From: n...@arkansas.net (Network Operations Center)
Subject: Linux vs. SCO - SCO losing
Date: 1997/05/16
Message-ID: <337be7f8.4877975@news.anc.net>#1/1
X-Deja-AN: 242690979
Organization: A poorly-installed InterNetNews site
Newsgroups: comp.unix.sco.misc
You know, when I first chose SCO as my operating system I was really
happy with its stability but it really *really* lacked in the speed
department. I was running a USENET news server on a Pentium Pro 200
with 128M of RAM and SCO's virtual disk manager. The machine was
completely dragging its ass and after a while (couple of months) I
started having problems. I was indirectly informed that HPFS
(misnamed, if you ask me, as its not high-performance at all) didn't
scale well with partions above 2G in size. Well.... I switched over
to DTFS w/ compression turned off but the server continued to drag its
ass.
Well, someone told me that Linux's ext2fs (of all things - can you
imagine, a freeware UNIX system in a commercial environment) was
fairly fast. I figured I had nothing to lose so I switched. Well, my
server is *hauling ass* now -- the load average which was hovering
around 2.0 in SCO with the processor being tied up 100% of the time
(15% sys, 5% usr, 80% wio!) is now happily humming around .02 with
the CPU only being tied up between 1-2% of the time.
Can someone please tell me why a commercial UNIX OS that I spent
over $1000 for runs like shit compared to a FREEWARE operating
system?! I really want to know? With Linux's iBCS module, I can
even run SCO binaries. Why would anyone pay for SCO?
Anyhow, I've ranted and raved enough for one day. I'm finally happy
with my USENET news server. I must admit my SCO-based web and
mail server are running very nicely (no crashes with an extremely
long uptime) but I bet if I switched those over to Linux they'd be
even faster!
From: "Mark A. Davis" <ma...@XXXlaketaylor.org>
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/16
Message-ID: <337C5189.41C6@XXXlaketaylor.org>#1/1
X-Deja-AN: 242705897
References: <337be7f8.4877975@news.anc.net>
Organization: Lake Taylor Hospital
Newsgroups: comp.unix.sco.misc
Network Operations Center wrote:
>
> You know, when I first chose SCO as my operating system I was really
> happy with its stability but it really *really* lacked in the speed
> department. I was running a USENET news server on a Pentium Pro 200
> with 128M of RAM and SCO's virtual disk manager. The machine was
> completely dragging its ass and after a while (couple of months) I
> started having problems. I was indirectly informed that HPFS
> (misnamed, if you ask me, as its not high-performance at all) didn't
> scale well with partions above 2G in size. Well.... I switched over
> to DTFS w/ compression turned off but the server continued to drag its
> ass.
OK, I will speculate. Firstly- you say you were are using the virtual
software RAID. This will eat up a fair amount of time when doing disk
I/O. I would NEVER recommend someone with lots of heavy machine load to
use virtual disk manager- you are asking the CPU/OS to do all the work
of what really should be left upto hardware (a RAID controller). Did
you try any tests with this off?
> Well, someone told me that Linux's ext2fs (of all things - can you
> imagine, a freeware UNIX system in a commercial environment) was
> fairly fast. I figured I had nothing to lose so I switched. Well, my
> server is *hauling ass* now -- the load average which was hovering
> around 2.0 in SCO with the processor being tied up 100% of the time
> (15% sys, 5% usr, 80% wio!) is now happily humming around .02 with
> the CPU only being tied up between 1-2% of the time.
SO, I suppose you were NOT using virtual RAID on the Linux machine,
since that is not available. Also, ext2fs *WILL* be faster than HPFS on
SCO, but keep in mind that HPFS is a journaling robust filesystem, and
ext2fs is not.
> Can someone please tell me why a commercial UNIX OS that I spent
> over $1000 for runs like shit compared to a FREEWARE operating
> system?! I really want to know?
In my mind, your statements are totally unsubstantiated. Why don't you
tell us your COMPLETE hardware configuration as a start. Then, do you
have any benchmarks or tests to provide any objective data??
> With Linux's iBCS module, I can
> even run SCO binaries.
