Here's my DNS histograms.  I visited cnn.com, bbcnews.com, news.com.
All three sites took an extremely long time to load, and the first two
sites failed DNS lookup the first time, I had to hit the refresh
button.  I'm using Vmware Fusion 2 running windows XP, Chrome
1.0.154.48, with OpenDNS as DNS host.  I do not see a problem with
"Use DNS pre-fetching to improve page load performance" turned off
under Vmware,  and I also do not see this problem in non-vmware
environment on my other windows XP laptop.


Histogram: DNS.PrefetchCacheEviction recorded 4 samples, average =
8987.0, standard deviation = 8090.6 (flags = 0x1)
     0 ...
   533
------------------------------------------------------------------------
O (1 = 25.0%) {0.0%}
   633 ...
  3553
------------------------------------------------------------------------
O (1 = 25.0%) {25.0%}
  4222 ...
  8416
------------------------------------------------------------------------
O (1 = 25.0%) {50.0%}
 10000
------------------------------------------------------------------------
O (1 = 25.0%) {75.0%}


Histogram: DNS.PrefetchFoundNameL recorded 28 samples, average =
5403.7, standard deviation = 3934.2 (flags = 0x1)
     0 ...
   539 --------------
O                                                           (1 = 3.6%)
{0.0%}
   730 --------------
O                                                           (1 = 3.6%)
{3.6%}
   989 -----------------------------
O                                            (2 = 7.1%) {7.1%}
  1340 --------------
O                                                           (1 = 3.6%)
{14.3%}
  1816 -------------------------------------------
O                              (3 = 10.7%) {17.9%}
  2460 -------------------------------------------
O                              (3 = 10.7%) {28.6%}
  3333
------------------------------------------------------------------------
O (5 = 17.9%) {39.3%}
  4516 --------------
O                                                           (1 = 3.6%)
{57.1%}
  6118
------------------------------------------------------------------------
O (5 = 17.9%) {60.7%}
  8289 ----------------------------------------------------------
O               (4 = 14.3%) {78.6%}
 11230 -----------------------------
O                                            (2 = 7.1%) {92.9%}
 15215 ...


Histogram: DNS.PrefetchQueue recorded 32 samples, average = 527.7,
standard deviation = 992.4 (flags = 0x1)
    0
------------------------------------------------------------------------
O (12 = 37.5%)
    1 ------------------------------------------------
O                         (8 = 25.0%) {37.5%}
    2 ------
O                                                                   (1
= 3.1%) {62.5%}
    3 ...
    5 ------------------
O                                                       (3 = 9.4%)
{65.6%}
    6 ...
  752 -
O
(1 = 3.1%) {75.0%}
  894
O
(0 = 0.0%) {78.1%}
 1062 -
O
(1 = 3.1%) {78.1%}
 1262
O
(0 = 0.0%) {81.3%}
 1500 -
O
(1 = 3.1%) {81.3%}
 1782 -
O
(1 = 3.1%) {84.4%}
 2117 -
O
(1 = 3.1%) {87.5%}
 2516 --
O
(2 = 6.3%) {90.6%}
 2990 -
O
(1 = 3.1%) {96.9%}
 3553 ...


Histogram: DNS.UnexpectedResolutionL recorded 27 samples, average =
4771.7, standard deviation = 6310.0 (flags = 0x1)
     0
------------------------------------------------------------------------
O (6 = 22.2%)
     1 ...
   294 -------
O                                                                  (3
= 11.1%) {22.2%}
   398 --
O
(1 = 3.7%) {33.3%}
   539 -------
O                                                                  (3
= 11.1%) {37.0%}
   730 --
O
(1 = 3.7%) {48.1%}
   989
O
(0 = 0.0%) {51.9%}
  1340 -----
O
(2 = 7.4%) {51.9%}
  1816 ...
  3333 -----
O
(2 = 7.4%) {59.3%}
  4516 --
O
(1 = 3.7%) {66.7%}
  6118 --
O
(1 = 3.7%) {70.4%}
  8289 -----
O
(2 = 7.4%) {74.1%}
 11230 ----------
O                                                               (4 =
14.8%) {81.5%}
 15215
O
(0 = 0.0%) {96.3%}
 20614 --
O
(1 = 3.7%) {96.3%}
 27929 ...


On Feb 10, 9:41 am, Jim Roskind <[email protected]> wrote:
> Your browser is instrumented to reveal some data that would be helpful
> in understanding this performance.  Since you're seeing times greater
> than 60 seconds, something big should show up.  If possible, please
> visit (in chromium):
>
> about:histograms/DNS
>
> You'll see a pile of ASCII art histograms, which reflect what has been
> going on with regard to DNS fetching.  None of the information should
> be very confidential, so if possible, please cut/paste/post a copy
> here for folks to look at.
>
> It may, for example, be that the DNS resolver in Vmware is restricting
> the the rate of resolutions, rather than allowing them to happen in
> parallel.  Without taxing most computers or networks, Chromium
> currently tries to do up ot 8 resolutions in parallel.  If this were a
> problem, then you'd see a large average queueing delay (as resolutions
> backed up into the queue) in the histogram labeled DNS.
> PrefetchQueue.  The median for that histogram should be tiny (around
> 1ms), but the mean could be much larger if you visit pages with a LOT
> of domains (wikipedia comes to mind).    Even so, I'd be rather
> surprised to hear about queuing delay over 1 second.
>
> As a second example, for some reason your DNS resolution time may be
> terrible.  If so, you'd see this in the histogram labeled
> DNS.PrefetchFoundNameL (where I'd commonly expect to see a median
> value around 85ms, and mean value up closer to 200-600ms).  If you see
> a long tail (resolutions larger than a second or two, meaning larger
> than 1000-2000ms along the axis on the left), then there may be a
> problem with your resolver or your network.  I've seen especially long
> tails in negative results (host not found) on networks when  netbios
> kicks in and does various work to broadcast in order to find local
> hosts.
>
> Again, please post a copy of your  DNS histograms after you encounter
> this problem.
>
> Thanks,
>
> Jim
>
> On Feb 9, 8:06 pm, Bob Doe <[email protected]> wrote:
>
> > I'm running Chrome under Vmware Fusion 2.0.  If I have "Use DNS pre-
> > fetching to improve page load performance" turned on, many websites I
> > visit seem to stall for a long time (> 60 seconds) at "Resolving
> > host..."  before web pages are rendered.  I have found other people on
> > the web complain about the same problem, what is the underlying cause,
> > and is there plan to fix this problem in upcoming releases?
--~--~---------~--~----~------------~-------~--~----~
Chromium Discussion mailing list: [email protected] 
View archives, change email options, or unsubscribe: 
    http://groups.google.com/group/chromium-discuss
-~----------~----~----~----~------~----~------~--~---

Reply via email to