AR for Mary: Contact Nathan Lynch to ask him to validate the current scripts/tests for interrupt migration in CPU hotplug testing. AR for Martine: Report the issue w/ the fragmentation avoidance patches in the next DCL TWG meeting Attendees: Mary Edie Meredith, OSDL Joel Schopp, IBM Natalie Protasevich, Unisys George Mann, Indivual Contributor Martine Silbermann, HP * Fragmentation Avoidance Issue. The last submission of the fragmentation avoidance patches led to a lengthy discussion over the proposed approach. There were basically 2 schools of thoughts, one promoting fragmentation avoidance and the other favoring zones to allow memory removal. Mel Gorman gave a detailed comparisons of pros and cons of both approaches. The zone approach showed a lot of potential problems with performance and needed some specific tuning but the people supporting that approach felt that the users wanting hotplug memory remove would deal w/ the performance hit and make the effort to tune their systems. The fragmentation avoidance approach doesn't lead to any performance degradation but adds some complexity to the memory allocator which is a source of concern. At this point it doesn't seem likely that these patches will progress any further. It is unfortunate that the controversy over this approach didn't surface earlier in the development considering that this was rev 19 of the patches and the approach had been presented a few times in the past including at LKS. The developers working on the fragmentation avoidance patches are trying to do lots of simplification to minimize the invasive nature of the approach and increase the chance of the patches still being accepted. We discussed the fact that some of the patches currently under discussion will provide valuable features for data center support and should therefore be of concern to this group, hot-remove is one of them. A proposal was made to support the fragmentation patches as a separate subtree dedicated specifically for data center features but unless we are fairly confident that those patches will eventually be accepted in mainline this just seems to be wasted time and resources. The memory migration patches are in good shape and will probably make it into mainline soon. AR for Martine: Report the issue w/ the fragmentation avoidance patches in the next DCL TWG meeting * Update on memory tests. IBM's tests for user memory removal require a tool that is not open source. Considering the latest development in hot remove there's no real need or incentive to work on open sourcing that tool. HP has an open source tool for page migration that will be added to the OSDL test suite. * Update on CPU hotplug patches. Natalie reported that all her CPU hotplug patches have been accepted in mainline. * Last meeting's ARs. Mary is waiting for feedback from the lhcs list on what the content of /proc should be after a CPU is brought back online. Martine contacted Ric Wheeler from EMC who provided a direct contact in VMware, discussion on possible collaboration w/ VMware are on-going. The original e-mail exchange regarding interrupt migration tests for CPU hot-remove was between Mark Wong and Nathan Lynch. AR for Mary: Contact Nathan Lynch to ask him to validate the current scripts/tests for interrupt migration in CPU hotplug testing. Next meeting is scheduled for November 22nd at 11:00am -12:00pm PST, 2:00pm - 3:00pm EST. Thanks for your participation. Martine J. Silbermann
