[clug] Canberra Linux Users Group meeting

Christopher Zhang u4123459 at anu.edu.au
Tue Feb 20 07:15:35 GMT 2007


Canberra Linux Users Group Meeting - 22nd February 2007
	======================================================

Date:		22nd February 2007 (Fourth Thursday of the month)

Time:		19:00 - 21:00 (or when it finishes)

Speaker:	Jeremy Kerr,  Hugh Blemings and Nick Piggin

Abstract:	
			IBM's OzLabs team Open Source Software hackers Hugh Blemings and  
Jeremy Kerr will
			share this opportunity to offer us a talk on running Linux on the  
CELL processor and PS3.
			Followed by Nick Piggin, who is currently working for Novell/Suse,  
will be giving us a presentation
			on his NUMA replicated page-cache for Linux.

			Hugh Blemings and Jeremy Kerr¿s talk will explain the new Cell  
Broadband Engine (CBE), collaboratively
			developed by Sony, Toshiba and IBM. The CBE's novel design aims to  
provide a new level of performance
			for its most well-known application, the PlayStation 3.

			Linux has been a key part of the CBE development process, and is  
under constant development within the IBM
			Linux on Cell team.

			Topics covered include:

			* An introduction to the Cell hardware
			* State of the Linux on Cell effort
			* The Cell programming model, interfaces and methods
			* Current applications and performance
			* Linux on the Playstation 3
			* How to get involved
			* Live demonstrations of the above on a Playstation3

			Followed is Nick¿s NUMA page-cache replication talk, which is  
based on the following:

			In a NUMA system, blocks of memory are divided up in nodes, and  
the memory throughput and latency depends
			on the proximity of the node to the CPU operating on that memory.  
A CPU has a local node, in which it achieves the
			optimal memory performance, and remote nodes, which are slower.

			In the most ideal situation, each CPU in the system would operate  
only on its local memory. Linux tries to achieve
			this goal by allocating local memory by default. This is a  
reasonable strategy, but what happens if memory needs to be
			accessed by different CPUs with different local nodes?

			One approach is to replicate the shared memory. That is, make a  
copy of that memory which is local to each
			CPU that accesses it. This can be achieved by replicating pages  
that are only being read from, and discard the replicas
			when the memory is written to. Despite the limitation to read-only  
memory, replication is applicable to a number of
			important cases such as shared libraries, program files, and  
common data files.


Venue:		Room N101
			Computer Science and Information Technology Building
			North Road
			The Australian National University

			See http://clug.org.au/ for more directions and a map

Food/drink:	Pizza, Sushi and soft drink/juice. Come hungry and bring
			about $6 for Pizza or $8 for Sushi(booking required) to cover the  
cost of your share if you want
			some.

If you would like to give a talk at a future meeting, please email me.



More information about the linux mailing list