<html>
<head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
</head>
<body text="#000000" bgcolor="#FFFFFF">
Hello Mr. Eromenko,<br>
<br>
thank you for porting this patch on beta version.<br>
<br>
As for performance<br>
<blockquote
cite="mid:CAOJ6w=FDiYeANeVGuHoz6BR3kKBtBV91hi_s1mnvX6JPdfSCwQ@mail.gmail.com"
type="cite">
<pre wrap="">Command to clone 70 VMs:
$ for i in `seq 1 70`; do VBoxManage clonevm "ZeroVM" --mode all
--register --name "ZeroVM Clone $i"; done
70 VMs, Host-only-interface mode:</pre>
</blockquote>
I think it can happen because all your VMs have only one interface
vboxnet0 (if I'm not mistaken?). In our case we have separate
interfaces for every VM (<i>vboxnet1..80</i>), and all of them are
in <i>/etc/network/interfaces</i>, so host interfaces list updating
takes significantly longer time.<br>
<br>
<blockquote
cite="mid:CAOJ6w=FDiYeANeVGuHoz6BR3kKBtBV91hi_s1mnvX6JPdfSCwQ@mail.gmail.com"
type="cite">
<pre wrap="">unpatched:
2:30 min boot. -- 2:33 min (2nd run)
patched: (slower by margin of error)
2:35 min boot. .. 2:34 (2nd run)</pre>
</blockquote>
What does this time mean? If I understand you right, you start VMs
without OSes, but for simple VM starting this time looks too long...<br>
<br>
<blockquote
cite="mid:CAOJ6w=FDiYeANeVGuHoz6BR3kKBtBV91hi_s1mnvX6JPdfSCwQ@mail.gmail.com"
type="cite">
<pre wrap="">Timed manually from smartphone, so 1 second mistake is possible. (not
via host OS timer)
</pre>
</blockquote>
We use time field in VBoxSVC log for chronometry (we attached <i>VBoxSVC.log</i>s
to first letter). This way you can measure execution time for
specific procedure (<i>FindHostNetworkInterfaceByName</i> in our
case) and see how this patch affects this.<br>
<br>
Regards,<br>
Alexander<br>
<br>
</body>
</html>