You can run MANY SCO Binaries. But not ones requiring any shared
libraries, unless you want to also buy SCO Unix.
> Why would anyone pay for SCO?
Duh- I think you better recheck the value of SCO Unix in regards to
centralized control, tighter standards, Motif, Merge, documentation,
journaling file systems, driver support, and the logic and flow of the
layout of the OS. Plus, being commercial had other advantages, not the
least of which is firm support from commercial software producers (end
applications).
> Anyhow, I've ranted and raved enough for one day. I'm finally happy
> with my USENET news server. I must admit my SCO-based web and
> mail server are running very nicely (no crashes with an extremely
> long uptime) but I bet if I switched those over to Linux they'd be
> even faster!
There is more in the world than just fast. Yes, I believe Linux to be
faster for most things, but certainly nowhere NEAR the difference you
seem to be implying. And yes, I use BOTH.
--
/--------------------------------------------------------------------\
| Mark A. Davis, |Lake Taylor| Voice: (757)-461-5001x431 8-4:30ET |
| Director of | Hospital | ma...@XXXtaylor.infi.net to reply |
|Information Systems|Norfolk, VA| from USENET remove anti-spam "XXX" |
\--------------------------------------------------------------------/
From: je...@comix.santa-cruz.ca.us (Jeff Liebermann)
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/16
Message-ID: <5ljjfe$e4v@comix.comix.santa-cruz.ca.us>#1/1
X-Deja-AN: 242050534
References: <337be7f8.4877975@news.anc.net> <337C5189.41C6@XXXlaketaylor.org>
Organization: COmmittee to Maintain an Independent Xenix
Newsgroups: comp.unix.sco.misc
Mark A. Davis (ma...@XXXlaketaylor.org) wrote:
: OK, I will speculate. Firstly- you say you were are using the virtual
: software RAID. This will eat up a fair amount of time when doing disk
: I/O. I would NEVER recommend someone with lots of heavy machine load to
: use virtual disk manager- you are asking the CPU/OS to do all the work
Probably true. However SCO delivers a better way to screw up
disk performance. The hard disk cache is rediculously small.
In 3.2v4.2, NBUF was limited to about 600KB. In OSR5,
the disk cache is "dynamic" but only allocated 1.4MB of my
16MB ram. See /usr/adm/messages "i/o bufs = XXXX" message for
your NBUF. The algorithm is apparently the same in both
versions at 10% of physical memory. This is far too small.
I usually allocate about 1/3 of physical memory for disk buffering.
If I can afford the ram (for systems over 64MB ram), I may give
NBUF (and NHBUF) half the ram. The preformance increase is
spectacular.
News servers are a different animal. Unless you have a zillion
nntp feeds or nntp readers, disk caching does not yield as much
an improvement as an increase in the number of spindles. This
is conventional wisdom extracted from reading the news.software.nntp
newsgroup. The idea is to reduce the number and range of seeks.
Software RAID is a loser. Striping is a big win.
Software for news servers also have problems. INN 1.4 has a
memory leak that would gobble ram and not give it back. This
was apparently solved in 1.5.1. I'm still running 1.4 (I'm lazy)
and need to kill and restart INN about once a week.
Linux has a very efficient way of dealing with memory. The
various buffer pools are shared in common and both grow and
shrink as required. OSR5's are mostly fixed, can grow, but do not
shrink (my observation). The result is that FOR A GIVEN AMOUNT
OF RAM, Linux is much more efficient. My guess is that you need
to add about 16MB ($80) more RAM to an OSR5 server to get equal
performance with a Linux 2.0.30 box.
: In my mind, your statements are totally unsubstantiated. Why don't you
: tell us your COMPLETE hardware configuration as a start. Then, do you
: have any benchmarks or tests to provide any objective data??
I agree. The entire posting is subjective and lacks sufficient
information to offer improvements and/or explanations. The
supplied load average only indicates CPU activity. Yet, the
complaint was about disk performance. How about some disk benchmarks?
It would be interesting to know what various outputs of "sar" show.
Perhaps the results from SarCheck or SCO Doctor.
