Browse Prior Art Database

Coordinating Multiple Server Partitions to Enter Power-Save State

IP.com Disclosure Number: IPCOM000117857D
Original Publication Date: 1996-Jun-01
Included in the Prior Art Database: 2005-Mar-31
Document File: 6 page(s) / 166K

Publishing Venue

IBM

Related People

Kubala, JP: AUTHOR [+2]

Abstract

Disclosed is a method of coordinating multiple server logical partitions on a single computer to enter power save state during a utility power failure. The method allows the individual servers to prepare themselves for power save state and drive the process when they are ready. The hypervisor coordinating the server partitions then places the machine in power save state when all server partitions are ready. Power save state continues to preserve the memory content of the machine, drawing power at much lower rate than normal from an internal battery unit or universal power supply, until utility power is restored. The hypervisor then restarts the server logical partitions to allow continued operation without the loss of data.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 35% of the total text.

Coordinating Multiple Server Partitions to Enter Power-Save State

      Disclosed is a method of coordinating multiple server logical
partitions on a single computer to enter power save state during a
utility power failure.  The method allows the individual servers to
prepare themselves for power save state and drive the process when
they are ready.  The hypervisor coordinating the server partitions
then places the machine in power save state when all server
partitions are ready.  Power save state continues to preserve the
memory content of the machine, drawing power at much lower rate than
normal from an internal battery unit or universal power supply, until
utility power is restored.  The hypervisor then restarts the server
logical partitions to allow continued operation without the loss of
data.

      In this system, a change in power status is indicated by a
machine check.  The actual power status is then read with another
instruction called STORE-WARNING-INFORMATION.  The delivery mechanism
of the status changes are not important here as long as the
hypervisor and the server logical partitions can be made aware of all
the changes.

      The hypervisor maintains a state table for the known power
state.  The initial power state is power-normal, indicating utility
power is present.  The hypervisor must make changes to the power
state and initiate actions when a change in power status is received.
If the hypervisor's power state is power-normal, the hypervisor
starts a task that issues a STORE-WARNING-INFORMATION instruction to
find out what the power status is.  If utility power is found to be
off, the hypervisor maintained power state is changed to power-off.
A power monitor task is then set up to check on the state of logical
partitions, as well as power, every second until either all server
logical partitions have checked in ready to enter power-save state or
utility power is restored.  This task will system reset any
non-essential logical partitions (non-servers) to prepare the
physical system for entering power-save state.

      If utility power is restored prior to all servers checking in,
the state is changed to power-on and any server logical partitions
that had checked in are allowed to resume normal operation as
power-save state has now ended for them.  If all servers check in
while utility power is still off, the hypervisor power monitor task
will issue the real instruction (ENTER-POWER-SAVE-STATE) to place the
entire machine into power-save state.  At the point of issuing the
instruction, all logical partitions have been system reset.  When
utility power is restored, the hypervisor gets control back and
resumes all server logical partitions in the system.  Power state is
then returned to power-normal.

Hypervisor power-save State machine transitions

The following power-save state transitions are depicted in the
Figure:
  o  Power-Normal (0) > Power-Off (1)
     A machine check while in powe...