On Thu, Dec 04, 2008 at 03:16:58PM +0000, Chris Webb wrote: ... > This is no longer the case: I fixed this using BPF/linux socket filters in a > patch that was distributed with vblade-16 and went into the main vblade > codebase in vblade-18. A vblade instance will now only wake up for a AoE > packet which is destined for the matching slot and shelf number. > > I'm exported several hundred AoE devices per host over gigabit ethernet, and > see marginally better performance from a collection of vblade processes than > a qaoed instance, despite the greater code complexity of the latter.
Yes, that work helped vblade to scale more than was possible before. It's also worth noting that the Linux kernel doesn't use RAM to store multiple copies of read-only text (executable data) from binaries or from the libraries they use, so I would expect a hundred idle vblades less than one hundred times as much RAM as a single vblade. -- Ed Cashin <[EMAIL PROTECTED]> ------------------------------------------------------------------------------ SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ _______________________________________________ Aoetools-discuss mailing list Aoetools-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/aoetools-discuss