[x] Email to author [ ] To mailing list [x] Posted to newsgroup
--
# Jeff Liebermann Liebermann Design 150 Felker St #D Santa Cruz CA 95060
# 408.336.2558 voice wb6ssy@ki6eh.#cca.ca.usa.noam wb6ssy.ampr.org 44.4.18.10
# 408.699.0483 digital_pager 73557,2074 cis [don't]
# je...@comix.santa-cruz.ca.us http://www.cruzio.com/~jeffl
From: n...@arkansas.net (Network Operations Center)
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/17
Message-ID: <337df679.45122758@news.anc.net>#1/1
X-Deja-AN: 242125418
References: <337be7f8.4877975@news.anc.net>
Organization: A poorly-installed InterNetNews site
Newsgroups: comp.unix.sco.misc
On Fri, 16 May 1997 04:59:05 GMT, n...@arkansas.net (Network Operations
Center) wrote:
[original message snipped - summary: Linux is faster than SCO as a
USENET news server]
: In my mind, your statements are totally unsubstantiated. Why don't you
: tell us your COMPLETE hardware configuration as a start. Then, do you
: have any benchmarks or tests to provide any objective data??
: : information to offer improvements and/or explanations. The
: : supplied load average only indicates CPU activity. Yet, the
: : complaint was about disk performance. How about some disk benchmarks?
: : It would be interesting to know what various outputs of "sar" show.
: :Perhaps the results from SarCheck or SCO Doctor.
My hardware configuration is/was as follows:
1) Pentium Pro 200
2) 128M RAM
3) Adaptect 2940UW controller (single controller)
4) SCO OS5 (part of Internet Fast Start)
5) 3 Quantum Fireball hard drives (4.3G in size)
6) HPFS for file system (although I switched to DTFS
after someone informed me of a scalability problem
with HPFS)
device %busy avque r+w/s blks/s avwait avserv
Sdsk-0 34.38 2.81 29.70 64.39 20.96 11.57
Sdsk-1 27.24 6.02 14.30 102.33 95.60 19.05
Sdsk-2 15.62 1.81 21.69 59.31 5.86 7.20
I wish I could give more current data - this is a snapshot of one
the average of one day's worth of performance from the SCO.
This is with the SCO VDM running a RAID0 (striping) array across
all three drives. The drives are Quantum Fireballs (fast SCSI-II).
This is on a particularly good day. However, as the news load
increased (ie: as data piled up on my news server) these numbers
would skyrocket. A typical sar -u would show something like:
00:00:01 %usr %sys %wio %idle
Average 3 5 85 7
This would get particularly high during the expiry process.
Also, netstat -m would show SEVERAL failures to allocate memory,
particularly for the class 8, class 9, class 10 range. I attempted to
increase the available amount of STREAMS memory, but that only
gimped the box up even further.
I also ran a variety of INN software. I started off with INN1.4sec
but was informed that 1.5.1 was better, faster and supported
streaming news feeds. When I switched to 1.5.1 it sped things
up at first, but it was a constant battle over time.
When I was running SCO, expire would take up to *9* hours
to complete when I had somewhere near 500,000 entries in
my history database. Under Linux, it currently takes 38 minutes
for the same amount of articles to be expired.
I would give you some similar disk statistics under Linux, but
I'm not sure how to extract the data. All I know is that the machine
responds faster, has more available memory, is stable (ie: has not
crashed with an unknown HPFS error). The load average is
extremely low, and it does not show the CPU being constantly
used.
*shrug* benchmarks or no benchmarks, the machine runs faster.
From: jo...@sco.COM (John DuBois)
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/19
Message-ID: <5lqckc$99a@hobbes.sco.com>#1/1
X-Deja-AN: 242459381
References: <337be7f8.4877975@news.anc.net>
Organization: The Santa Cruz Operation, Inc.
Newsgroups: comp.unix.sco.misc
In article <337be7f8...@news.anc.net>,
Network Operations Center <n...@arkansas.net> wrote:
+You know, when I first chose SCO as my operating system I was really
+happy with its stability but it really *really* lacked in the speed
+department. I was running a USENET news server on a Pentium Pro 200
+with 128M of RAM and SCO's virtual disk manager. The machine was
+completely dragging its ass and after a while (couple of months) I
+started having problems. I was indirectly informed that HPFS
+(misnamed, if you ask me, as its not high-performance at all) didn't
+scale well with partions above 2G in size. Well.... I switched over
+to DTFS w/ compression turned off but the server continued to drag its
+ass.
