[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: Re: [Xen-users] pfSense HVM
On Sun, 2010-06-06 at 01:29 +0100, Jonathan Tripathy wrote: > On 06/06/10 01:23, Matej Zary wrote: > > Hello, > > > > looks like you have much better HW than I used - when I benched Linux HVM > > guest on Athlon X2 1,9 GHz I got scary results (in a bad way of course) :D. > > 90 mbit/s is not bad at all for emulated hvm guest - the questions is, how > > much CPU power it consumes in the dom0 (qemu-dm process) and how many > > availabe cores do you have. PV driver should solve this issue (or pci > > passthru for the WAN NIC and PV driver for the LAN NIC). I got better > > results is some cases with the default emulated NIC (realtek?), but in > > majority, the emulated e1000 was better. > > > > > > Regards > > > > > > Matej > > > > ________________________________________ > > From: xen-users-bounces@xxxxxxxxxxxxxxxxxxx > > [xen-users-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Jonathan Tripathy > > [jonnyt@xxxxxxxxxxx] > > Sent: 06 June 2010 00:35 > > To: Nicolas Vilz; Xen-users@xxxxxxxxxxxxxxxxxxx > > Subject: Re: [Xen-users] pfSense HVM > > > > > > Hi Matej, > > Well it's a quad core Xeon X3430 with 8GB of RAM. I'm using > PCI-Passthrough for the WAN NIC, and e1000 for the LAN NIC. Since it's > pfsense, it isn't easy (if at all possible) to get this going with PV. > > So even though e1000 is a "gigabit" driver, 90Mb/s is ok for HVM you think? > > Thanks You can try bench the throughput between wan and the pfsense (how fast ist the pci-passthru) and the between the pfsense and another domu or Dom0(how slow is the emulation). Iperf is neat CLI utility for this. :) Also when benching the pfsense-domx, watch the CPU utilization in the Dom0 (top, or better with dstat). The emulation tends to induce quite a cpu load in the dom0 via the qemu-dm process (at least on my shoddy old hw :) ). Emulation is really ineffective and I think that the various emulation NIC models don't have that significant impact on the overall speed - the "slowness" of this method lies in the nature of emulation. :) I attached one graph from my benchmarks - all results are in Mbit/s, iperf TCP bench with 3 different overall frame lengths. PC was another physical computer (same HW config) - so this is physical to virtual bench. Just for illustration of the PV drivers impact. :) Regards Matej <<attachment: winmail.dat>> _______________________________________________ Xen-users mailing list Xen-users@xxxxxxxxxxxxxxxxxxx http://lists.xensource.com/xen-users
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |