From: Gerard Henry (ghenry@cmi.univ-mrs.fr)
Date: Tue May 18 2004 - 05:21:26 EDT
Sorry for this late summary, i was very busy with many others concerns.
First i will thank people who responded, and responses was very instructives.
Lars Hecking:
Tim Chipman
Chris Hoogendyk
Marcelino Mata
Daniel.Jaime
I don't receive what i was waiting for, but some clues (sorry for my poor english vocabular):
First, use firefox, and i try it on a blade100, it seems to work well, except for printing. You have to use Xprint explained here:
Sun InfoDoc #ID71966
but i don't want to install GISWxprint and GISWxprintglue on my prod server.
And it needs somme scripting if you want your users use firefox and thunderbird instead of mozilla, config files are not in .mozilla directory
M. Mateo try to explains me the measures of footprint, but i think i need to compare it with some others products like firefox
Two points:
- first, i want to use sun packages instead of other contributed packages, on a prod server, it simplifies management, and i have some support from sun in case of misfunction
- two, T. Chipman talk about apps launch from a linux server, yes i think he's right, but as long as i can buy sparc machines, i don't want to administer pc machines in server's office, it's a choice...
here is the responses i received:
Lars Hecking:
As you mention junk control, I presume they are using mozilla for mail?
If it's not too big a deal, I'd recommend to use mozilla thunderbird for
mail, and firefox for the web. thunderbird can only do pop and imap, though,
it can't read a (local of NFS mounted) mail spool.
I don't use thunderbird myself. We have 0.6 installed on Linux, and 0.5
on Solaris, I'm waiting for a contributed build (there is one for Solaris
8 x86 already). None of the thunderbird users have come back with feedback
yet, but there aren't many.
We have firefox 0.8 on Solaris and Linux, and I have yet to run into any
bugs. I made it my browser of choice nearly 3 months ago, and did not go
back to mozilla. The only problem is that an additional package (xprint)
is required for printing, but I don't print off the web very often.
Tim Chipman
for what it may be worth: we're running a sunray environment here, and
I gave up on running netscape/mozilla on the sunray server itself a long
time ago. We deployed a "small" dual-2400mhz athlon (linux) app-server
where we run mozilla, openoffice, etc... and then xforward these apps to
users on the sunray environment. Since we're using centralized kerberos
authentication, this can all be done transparently - the user on the
sunray doesn't really even know their mozilla (etc) are not running
locally. The net result is that the sunray server loads are infinitely
more reasonable ; the sunray sessions are responsive ; the athlon app
server does all the "grunt work" of actually running mozilla
(openoffice, gimp, etc) and of course, since athlon app servers are
virtually free, we prolong the life of our sunray server at minimal net
effective cost.
Of course, our deployment here isn't huge - we have typically between 20
and 25 concurrent sunray users, on a total of about 30-35 sunray
workstations - and we're using an 4x400mhz e450 as the sunray server.
(we had been using a 2x400 e250 until about 6 months ago, but had an
easy opportunity to swap in a 450. With current usage, I believe we can
easily scale to 50 concurrent users before we need to get worried about
needing any more upgrades...)
and, to say it clearly: The linux app server, despite being a "generic
clone box" system, has been absolutely rock-solid. No crashes,
instabilities, and the performance is phenomenal. (in my testing,
athlons run approx equivalent MHZ yield as per UltraSparc II or III
CPU's, ie, early benchmarks I did on an older dual-athlon indicates that
a 2x1600mhz athlon system with "3200mhz total" had virtually identical
CPU performance to an 8x400mhz {ie, also "3200mhz total"} e3500 system
when running "Blast", a CPU-intensive app we use here frequently).
Hope this is of some use/interest,
Chris Hoogendyk
for that very reason, I have been thinking about going to Mozilla
FireFox. haven't had time to get to it yet. take the savings in memory
useage on the one server and multiply it by the dozens of SunRay users
on that server, and it's not small peanuts.
Marcelino Mata
Using RMCmemtool memps command I get the following :
# ./memps
PID Size Resident Shared Anon Process
18130 53424k 37800k 26024k 11776k Mozilla.org 1.6 build for
Solaris
18202 46920k 37216k 27616k 9600k SUN Mozilla 1.2.1
20432 42328k 40152k 28136k 12016k SUN Mozilla 1.4
18224 40032k 32128k 21352k 10776k Mozilla.org 1.5 build for
Solaris
I started Mozilla with blank screen. I agree that the memory foot print is
getting bigger. I'm not sure if memory size should be based on Size or
Resident+Shared. Using Resident+Shared, 1.2.1 is 64.8Mb vs 68.2Mb for 1.4.
How are you getting 30Mb for 1.2.1? From research, RMCmemtool gives more
accurate information vs Solaris "ps -ef", top, Process Manager etc.
I understand that Mozilla 1.7 is suppose to decrease the foot print to
Mozilla 1.5 level. I do not know if that will mean anything to SUN released
builds. It looks like Sun 1.4 includes some code from the 1.6 release (i.e.
single sign-on).
In general I have noticed that since Mozilla is based on the reference
platform of Linux and Windows, that the Solaris build is larger since it
based on older GNOME technology and does not have certain Linux libraries
like it's font server etc...Optimization wise, Mozilla will probably be
smaller on Linux.
I was getting different numbers so I rebooted and tested again. I also
installed Mozzila 1.7rc1. Installing SUN 1.4 unstalled 1.2.1 so I don't
have those numbers...
Size Resident Shared Anon Process
57560k 42208k 28056k 14152k mozilla 1.6
37384k 30568k 19840k 10728k mozilla 1.5
44008k 41960k 33432k 8528k SUN 1.4
56424k 39544k 28776k 10768k mozilla 1.7rc1
Here's the explanation of memory usage from memtool docs....attached is the
docs for your reference.
Resident - The amount of physical memory that this file has associated with
it.
Used - The amount of physical memory that this file has mapped into a
process segment or SEGMAP. Generally the difference between this and the
resident figure is what is on the free list associated with this file.
Shared - The amount of memory that this file has in memory that is shared
with more than one process.
Daniel.Jaime
why don't you install the actual official stable release 1.6 from
http://ftp.mozilla.org/pub/mozilla.org/mozilla/releases/mozilla1.6/contrib/mozil
+la-sparc-sun-solaris2.8-1.6.tar.gz
.
A lot of severe security problems were fixed in this version. The README
file can be found at
http://ftp.mozilla.org/pub/mozilla.org/mozilla/releases/mozilla1.6/contrib/mozil
+la-sparc-sun-solaris2.8-1.6.readme
,
the Downloadpage at http://www.mozilla.org/releases .
Concerning the print environment, you can find in the README file the following
information
* Printing is supported via Xprint
(see http://www.mozilla.org/releases/mozilla1.4rc3/index.html#xprint)
xprint.mozdev.org provides various Solaris packages including
"GISWxprint"
and "GISWxprintglue", for people without (root) permission to install
packages
the plain tarballs are provided, an alternative prodecure is described
in
Sun InfoDoc #ID71966 about the same issue (see
http://sunsolve.sun.com/pub-cgi/retrieve.pl?doc=finfodoc%2F71966 - "How
to add
printers to Netscape[TM] and Mozilla[TM]?").
I hope this can help you.
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:28:41 EDT