Text 6146, 383 rader
Skriven 2005-07-18 16:17:42 av Rich (1:379/45)
Kommentar till text 6143 av Geo (1:379/45)
Ärende: Re: Disk perf
=====================
From: "Rich" <@>
This is a multi-part message in MIME format.
------=_NextPart_000_00AA_01C58BB4.3A406500
Content-Type: text/plain;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
I don't believe you know how the cache works.
Rich
"Geo" <georger@nls.net> wrote in message news:42dc211c@w3.nls.net...
1 9 ms 11 ms 9 ms [216.144.26.33]
2 45 ms 86 ms 55 ms r-cle2.nls.net [216.144.8.2]
3 46 ms 65 ms 44 ms r-cle1.nls.net [216.144.8.1]
4 59 ms 55 ms 53 ms sl-gw4-roa-4-0-ts2.sprintlink.net =
[144.228.252.233]
5 72 ms 53 ms 65 ms sl-bb22-roa-2-1.sprintlink.net =
[144.232.17.205]
6 76 ms 73 ms 56 ms sl-bb22-chi-6-1.sprintlink.net =
[144.232.8.81]
7 60 ms 59 ms 58 ms sl-st21-chi-11-0.sprintlink.net =
[144.232.20.21]
8 58 ms 53 ms 91 ms sl-xocomm-15-0.sprintlink.net =
[144.223.241.66]
9 56 ms 105 ms 56 ms p5-0-0.rar2.chicago-il.us.xo.net =
[65.106.6.137]
10 58 ms 61 ms 94 ms p0-0-0d0.rar1.chicago-il.us.xo.net =
[65.106.1.85]
11 79 ms 114 ms 85 ms p6-0-0.rar2.washington-dc.us.xo.net =
[65.106.0.46]
12 85 ms 91 ms 82 ms p0-0-0d0.rar1.washington-dc.us.xo.net =
[65.106.1.13]
13 78 ms 75 ms 161 ms p6-0-0.rar2.nyc-ny.us.xo.net =
[65.106.0.1]
14 76 ms 81 ms 286 ms p0-0-0d0.rar1.nyc-ny.us.xo.net =
[65.106.1.1]
15 114 ms 76 ms 79 ms p0-0-0.mar1.nyc-ny.us.xo.net =
[65.106.3.46]
16 76 ms 99 ms 78 ms p0-0.chr1.nyc-ny.us.xo.net =
[207.88.86.166]
You probably have about 100ms more lag cause of the distance. Cache =
will still be slow if you have lag because it's still got to resolve and =
check the linked graphics and such before it can match them up with the =
cache.
Geo.
"Rich" <@> wrote in message news:42db0176@w3.nls.net...
It can't get any faster than the network traffic. I'm at least =
15 hops from the nyt. If their servers are on the east cost you will be = much
closer. Since all the downloads are small latency matters more = than
throughput which is why with everything cached it still took six = seconds for
me.
Rich
"Geo" <georger@nls.net> wrote in message =
news:42dafff1@w3.nls.net...
I just clicked and counted, 1...2...3..finished
I'm serious though, set your cache to 1mb, purge it, go to that =
/content.ie5 directory and purge it as well then give it a try.
Geo.
"Rich" <@> wrote in message news:42dafba4@w3.nls.net...
You can't compare two machines connected to different =
networks at different times of the day.
Did you measure the 4 seconds by counting or measuring? I =
used a network capture with timestamps for all the packets. The time I =
reported is from the first request to the last request. I should have =
included the last response so my numbers, which I rounded down to whole =
seconds, would actually be a bit longer.
Rich
"Geo" <georger@nls.net> wrote in message =
news:42dae95a$1@w3.nls.net...
I clicked on your NYT link and my browser started up and the =
page came up in less than 4 seconds total. Perhaps instead of an empty = cache
you should try it with your cache set to 1mb.
Geo.
"Rich" <@> wrote in message news:42d9397d@w3.nls.net...
Because the cache still has a very significant =
performance benefit and even with DSL speeds pages take far longer than = a
blink to appear. For example, I just tried www.nytimes.com with an = empty
cache. There are 86 HTTP requests that take 10 seconds. With the = cache it
only took 6 seconds.
Your ideas on DNS are equally silly, particularly because =
DNS is explicitly a multi-level cache.
Rich
"Geo" <georger@nls.net> wrote in message =
news:42d9304a$1@w3.nls.net...
"Frank Haber" <frhaber@N0SPMrcn.com> wrote in message
news:42d84be9$1@w3.nls.net...
> (Browser cache)
>
> Firefox and Moz come set to 50MB. I find that a good =
round number.
When I was on dialup, 5mb seemed like a good cache (50 =
munutes to download
5mb), because cache made a difference. On todays DSL =
connected computers,
nothing but the previous 4 or so pages should be cached, =
it's just a huge
waste since it only takes a blink to bring up a page on =
DSL. Why todays
browsers continue this outdated practice and don't sense =
the connection
speed and adjust accordingly is beyond me.
