Oracle: Unbreak my Linux Heart

Oracle Linux Tests are designed to verify Linux kernel functionality and stability essential for the Oracle Database. The Oracle Linux Test (OLT) kit , which is distributed as a set of rpms, provides an automated mechanism to define, execute and analyze tests. The kit contains tools such as DBT2 and DBT3 workloads against Oracle single instance or clustered databases, as well as specialized workload simulators. OLT can be used for running tests on Oracle Enterprise Linux, RHEL and SLES distributions and a variety of topologies (storage attached network, network attached storage).

Since we are virtualizing many of our Linux servers, this post was of interest to us: "VMmark 1.0 utilized only 32-bit workloads, which was a reasonable mix when the benchmark was first defined roughly three years ago. However, 64-bit applications and OSes are becoming much more prevalent and we need the ability to characterize this more complex reality. To address this, [ for vmware 1.1 ] we have transitioned three of the VMmark workloads - Java server, database server, and web server - to 64-bit. In order to maintain comparability with the existing version 1.0 results, we have retained the underlying virtual hardware definitions and load levels for each workload." This will let us benchmark our specific ESX host setup and test Oracle Linux VM servers under various settings...

While we ourselves did not have any performance issues of our Dell Poweredge 2950 using 2 Qlogic qle2460 HBAs to connect to an EMC DMX-3 SAN over Cisco FC switches, some of our colleagues at another subsidiary did have problems: only 50MB/s using the bonnie disk benchmark where we got 240MB/s... Clearly something was different. Investigations showed they were using LVM on the LUNs and created partitions within LVM. We did not use LVM but gave bare LUNs to OracleASM to manage and control. This led to believe that misalignment of the partition to physical disk cylinder may have caused additional overhead. Because LVM created an MBR on the disk's first set of cylinders, the first partition was created starting at cylinder 63. As this is not a power of 2, the disk subsystem experiences overhead as it translates the numbers back and forth. Research showed that both Oracle and EMC's best-practices both recommend to align your partitions on power of two cylinders! There is a best-pra...

Found a nice article on a benchmark study of the new, modern file systems that are coming up. ZFS, XFS, and EXT4 file systems compared shows how ZFS, XFS and ext4 measure up against each other. Of course, there is no reference measurement of zero-line to compare against. But perhaps other benchmark studies can provide this...

Today, Oracle announced a new TPC-C benchmark result on an HP ProLiant ML350 with the best ever price/performance. This result sets a world record in the price/performance category, which measures the cost per transaction of the configuration benchmarked. It demonstrates that Oracle Database not only provides customers with a highly performant and scalable data management solution, but can also deliver unparalleled cost savings with record-breaking performance on HP ProLiant systems, ideally suited to small and medium sized businesses (SMB). Achieving 100,926 transactions per minute with a price/performance of $.78/tpmC, Oracle Database 10g Release 2 Standard Edition One delivered 45 percent more performance at 14 percent less cost than its nearest competitor. To set this new TPC-C price performance record, Oracle Database 10g Release 2 Standard Edition One was run on an HP ProLiant ML350 server with one Intel Quad-Core Xeon 2.6 GHz processor and HP StorageWorks arrays. The operating s...

Kevin Closson points to a test Dell did using their hardware and an Oracle et-up usign RAC and one without RAC. The results are interesting! The tests are run using Dell's DVD Store Test Application using a PowerEdge 2950 cluster and one PowerEdge 6850 .