neal@mnopltd.UUCP (04/24/91)
The current (maybe hopefully past) ankle-biting parade over SCO is of no
interest to me. SCO is as great as the world deserves. Now, on to
business.
I noticed that the Xenix 2.3.2 kernel is about 1000K. The SCO Unix
kernel is about 2000K. All worthwhile, no doubt. However, one of the
new machines I just brought up is about 1000k short of memory and is
paging under heavy load. (> 18 users)
Are there any things I can safely reconfigure out which will get me
down to the size of Xenix? I was hoping in terms of features we don't
use like RFS, NFS, TCP, etc.
------------------------------------------------------------------------------
Neal Rhodes MNOP Ltd (404)- 972-5430
President Lilburn (atlanta) GA 30247 Fax: 978-4741
emory!mnopltd!neal
gatech!emory!mnopltd!neal
------------------------------------------------------------------------------marc@jahangir.UUCP (Marc Rossner) (04/24/91)
In article <204@mnopltd.UUCP>, neal@mnopltd.UUCP writes: > The current (maybe hopefully past) ankle-biting parade over SCO is of no > interest to me. SCO is as great as the world deserves. Now, on to > business. > > I noticed that the Xenix 2.3.2 kernel is about 1000K. The SCO Unix > kernel is about 2000K. I thought that the advantage of Xenix was that it was supposed to be tiny. I also thought that I had a rather massive kernel on my ISC 2.2. My kernel is 750K. Are you sure you have your numbers right? Marc Rossner jahangir!marc@uunet.uu.net
rbraun@spdcc.COM (Rich Braun) (04/25/91)
marc@jahangir.UUCP (Marc Rossner) writes: >I thought that the advantage of Xenix was that it was supposed to be tiny. >I also thought that I had a rather massive kernel on my ISC 2.2. My >kernel is 750K. Are you sure you have your numbers right? Reading this thread, I can't help but to wonder: why worry about kernel size, these days? I've long been one to complain about the fact that software seems to get larger in direct proportion to the decrease in memory costs, and often slower due to its increasing complexity, but in the case of a reasonably well-performing O/S with lots of features, why worry so much about kernel size? Add another megabyte to the system and the problem will go away. Seems a fairly simple and economical solution. Even at 1-2Mb, kernels remain significantly smaller than most applications. (As compared to ten years ago on mainframe computers, when a kernel was typically many times larger than an application.) -rich
ronald@robobar.co.uk (Ronald S H Khoo) (04/25/91)
marc@jahangir.UUCP (Marc Rossner) writes: > In article <204@mnopltd.UUCP>, neal@mnopltd.UUCP writes: > > I noticed that the Xenix 2.3.2 kernel is about 1000K. The SCO Unix > > kernel is about 2000K. > I thought that the advantage of Xenix was that it was supposed to be tiny. Well, it's *comparatively* tiny. It's *comfortable* to do useful work on a 2 Mb RAM 40 Mb disc Xenix machine, so long as no networking is required. That's HUGE by comparison with what you needed to do useful work under V7 on a PDP, but tiny compared to what you need for any comparable 386 System V. > I also thought that I had a rather massive kernel on my ISC 2.2. My > kernel is 750K. Are you sure you have your numbers right? Hmm.. Let's see $ size /unix # SCO System V/386 3.2.0, no networking, no STREAMS 426840 + 70352 + 285236 = 782428 $ ls -l /unix ----r--r-- 1 bin mem 657193 Mar 19 18:08 /unix Ok. Now, let's check a Xenix kernel out. # mount /dev/fd096 /mnt -r # SCO Xenix 386AT 2.3.2 N1 boot disc $ size /mnt/xenix 240256 + 39492 + 82860 = 362608 = 0x58870 $ ls -l /mnt/xenix -rw-r--r-- 1 sys sys 311531 Jan 12 1989 /mnt/xenix Remember, this kernel is quite happy to execute System V/386 3.2 COFF binaries, including ones linked against the shared libc. And the default configuration is for a small, non-networked machine, but it's fine for a couple of people to do real work with. OK, let's get the figures for a Xenix kernel with TCP/IP and configured up with tape drivers and enough kernel resources to cope with 15-20 users, etc etc. ls -l gives 860169 # yeah, I'm typing this in from a size gives 390424 + 423636 + 0 = 814060 # cu(1) session in the other window. I don't have the figures for an equivalently configured System V/386 kernel. Someone else ? -- Ronald Khoo <ronald@robobar.co.uk> +44 81 991 1142 (O) +44 71 229 7741 (H)
bill@bilver.uucp (Bill Vermillion) (04/25/91)