+
+Well, someone told me that Linux's ext2fs (of all things - can you
+imagine, a freeware UNIX system in a commercial environment) was
+fairly fast. I figured I had nothing to lose so I switched. Well, my
+server is *hauling ass* now -- the load average which was hovering
+around 2.0 in SCO with the processor being tied up 100% of the time
+(15% sys, 5% usr, 80% wio!) is now happily humming around .02 with
+the CPU only being tied up between 1-2% of the time.
Aside from what others have noted, I should point out that you are
misinterpreting one of the CPU usage statistics you quote. "15% sys, 5% usr,
80% wio" means that the CPU was being used 20% of the time. The 80% wio means
it was idle waiting for IO 80% of the time; if there had been anything needing
that CPU time it would have gotten it. This mistake is natural when taken with
the "load of 2"; the load factor reported by various utilities is incorrect.
My experience with HTFS is that the key to keeping it fast is to keep the
usage below 90%. Beyond that it really does slow down unacceptably. Slowing
down as usage approaches 100% is common with filesystems in general but much
more noticable with HTFS.
John
--
John DuBois jo...@sco.com KC6QKZ
From: jo...@sco.COM (John DuBois)
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/19
Message-ID: <5lqcgu$991@hobbes.sco.com>#1/1
X-Deja-AN: 243299981
References: <337be7f8.4877975@news.anc.net>
Organization: The Santa Cruz Operation, Inc.
Newsgroups: comp.unix.sco.misc
In article <337be7f8...@news.anc.net>,
Network Operations Center <n...@arkansas.net> wrote:
+You know, when I first chose SCO as my operating system I was really
+happy with its stability but it really *really* lacked in the speed
+department. I was running a USENET news server on a Pentium Pro 200
+with 128M of RAM and SCO's virtual disk manager. The machine was
+completely dragging its ass and after a while (couple of months) I
+started having problems. I was indirectly informed that HPFS
+(misnamed, if you ask me, as its not high-performance at all) didn't
+scale well with partions above 2G in size. Well.... I switched over
+to DTFS w/ compression turned off but the server continued to drag its
+ass.
+
+Well, someone told me that Linux's ext2fs (of all things - can you
+imagine, a freeware UNIX system in a commercial environment) was
+fairly fast. I figured I had nothing to lose so I switched. Well, my
+server is *hauling ass* now -- the load average which was hovering
+around 2.0 in SCO with the processor being tied up 100% of the time
+(15% sys, 5% usr, 80% wio!) is now happily humming around .02 with
+the CPU only being tied up between 1-2% of the time.
Aside from what others have noted, I should point out that you are
misinterpreting one of the CPU usage statistics you quote. "15% sys, 5% usr,
80% wio" means that the CPU was being used 20% of the time. The 80% wio means
it was idle waiting for IO 80% of the time; if there had been anything needing
that CPU it would have gotten it. This mistake is natural when taken with the
"load of 2"; the load factor reported by various utilities is incorrect.
My experience with HTFS is that the key to keeping it fast is to keep the
usage below 90%. Beyond that it really does slow down unacceptably. Slowing
down as usage approaches 100% is common with filesystems in general but much
more noticable with HTFS.
John
--
John DuBois jo...@sco.com KC6QKZ
From: n...@arkansas.net (Network Operations Center)
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/21
Message-ID: <33827d84.2291073@news.anc.net>
X-Deja-AN: 242753187
References: <337be7f8.4877975@news.anc.net>
Organization: Arkansas.Net/ANCI
Newsgroups: comp.unix.sco.misc
On Fri, 16 May 1997 04:59:05 GMT, n...@arkansas.net (Network Operations
Center) wrote:
[original message snipped - summary: SCO poor performer
as a USENET news server]
I've read over all of the replies to my original post complaining
about SCO as a USENET news server. If you don't include the
narrow-minded posts aimed at my typos and mangled acronyms I have yet
to hear anyone support the point that SCO makes a fast, robust news
server.