It's the same logic as the DNS client cache, both these =
caching functions
create limited problems from time to time, the goal should =
be to get away
from problematic methods and go for reliability. Of course =
that's a concept
that's foreign to most programmers today, they are blind =
to the idea of
dropping an obsolete function and would rather waste time =
trying to cure the
problems it creates.
Geo.
------=_NextPart_000_00AA_01C58BB4.3A406500
Content-Type: text/html;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META http-equiv=3DContent-Type content=3D"text/html; =
charset=3Diso-8859-1">
<META content=3D"MSHTML 6.00.2900.2668" name=3DGENERATOR>
<STYLE></STYLE>
</HEAD>
<BODY bgColor=3D#ffffff>
<DIV><FONT face=3DArial size=3D2> I don't believe you know =
how the cache=20
works.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>Rich</FONT></DIV>
<DIV> </DIV>
<BLOCKQUOTE dir=3Dltr=20
style=3D"PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; =
BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
<DIV>"Geo" <<A =
href=3D"mailto:georger@nls.net">georger@nls.net</A>> wrote=20
in message <A=20
=
href=3D"news:42dc211c@w3.nls.net">news:42dc211c@w3.nls.net</A>...</DIV>
<DIV><FONT face=3DArial size=3D2> 1 9=20
ms 11 ms 9 ms =20
[216.144.26.33]<BR> 2 45 ms =
86=20
ms 55 ms r-cle2.nls.net =
[216.144.8.2]<BR> =20
3 46 ms 65 ms 44 =
ms r-cle1.nls.net [216.144.8.1]<BR> 4 59 =
ms 55 ms 53 ms =20
sl-gw4-roa-4-0-ts2.sprintlink.net [144.228.252.233]<BR> =20
5 72 ms 53 ms 65 =
ms sl-bb22-roa-2-1.sprintlink.net [144.232.17.205]</FONT></DIV>
<DIV><FONT face=3DArial size=3D2> 6 76=20
ms 73 ms 56 ms =20
sl-bb22-chi-6-1.sprintlink.net [144.232.8.81]<BR> =
7 60=20
ms 59 ms 58 ms =20
sl-st21-chi-11-0.sprintlink.net [144.232.20.21]</FONT></DIV>
<DIV><FONT face=3DArial size=3D2> 8 58=20
ms 53 ms 91 ms =20
sl-xocomm-15-0.sprintlink.net [144.223.241.66]<BR> =
9 =20
56 ms 105 ms 56 ms =20
p5-0-0.rar2.chicago-il.us.xo.net [65.106.6.137]</FONT></DIV>
<DIV><FONT face=3DArial size=3D2> 10 58=20
ms 61 ms 94 ms =20
p0-0-0d0.rar1.chicago-il.us.xo.net [65.106.1.85]</FONT></DIV>
<DIV><FONT face=3DArial size=3D2> 11 79 =
ms 114=20
ms 85 ms p6-0-0.rar2.washington-dc.us.xo.net=20
[65.106.0.46]<BR> 12 85 ms 91 =
ms 82 ms p0-0-0d0.rar1.washington-dc.us.xo.net =
[65.106.1.13]<BR> 13 78 ms 75 =
ms 161 ms p6-0-0.rar2.nyc-ny.us.xo.net=20
[65.106.0.1]<BR> 14 76 ms 81=20
ms 286 ms p0-0-0d0.rar1.nyc-ny.us.xo.net=20
[65.106.1.1]<BR> 15 114 ms 76=20
ms 79 ms p0-0-0.mar1.nyc-ny.us.xo.net=20
[65.106.3.46]<BR> 16 76 ms 99 =
ms 78 ms p0-0.chr1.nyc-ny.us.xo.net=20
[207.88.86.166]</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>You probably have about 100ms more =
lag cause of=20
the distance. Cache will still be slow if you have lag because it's =
still got=20
to resolve and check the linked graphics and such before it can =
match=20
them up with the cache.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>Geo.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<BLOCKQUOTE dir=3Dltr=20
style=3D"PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; =
BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
<DIV>"Rich" <@> wrote in message <A=20
=
href=3D"news:42db0176@w3.nls.net">news:42db0176@w3.nls.net</A>...</DIV>
<DIV><FONT face=3DArial size=3D2> It can't get any =
faster than the=20
network traffic. I'm at least 15 hops from the nyt. If =
their=20
servers are on the east cost you will be much closer. Since =
all the=20
downloads are small latency matters more than throughput which is =
why with=20
everything cached it still took six seconds for me.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>Rich</FONT></DIV>
<DIV> </DIV>
<BLOCKQUOTE dir=3Dltr=20
style=3D"PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; =
BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
<DIV>"Geo" <<A =
href=3D"mailto:georger@nls.net">georger@nls.net</A>>=20
wrote in message <A=20
=
href=3D"news:42dafff1@w3.nls.net">news:42dafff1@w3.nls.net</A>...</DIV>
<DIV><FONT face=3DArial size=3D2>I just clicked and counted,=20
1...2...3..finished</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>I'm serious though, set your =
cache to 1mb,=20
purge it, go to that /content.ie5 directory and purge it as well =
then give=20
it a try.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>Geo.</FONT></DIV>
