On Sat, Jul 24, 2010 at 4:01 PM, Joerg Schilling <[email protected]> wrote: > Moinak Ghosh <[email protected]> wrote: > >> I am taking guesses here but I suspect this boot performance may >> be relative. AFAIK SMF does not restrict the extent of concurrency. >> So if there are 100 non-interdependent services ready to be started >> at the current graph state then it will start all 100 concurrently. This >> might be good for multi-way big-iron boxes but can be at the >> diminishing returns point for smaller workstations and laptops where >> there is neither the disk/memory bandwidth and nor the core-count to >> handle such a load. In this case a feature to have a graceful reduction >> in concurrency level in SMF should help. > > It may even be seek times are the main problem... >
True. Too much concurrency on spinning platters can cause seek time bottlenecks. Possibly an iostat log during boot can provide some clues. If true then the situation will be better with SSDs. Regards, Moinak. -- ================================ http://www.belenix.org/ http://moinakg.wordpress.com/ _______________________________________________ opensolaris-discuss mailing list [email protected]
