugin [1] to the GridKa tape backend. Communication between the java plugin and the tape backend is implemented via a parallel file system.

Task

• Implementation of a new software layer which manages the tape requests stored in the file system and hands them over to the tape backend.
• Re-implementation of the dCache plugin and the system which communicates with the backend to use a message queue system instead of the parallel file system for communication.

Requirements

• good knowledge of python or another modern scripting language
• knowledge of java (dCache is written in java)

Supervisors

Haykuhi Musheghyan haykuhi.musheghyan@kit.edu
Andreas Petzold petzold@kit.edu

References

[0] dCache

[1] dCache TSM Plugin