In article <545@jahangir.UUCP> marc@jahangir.UUCP (Marc Rossner) writes: >In article <204@mnopltd.UUCP>, neal@mnopltd.UUCP writes: >> I noticed that the Xenix 2.3.2 kernel is about 1000K. The SCO Unix >> kernel is about 2000K. ^^^^^^^^ >I thought that the advantage of Xenix was that it was supposed to be tiny. ^^^^^ SCO manufactures BOTH Xenix and Unix. Xenix is small in comparison to Unix. -- Bill Vermillion - UUCP: uunet!tarpit!bilver!bill : bill@bilver.UUCP
shwake@raysnec.UUCP (Ray Shwake) (04/25/91)
rbraun@spdcc.COM (Rich Braun) writes: >Reading this thread, I can't help but to wonder: why worry about >kernel size, these days? I've long been one to complain about the fact >that software seems to get larger in direct proportion to the decrease >in memory costs, and often slower due to its increasing complexity, but >in the case of a reasonably well-performing O/S with lots of features, >why worry so much about kernel size? Recent history proves this assumed correlation false. The radical increase in memory demands (roughly 1987-90) resulted from the move to RISC, windowing environments (MS and X Windows), heavy networking, etc. That same period also witnessed *increased* memory prices partly resulting from the move to 1 MB chips (and general shortages of same) and the foolish memory floor price agreement with Japan. >Add another megabyte to the system and the problem will go away. Seems >a fairly simple and economical solution. Even at 1-2Mb, kernels remain >significantly smaller than most applications. (As compared to ten >years ago on mainframe computers, when a kernel was typically many >times larger than an application.) This line of thought assumes, at minimum, that the system architecture supports minimal, and inexpensive, memory increases. That's not always the case. Examples: Try to take a NEC Powermate from 2 MB to 3, 4 MB to 6, or 8 MB to 12. Can't do it. Another example. To increase a NeXT workstation from 8 MB, one must take it all the way to *20* MB (16 if one gets rid of surplus 1 MB SIMMS). Inexpensive it won't be. In summary? End the bloat. Bring on the UN!X Lite. ----------- uunet!media!ka3ovk!raysnec!shwake shwake@rsxtech
peter@ficc.ferranti.com (Peter da Silva) (04/26/91)
In article <7395@spdcc.SPDCC.COM> rbraun@spdcc.COM (Rich Braun) writes: > Reading this thread, I can't help but to wonder: why worry about > kernel size, these days? Memory is cheap. But it isn't free. > Add another megabyte to the system and the problem will go away. You want me to stop complaining, you buy me new motherboards for all the old Compaq 386es with 1MB on the motherboard, 4MB max, and anything over 2MB real expensive that we're stuck with. > Seems > a fairly simple and economical solution. Even at 1-2Mb, kernels remain > significantly smaller than most applications. What are you RUNNING? The biggest things I run are O(100K). -- Peter da Silva. `-_-' peter@ferranti.com +1 713 274 5180. 'U` "Have you hugged your wolf today?"
sef@kithrup.COM (Sean Eric Fagan) (04/29/91)
In article <1991Apr25.153709.552@bilver.uucp> bill@bilver.uucp (Bill Vermillion) writes: >In article <545@jahangir.UUCP> marc@jahangir.UUCP (Marc Rossner) writes: >>In article <204@mnopltd.UUCP>, neal@mnopltd.UUCP writes: >>> I noticed that the Xenix 2.3.2 kernel is about 1000K. The SCO Unix >>> kernel is about 2000K. ^^^^^^^^ >>I thought that the advantage of Xenix was that it was supposed to be tiny. > ^^^^^ >SCO manufactures BOTH Xenix and Unix. Xenix is small in comparison to >Unix. kithrup 1> size /unix 591700 + 79492 + 465460 = 1136652 I wish kithrup still had it's xenix kernel, so you could see the difference. What was it, you ask? Well, kithrup's xenix kernel was about 500k total, and that included STREAMS and whatnot (my ex-housemate ran X on it, so we bumped up some limits, and put in a few extra device drivers). kithrup is running with networking, STREAMS, and whatnot. (No NFS, though.) I have seen SCO UNIX kernels approaching 4Mbytes (lots and lots and lots of drivers, mostly); I have also seen xenix kernels approaching 2Mbytes. The main difference, at that point, is that there *is* so much more for unix: filesystems (cd-rom, dos, nfs, rfs), more device drivers out there, etc. -- Sean Eric Fagan | "I made the universe, but please don't blame me for it; sef@kithrup.COM | I had a bellyache at the time." -----------------+ -- The Turtle (Stephen King, _It_) Any opinions expressed are my own, and generally unpopular with others.
neal@mnopltd.UUCP (04/29/91)
->
->In article <204@mnopltd.UUCP>, neal@mnopltd.UUCP writes:
->> The current (maybe hopefully past) ankle-biting parade over SCO is of no
->> interest to me. SCO is as great as the world deserves. Now, on to
->> business.
->>
->> I noticed that the Xenix 2.3.2 kernel is about 1000K. The SCO Unix
->> kernel is about 2000K.
