To optimize performance, operating system and WebSphere Application Server paramaters are automatically tuned in the IBM® WebSphere® Application Server Hypervisor Edition virtual image.
1 echo 1 > /proc/sys/kernel/sched_compat_yield
Passing the clock=pmtmr boot parameter to a virtual image with a VMI kernel might cause a performance degradation
The default virtual machine interface (VMI) kernel does not pass the clock=pmtmr parameter to the kernel. However, if you pass the clock=pmtmr boot parameter to a virtual image with a VMI kernel, you might see a performance degradation. You can check for a performance degradation with standard performance bench-marking software, such as SPECjAppServer. To solve this problem, use an editor, such as vi, and open the /boot/grub/menu.lst file and remove the clock=pmtmr boot parameter on the kernel command line for the desired VMI kernel. Example: /boot/grub/menu.lst
WebSphere Hypervisor Edition automatically tunes application servers and deployment managers based on the virtual resources allocated to the virtual image. The automatic tuning attempts to optimize your virtual system for performance, but can easily be changed to meet the specific needs of your application. The Java heap size is sized based on the allocated virtual memory; WebSphere Application Server web container threads are sized based on the number of virtual CPUs.
Minimum/Maximum JVM Heap = max(min((Virtual Memory - 768MB) / # Application Server Instances, 2048MB), 256MB)Depending on the amount of virtual memory allocated to a virtual image, distribute as much memory as possible to each application server and deployment manager instance. After reserving 768MB for the OS and other native processes, the remaining memory is divided evenly between JVMs. The automatic tuning algorithm will assign no less than 256MB, and no more than 2GB for a given JVM.
Minimum/Maximum Threads = # Virtual CPUs * 10A factor of 10 modifies the number of web container threads, according to the number of virtual processors in the virtual image. The purpose is to provide additional worker threads to effectively utilize systems with more virtual CPU resource.