« previous 1 2 3 4 5 6 7 next »

Sit back and relax until the next call, part two

In version 2.6.24, developers also integrated High Resolution Timer support for x86-64, PPC and MIPS architectures - x86-32 systems have already been supported since Linux 2.6.21. The kernel can now co-operate with hardware which supports HPET (High Precision Event Timers) or other precision timing technologies to execute predefined tasks at a predefined time much more precisely than before. Among other things, this is important for real time applications but also offers advantages for other types of time-critical software - for example for audio processing.

Like the High Resolution Timers for 32-bit systems in 2.6.21, the varieties for other architectures come with Dynamic Ticks (Tickless Idle) functionality. In modern systems, the kernel can use these extensions to program the timer interrupt for the next scheduled task. This way, the processor can rest until its next call in an otherwise idle system, and can enter energy saving mode until the next task or IRQ comes up. This is intended to reduce power consumption and prolong the battery life of notebooks, as so far, the CPU had to wake up and process the constantly ticking timer IRQ between a hundred and a thousand times per second, depending on kernel configuration. However, in order to achieve a real reduction in power consumption, applications must not wake up the processor unnecessarily either - this is only the case with the very latest distributions which have been optimised in this way. The Powertop program helps to find applications which wake up the processor unnecessarily.

Many x86 notebooks with 32-bit distribution should also have more battery life with Linux 2.6.24 as the new kernel offers cpuidle for more efficient power saving with modern x86 processors. Support of power saving techniques is said to be particularly effective in connection with the HD audio sound drivers of newer (Centrino) notebooks, where it is said to lower power consumption by half a watt in some notebooks. The newly integrated SATA Link Power Management in the ahci driver should be an advantage mainly for newer devices - according to the developers, potential savings could be up to one watt per SATA connection. However, aggressive ATA power saving reduces processing speed.

Virtualisation

2.6.24 offers various improvements for container/control group virtualisation technologies like OpenVZ or Linux VServer; for example /proc/net directory abstraction, separate namespaces for the processes of a control group, process scheduler support, the veth virtual network driver and write-protected bind-mount support.

The kernel's own KVM virtualisation component now emulates the various different interrupt controllers found in modern systems (PIC, Local APIC, I/O APIC) itself. This is intended to improve performance and compatibility, which is also true for emulating HTL instructions. After adjustments, x64 Windows is now also said to run under KVM.

From now on, fast data exchange between guest and host systems is to be achieved via the Virtio interface. Developers hope that the various virtualisation approaches will emulate their storage or network controllers in a Virtio-compatible way - due to the generic Virtio driver, for example, a guest system using Lguest should also boot under KVM.

« previous 1 2 3 4 5 6 7 next »