This thesis is about interaction between different architectures inhigh performance computing for file system I/O. This is evaluated byperformance, scalability and fault handling. What excel in a looselycoupled system fail in a tightly connected system and vice versa.
The I/O-path from disk to application have been examined boththeoretically and with tests for local and distributed file systems.The impact of different levels of cache is shown using various tests.
This test results has been used to design and implement a protocolgiving SCI the semantics of TCP/IP, thereby replacing TCP/IP inPVFS. SCI is a low latency, high throughput interconnect withdecentralized routing. In PVFS interconnect latency have only provenimportant for meta data operations. For I/O operations the pipelininghides the latency with the protocol window. PVFS have as expectedshown increased read and write performance with increased interconnectthroughput. Throughput have been increased by a factor of 5 byintroducing SCI from 100Mb/s Ethernet. To limit overloading in theinterconnect, two different techniques have been evaluated. Exponential backoff as in TCP/IP and a token based scheme.