In a private message with one SCO user, he concurred that there were
some real problems with the HTFS file system. Additionally, serveral
posts supported this fact:
>My experience with HTFS is that the key to keeping it fast is to keep the
>usage below 90%. Beyond that it really does slow down unacceptably. Slowing
>down as usage approaches 100% is common with filesystems in general but much
>more noticable with HTFS.
Although I'd tend to disagree that the problem appears as late as 90%
of utilization. It occurrs earlier. Further, I corresponded with
Bela a long time before I got to the point of being upset looking for
*recommendations* on how to make my news server run better under SCO.
Some pertinent quotes:
From: Bela Lubkin <be...@sco.COM>
Date: Sun, 9 Feb 1997 04:24:53 -0800
>>The performance of
>> the machine is fine until it is time to expire news articles.
>> When this happens, my %wio skyrockets, slowing the machine down
>> and causing it to take nearly 6 hours to run expiration.
>
>This is probably due to a known issue with HTFS on news filesystems.
>Something about the pattern of filesystem activity in a news server
>causes slowdowns. A fix for this is under development but has no set
>delivery date.
[...]
>Changing the underlying RAID arrangement wouldn't help since it is a
>problem in the filesystem code.
>
>SCO's news server is currently running on a 3x4GB RAID arrangement
>(striped or concat, I'm not sure which), using DTFS with compression
>turned off (mount -o nocomp).
[..]
>Your current problem is much more due to filesystem behavior issues than
>RAID issues; don't worry about the RAID setup right now.
If those points aren't rock-solid enough, I don't know what else to
say? Bela and I corresponded for a few days and I took *ALL* of the
suggestions he gave (switch from concat array to RAID0 and finally
to DTFS w/ no compression) but the problems did not go away.
Anyhow, one user in the newsgroup posted:
>SO, I suppose you were NOT using virtual RAID on the Linux machine,
>since that is not available. Also, ext2fs *WILL* be faster than HPFS on
>SCO, but keep in mind that HPFS is a journaling robust filesystem, and
>ext2fs is not.
For one, software RAID0 is available on Linux using the 'md' driver
which is a standard part of the kernel. However, the person is right
- I am using a hardware RAID controller which is supported by Linux.
The fact is, HTFS is supposed to be the 'high-throughput' file system
and appears to be the fastest supported file system offered by SCO.
If not, what other choice is there? DTFS w/ compression turned off
suffered the same problems.
Another point people *tried* to make was that the VDM (software RAID)
was an inappropriate tool because I was expecting the hardware (CPU)
to do all the work. However, one person commented:
>I want to think about this one for a while and see what other
>people think. The SCO VDM on HTFS is slower than hardware RAID
>and/or Linux but not that much slower. Something else is going
>on and I sense (guess) that I'm missing something here.
Couple that with Bela's observation that the problem was with the
filesystem code and you can easily see its not the VDM that's causing
the problem. The fact is I was using a Pentium PRO 200 with 128M of
RAM. The machine should have had some spare CPU cycles for handling
the RAID tasks. A DPT controller uses a 68000 20Mhz processor for its
hardware RAID controller; I imagine that even while running INN, my
Pentium PRO 200 would have had a few spare CPU cycles to handle the
RAID tasks.
Further:
>> device %busy avque r+w/s blks/s avwait avserv
>> Sdsk-0 34.38 2.81 29.70 64.39 20.96 11.57
>> Sdsk-1 27.24 6.02 14.30 102.33 95.60 19.05
>> Sdsk-2 15.62 1.81 21.69 59.31 5.86 7.20
>
>Very bad. With 2-6 r/w instruction in the queue waiting to be
>chewed on by the hard disk adapter, this system will be a snail.
>I have never used the SCO Virtual Disk Manager. I only use DPT
>caching controllers with LOTS of memory. I had some bad experiences
>with Veritas RAID and decided that software RAID was a bad idea.
