Browse Prior Art Database

Daemon management to optimize performance in a virtualized environment

IP.com Disclosure Number: IPCOM000237023D
Publication Date: 2014-May-27
Document File: 2 page(s) / 21K

Publishing Venue

The IP.com Prior Art Database

Abstract

A method for daemon management to optimize performance in a virtualized environment is disclosed.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 01 of 2

Daemon management to optimize performance in a virtualized environment

Disclosed is a method for daemon management to optimize performance in a virtualized environment.

Virtualization is a hot trend in most operating systems, including AIX

                                           ® . AIX can run multiple instances of itself known as system workload partitions (WPARs), each appearing as an isolated operating system (OS) image. Operating systems were implemented before the concept of virtualization became mainstream, thus they are not designed to handle problems that are introduced by this new methodology. One problem is the fact that AIX, like other operating systems, have multiple daemons running within each instance of the OS. These daemons tend to average in the teens, and run periodically within AIX, conducting various tasks. The problem occurs when there are multiple WPARs running, each with multiple daemons per AIX instance. The number of daemon processes running on the system can quickly grow, and saturate the CPU resources. So even if the WPARs are idle, the daemons running within each WPAR, can saturate enough of the CPU, rendering the system unusable. There are a number of ways this exponential growth of daemons can be curtailed in OS virtualization such as WPAR technology. One solution would be over loading the daemon in the Global WPAR to service as a surrogate for each of the WPAR. For example type X daemon from each WPAR would register its queue address with type X daemon in the Global WPAR. Type X daemon in the WPAR will go into sleep woken by the type X daemon from the Global WPAR only when there is work (event) to process. In each WPAR, events of type X which are targeted for type X daemon get queued into the type X queue. Since the type X queue address is registered with type X daemon in the Global WPAR, every time type X daemon wakes up in Global WPAR on an interval checks the queue for events. If an event is waiting in the queue for WPAR1, then global WPAR type X daemon will wake type X daemon in WPAR1. The part where Global WPAR manages this overloading of daemon is called daemon management layer.

Since the global WPAR has access to each of the WPAR memory region, registering (passing) event queue address with Global WPAR or Global WPAR accessing event queue of the Guest WPAR is not a breach to the security and isolation of WPAR. In cases with a very large number of WPARs, a single type X daemon in the Global WPAR may not be able to scale without introducing latency, in which case multiple type X daemons in the global WPAR can be used to manage large number of WPARs.

In multi-node systems, Global WPAR can assign one type X daemon per node (if Global WPAR spans across nodes) to scale based on the number of WPARs within ea...