Abstract: A method, computer cluster and computer unit for performing storage access shift between computer units. A redundant computer cluster comprises computer units and a shared storage unit. The computer units interface the shared storage unit, which stores data storage resources to which access has been established by a replicated service unit executed the computer cluster. When switchover condition associated with the replicated service unit is detected, the data storage resources to which access has been established from the service unit are first released in the service unit replica that will enter standby state. Thereafter, the access is established to the data storage resources in the service unit replica that will enter active state. Only then the replica of the service unit is allowed to become active. The benefits of the invention are related to the improved performance and reliability of file system access from a redundant multi-node computer cluster.

Abstract: A problem with tape units (TU) arises when writing of data onto a tape is not successful. The computer does not know which blocks (1-25) have been stored successfully. Data blocks are buffered in a ring buffer for so long that one can be sure that the storing on the tape (TA) is successful. The ring buffer can be implemented as buffer areas (BA1, BA2) in disk units (W1, W2). Data blocks are stored not only in the tape unit but also in the first buffer (BA1), until it is filled. Then blocks arm stored in the second buffer (BA2), until this too is filled, and the first buffer (BA1) is used again. When there is a problem with the tape unit, the tape unit is re-initialized, whereby the data in the tape memory (TM) is lost. Loss of data is prevented by reading the data from the ring buffer.