ISS Technology Update, Vol. 7 Number 6

ISS Technology Update Volume 7, Number 6
8
Multi-core Technology
Traditionally, server manufacturers increase computing performance by plugging a higher frequency processor into the system
board socket. However, that tradition is quickly fading because higher frequency processors use more power (watts) and
generate more heat. Another reason is that the processor performance gains from increasing the frequency are partly negated
by memory latency—known as the Von Neumann bottleneck. As a result, the elevated power/heat/cooling levels coupled with
rising energy costs have caused customers to limit rack density and cap processor power use at the socket. The processor
power issue has motivated manufacturers to place multiple execution cores on a single die to continue the upward “ramp” to
higher processor performance. This processor performance ramp is very important because it is the fuel that drives the entire
computer industry. Some highlights of multi-core processors are described below.
Higher performance per watt
When comparing the performance of dual-core and single-core processors with similar power use, dual-core processor
performance is over 70 percent higher. To match dual-core processor performance, the single-core processor frequency would
have to be over-clocked by 20 percent, but its power use would increase by 10 percent.
Smaller is bigger
Multi-core technology becomes increasingly attractive as the silicon manufacturing process shrinks from 65 nm to 45 nm, and
eventually 32 nm. The reason is that smaller cores require less power, which permits more cores to be built into a single die.
Building 2, 4, 8, and soon 16 cores on a single die will not only have a significant impact on performance and heat
generation, it also reduces noise—less heat requires less fan power.
Virtualization to the core
Virtualization software, such as VMware Infrastructure, VMware Server, Microsoft® Virtual Server, and open source Xen virtual
machines, allows the user to create virtual machines that consolidate multiple operating systems or applications onto a single
physical server. Each virtual machine can have its own virtual processors, drives, NICs, and host bus adapters. Multi-core
processors enable the efficient, parallel operation of multiple operating environments across multiple cores. The ability to run
multiple OSs on multiple cores is cost effective in that the cores share system resources such as memory, I/O, and power
components. Virtualization on multi-core processors also eliminates the need for high-speed communication links between
systems.
Multi-threaded software is key
For customers to exploit the power-saving features of multi-core processors, software developers need to deliver more multi-
threaded applications. Through the operating system, a multi-threaded application sees each core as a logical processor and
can split, or thread, itself to run simultaneously on multiple cores. In contrast, single-threaded applications cannot use multiple
cores simultaneously. Therefore, processor manufacturers allow unused cores to be put in a low-power state when running
single-threaded applications.
Removing disabled servers from HP Systems Insight Manager
Since each customer has a unique way of defining “disabled” servers, HP does not offer a built-in, automated method to delete
disabled servers through HP Systems Insight Manager (HP SIM) other than the manual method. The reason for this is that
although a server may be de-commissioned, deleting it will also delete the event history of that server from the database, and
customers may want to keep that information around for some period of time for reporting purposes.
One manual method for disabling servers is to use the mxnode command available through the HP SIM command line interface
(CLI). The mxnode command can be used in a script to carry out the removal of a specified set of servers. The key would be to
define specific criteria that could be used to identify the devices. For example, a customer can use a string such as “servers that
became critical 30 days or more ago and have had no activity since.”