On Wed, Mar 19, 2008 at 03:10:12PM -0400, Chuck Robey wrote: >it might be possible to find some way to extend the work domain of an smp >system to stretch across machine lines, to jump across motherboards. Maybe >not to be global (huge latencies scare me away), but what about just going >3 feet, on a very high speed bus, like maybe a private pci bus? Not what >is, what could be?
This is definitely possible. DEC built a memory channel adapter which allowed multiple AlphaServers to share (part of) each other's RAM. You could also try looking at Amoeba - it is a cross between a "traditional" SMP system and a cluster. There's probably no reason why you couldn't build a kernel module to "share" RAM between hosts using Ethernet or similar - though it would be much slower than accessing local RAM. >small, but with a bunch of bandwidth. So, in that case, what really are >the differences between smp and clustering, besides the raw current size of >the implementation? Are there huge basic differences between the >clustering concept and by smp's actual tasks? The access time differences between local and remote RAM mean that there are different trade-offs: Memory coherence is extremely expensive so more effort is expended in avoiding operations that require coherence. In general tasks that work well in a clustered environment have very low inter-process communication requirements. -- Peter Jeremy Please excuse any delays as the result of my ISP's inability to implement an MTA that is either RFC2821-compliant or matches their claimed behaviour.
pgpYKpCyNy1aN.pgp
Description: PGP signature
