Previous topic |
Next topic |
Contents |
Contact z/OS |
Library |
PDF
Cached data writing z/OS Network File System Guide and Reference SC23-6883-00 |
|||||||||||||
The z/OS NFS Server always caches writes if out-of-sequence data packets are received or if a physical block of data is partially filled. If the NFS Server is processing in the binary data mode, the writes will remain cached until one of the following occurs: v4 CLOSE occurs, the write timeout for a data set has been reached or, if the logicalcache attribute is defined at less than 2GB, the number of cached packets exceeds the number specified in cachewindow . If the NFS Server is processing text data, the writes remain cached until v4 CLOSE occurs, or the write timeout for a data set occurs. If the logicalcache attribute is defined at less than 2GB and the number of cached packets exceeds the number specified in cachewindow, data processing ends with EIO error code. The missing data is padded with binary zeroes and record delimiters so that cached writes for text processing are written in the MVS data set on DASD at the location specified in each cached data packet. In the case of cached data packets for binary processing, only binary zeroes will be used to pad the missing data written at the specified location on DASD. See Table 1 . Attention: It is recommended that the application
and NFS client perform the writes in offset and length in multiple
of 512 bytes in order to optimize the mapping byte offsets to records
process described previously in this section and, therefore, reduce
data flush time. On the other hand, for the NFS version 3 commit
procedure, the z/OS server will only support committing the cached
data when the data set is timed out. For the NFS version 4 commit
operation, the z/OS Server will only support committing the cached
data upon receiving the close operation.
Note:
|
Copyright IBM Corporation 1990, 2014
|