<BLOCKQUOTE dir=3Dltr=20
style=3D"PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; =
BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
<DIV>"Rich" <@> wrote in message <A=20
=
href=3D"news:42dafba4@w3.nls.net">news:42dafba4@w3.nls.net</A>...</DIV>
<DIV><FONT face=3DArial size=3D2> You can't compare =
two machines=20
connected to different networks at different times of the=20
day.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2> Did you measure =
the 4 seconds=20
by counting or measuring? I used a network capture with =
timestamps=20
for all the packets. The time I reported is from the first =
request=20
to the last request. I should have included the last =
response so=20
my numbers, which I rounded down to whole seconds, would =
actually be a=20
bit longer.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>Rich</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<BLOCKQUOTE dir=3Dltr=20
style=3D"PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: =
5px; BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
<DIV>"Geo" <<A=20
href=3D"mailto:georger@nls.net">georger@nls.net</A>> wrote =
in message=20
<A=20
=
href=3D"news:42dae95a$1@w3.nls.net">news:42dae95a$1@w3.nls.net</A>...</DI=
V>
<DIV><FONT face=3DArial size=3D2>I clicked on your NYT link =
and my browser=20
started up and the page came up in less than 4 seconds total. =
Perhaps=20
instead of an empty cache you should try it with your cache =
set to=20
1mb.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>Geo.</FONT></DIV>
<BLOCKQUOTE dir=3Dltr=20
style=3D"PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: =
5px; BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
<DIV>"Rich" <@> wrote in message <A=20
=
href=3D"news:42d9397d@w3.nls.net">news:42d9397d@w3.nls.net</A>...</DIV>
<DIV><FONT face=3DArial size=3D2> Because the =
cache still=20
has a very significant performance benefit and even with DSL =
speeds=20
pages take far longer than a blink to appear. For =
example, I=20
just tried <A =
href=3D"http://www.nytimes.com">www.nytimes.com</A> with=20
an empty cache. There are 86 HTTP requests that take =
10=20
seconds. With the cache it only took 6 =
seconds.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2> Your ideas on =
DNS are=20
equally silly, particularly because DNS is explicitly a =
multi-level=20
cache.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>Rich</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<BLOCKQUOTE=20
style=3D"PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: =
5px; BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
<DIV>"Geo" <<A=20
href=3D"mailto:georger@nls.net">georger@nls.net</A>> =
wrote in=20
message <A=20
=
href=3D"news:42d9304a$1@w3.nls.net">news:42d9304a$1@w3.nls.net</A>...</DI=
V>"Frank=20
Haber" <<A=20
=
href=3D"mailto:frhaber@N0SPMrcn.com">frhaber@N0SPMrcn.com</A>>=20
wrote in message<BR><A=20
=
href=3D"news:42d84be9$1@w3.nls.net">news:42d84be9$1@w3.nls.net</A>...<BR>=
>=20
(Browser cache)<BR>><BR>> Firefox and Moz come set =
to=20
50MB. I find that a good round number.<BR><BR>When I =
was on=20
dialup, 5mb seemed like a good cache (50 munutes to=20
download<BR>5mb), because cache made a difference. On =
todays DSL=20
connected computers,<BR>nothing but the previous 4 or so =
pages=20
should be cached, it's just a huge<BR>waste since it only =
takes a=20
blink to bring up a page on DSL. Why todays<BR>browsers =
continue=20
this outdated practice and don't sense the =
connection<BR>speed and=20
adjust accordingly is beyond me.<BR><BR>It's the same =
logic as the=20
DNS client cache, both these caching functions<BR>create =
limited=20
problems from time to time, the goal should be to get =
away<BR>from=20
problematic methods and go for reliability. Of course =
that's a=20
concept<BR>that's foreign to most programmers today, they =
are=20
blind to the idea of<BR>dropping an obsolete function and =
would=20
rather waste time trying to cure the<BR>problems it=20
=
creates.<BR><BR>Geo.<BR><BR></BLOCKQUOTE></BLOCKQUOTE></BLOCKQUOTE></BLOC=
KQUOTE></BLOCKQUOTE></BLOCKQUOTE></BLOCKQUOTE></BODY></HTML>
------=_NextPart_000_00AA_01C58BB4.3A406500--
--- BBBS/NT v4.01 Flag-5
* Origin: Barktopia BBS Site http://HarborWebs.com:8081 (1:379/45)
|