When running Oracle 10g databases on Red Hat 5.5 with datafiles mounted over NFS v3 we saw suboptimal read performance, specifically db file scattered read requests showing high IO latency. The IO latency coming out of Disk was under 1ms. At the NFS server, we observed similar latency. NFS client was also showing similar latency plus the network latency. But Oracle was seeing latency multiple of what the NFS client was serving, almost 20+ms.