->
->I thought that the advantage of Xenix was that it was supposed to be tiny.
->I also thought that I had a rather massive kernel on my ISC 2.2. My
->kernel is 750K. Are you sure you have your numbers right?
->
->Marc Rossner
->jahangir!marc@uunet.uu.net
->
I quoteth /usr/adm/messages:
L4L5L6MNOPmem: total = 3712k, reserved = 4k, kernel = 1148k, user = 2560k
^^^^^
This includes: 2 Computone Boards
------------------------------------------------------------------------------
Neal Rhodes MNOP Ltd (404)- 972-5430
President Lilburn (atlanta) GA 30247 Fax: 978-4741
emory!mnopltd!neal
------------------------------------------------------------------------------neal@mnopltd.UUCP (04/29/91)
->marc@jahangir.UUCP (Marc Rossner) writes:
->>I thought that the advantage of Xenix was that it was supposed to be tiny.
->>I also thought that I had a rather massive kernel on my ISC 2.2. My
->>kernel is 750K. Are you sure you have your numbers right?
->
->Reading this thread, I can't help but to wonder: why worry about
->kernel size, these days? I've long been one to complain about the fact
->that software seems to get larger in direct proportion to the decrease
->in memory costs, and often slower due to its increasing complexity, but
->in the case of a reasonably well-performing O/S with lots of features,
->why worry so much about kernel size?
->
->Add another megabyte to the system and the problem will go away. Seems
->a fairly simple and economical solution. Even at 1-2Mb, kernels remain
->significantly smaller than most applications. (As compared to ten
->years ago on mainframe computers, when a kernel was typically many
->times larger than an application.)
->
->-rich
->
Why worry indeed: because on this box (NCR 3445) another MB costs $4000!
Unfortunate result of building a box with 16mb memory boards with
surface mount chips. We are sitting at 16mb now and would rather make it
work if possible.
BUT, if it were not for that brick wall I would tend to agree with you.
------------------------------------------------------------------------------
Neal Rhodes MNOP Ltd (404)- 972-5430
President Lilburn (atlanta) GA 30247 Fax: 978-4741
emory!mnopltd!neal
------------------------------------------------------------------------------davidsen@sixhub.UUCP (Wm E. Davidsen Jr) (05/01/91)
In article <TIZAOGB@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: | > Seems | > a fairly simple and economical solution. Even at 1-2Mb, kernels remain | > significantly smaller than most applications. | | What are you RUNNING? The biggest things I run are O(100K). You can run GNUemacs under MOTIF without paging in 32MB per user... Perhaps it's stuff like that. Actually I bet the compiler takes more than your 100k, but I'd bet I can run hours at a time without getting a process over a single MB. I'm unhappy with application bloat, too. MicroEmacs <120k even with some extensions and enhancements (and three sets of bug fixes). -- bill davidsen - davidsen@sixhub.uucp (uunet!crdgw1!sixhub!davidsen) sysop *IX BBS and Public Access UNIX moderator of comp.binaries.ibm.pc and 80386 mailing list "Stupidity, like virtue, is its own reward" -me
peter@ficc.ferranti.com (Peter da Silva) (05/02/91)
In article <3822@sixhub.UUCP> davidsen@sixhub.UUCP (bill davidsen) writes: > In article <TIZAOGB@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: > | > Seems > | > a fairly simple and economical solution. Even at 1-2Mb, kernels remain > | > significantly smaller than most applications. > | What are you RUNNING? The biggest things I run are O(100K). > You can run GNUemacs under MOTIF without paging in 32MB per user... > Perhaps it's stuff like that. Actually I bet the compiler takes more > than your 100k, but I'd bet I can run hours at a time without getting a > process over a single MB. -rwx--x--x 1 bin bin 21064 Sep 1 1987 /bin/cc -rwx--x--x 1 bin bin 31785 Sep 1 1987 /lib/p0 -rwx--x--x 1 bin bin 67345 Sep 1 1987 /lib/p1 -rwx--x--x 1 bin bin 109703 Sep 1 1987 /lib/p2 -rwx--x--x 1 bin bin 65711 Sep 1 1987 /lib/p3 /bin/cc: 15056 + 5818 + 2402 = 23276 = 0x5aec /lib/p0: 25184 + 6410 + 9606 = 41200 = 0xa0f0 /lib/p1: 51864 + 15290 + 8240 = 75394 = 0x12682 /lib/p2: 68657 + 40792 + 4978 = 114427 = 0x1befb /lib/p3: 54872 + 10648 + 6040 = 71560 = 0x11788 but then... -rw------- 1 sysinfo bin 239874 Apr 15 20:33 /xenix /xenix: 192411 + 19072 + 45786 = 257269 = 0x3ecf5 And that's with OpenNET support. -- Peter da Silva. `-_-' peter@ferranti.com +1 713 274 5180. 'U` "Have you hugged your wolf today?"