5 Conclusion

Abstract

were made for get_file operations on a disk to memory basis, i.e. the files were requested from the file servers disk and discarded on the receivers side. Thus we where able to neglect the influence of local disk write operations upon the results. This resembles a scenario where a user requests a certain file (e.g. a drawing) for viewing. Since drawings are typically stored as bitmaps (binary) or as graphic language files (text) we examined text and binary transfer modes. The results are satifying, compared with the maximum achievable throughput at TCP level of 1,185,840 bytes/sec. Especially if we think, that the throughput also includes the server disk reading delays. The vast difference of 200 Kbyte/ sec between binary and text modes is due to the extensive processing needed for ASN.1 character encoding/decoding. We described the design and implementation of a distributed file service which is part of a service platform that supports collaborative product development in engineering environments. Furthermore the file service is careful about openess assumption with respect to application systems (e.g. Computer Aided Design Systems). By avoiding architectural extensions to operating systems, it may serve for the a-posteriori integration of existing systems. The file service is based upon the ISO FTAM file transfer protocol and thus usable in heterogeneous computer environments. Name and location transparency are achieved by the integration of a so called CIM manager in the file transfer process. Furthermore the CIM manager supports joint work upon documents (=files) by allowing various degrees of sharing semantics and keeping a design history log. Based upon the FTAM architectural model, every application system may be file server and file service client at the same time. Through this design, the autonomy of the application systems is preserved while the shared files stored in the distributed file system are controlled by the CIM manager. We did not discuss security aspects (which would have been an important issue) since we use the simple, FTAM defined security mechanisms of user identification via passwords. We further described the ISODE based implementation of the file service and discussed its performance on the basis of measurements take at our prototype implementation. It was shown that the usage of FTAM did not result in performance drawbacks, compared with other file transfer protocols, like FTP.