>
I guess the overall point I'm trying to make is that I *did* try to
solve the performance problems with SCO. When I finally got
frustrated and switched to Linux, the performance problems
disappeared. I used to only receive 50,000 articles a day with the
server running dog slow with SCO. With Linux, I'm getting nearly
200,000 articles a day and the overall system performance is tons
faster and can support a much higher simultaneous reader load.
Why doesn't someone running SCO as a USENET news server post their
experiences? I haven't seen anyone say "I'm running SCO as a USENET
news server taking in a full news feed without any problems". I
wonder why?
From: ke...@shady.com (Kevin Smith, ShadeTree Software, Inc.)
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/21
Message-ID: <5lv3l9$lip@shady.shady.com>#1/1
X-Deja-AN: 243021472
References: <337be7f8.4877975@news.anc.net> <33827d84.2291073@news.anc.net>
Organization: ShadeTree Software, Inc.
Newsgroups: comp.unix.sco.misc
In article <33827d84...@news.anc.net>
n...@arkansas.net (Network Operations Center) writes:
>On Fri, 16 May 1997 04:59:05 GMT, n...@arkansas.net (Network Operations
>Center) wrote:
>
>[original message snipped - summary: SCO poor performer
>as a USENET news server]
>
>I've read over all of the replies to my original post complaining
>about SCO as a USENET news server. If you don't include the
>narrow-minded posts aimed at my typos and mangled acronyms I have yet
>to hear anyone support the point that SCO makes a fast, robust news
>server.
>
>... [lots of real info snipped]...
>
>Why doesn't someone running SCO as a USENET news server post their
>experiences? I haven't seen anyone say "I'm running SCO as a USENET
>news server taking in a full news feed without any problems". I
>wonder why?
>
I'd really like to second that. Does anybody have a high volume SCO
news server running (50000+ articles). 300000 articles a day for a
full news feed with 3-4 million articles on-line is not uncommon.
What did you have to do to make it work?
I'm currently running a small news server with 1500 articles a day or so
and 35000 articles (120MB) on-line. Expire takes 5 minutes. I have no
clue how that would scale.
--
Cool Beans! Kevin Smith
- Chump ShadeTree Software, Inc.
Philadelpha, PA Voice: 001-215-487-3811
sy...@bbs.cpcn.com Email: ke...@shady.com (shady!kevin)
From: n...@arkansas.net (Network Operations Center)
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/23
Message-ID: <3385d345.1066429@news.anc.net>#1/1
X-Deja-AN: 243434530
References: <337be7f8.4877975@news.anc.net>
]<33827d84.2291073@news.anc.net> <5lv3l9$lip@shady.shady.com>
Organization: Arkansas.Net/ANCI
Newsgroups: comp.unix.sco.misc
On 21 May 1997 11:16:57 -0400, ke...@shady.com (Kevin Smith, ShadeTree
Software, Inc.) wrote:
>>Why doesn't someone running SCO as a USENET news server post their
>>experiences? I haven't seen anyone say "I'm running SCO as a USENET
>>news server taking in a full news feed without any problems". I
>>wonder why?
>>
>
>I'd really like to second that. Does anybody have a high volume SCO
>news server running (50000+ articles). 300000 articles a day for a
>full news feed with 3-4 million articles on-line is not uncommon.
>What did you have to do to make it work?
>
I think that the overwhelming silence is testimony enough. I'm
sticking with Linux -- 200,000 articles a day is enough proof for me.
From: "Dennis Taylor" <spammers...@ipipeline.net>
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/26
Message-ID: <01bc69ea$ba43cac0$0257c2cf@dennis>#1/1
X-Deja-AN: 244024274
References: <337be7f8.4877975@news.anc.net> <33827d84.2291073@news.anc.net>
<5lv3l9$lip@shady.shady.com> <3385d345.1066429@news.anc.net>
Organization: Not Really
Reply-To: "Dennis Taylor" <spammers...@ipipelinet.net>
Newsgroups: comp.unix.sco.misc
> >I'd really like to second that. Does anybody have a high volume SCO
> >news server running (50000+ articles). 300000 articles a day for a
> >full news feed with 3-4 million articles on-line is not uncommon.
> >What did you have to do to make it work?
> >
> I think that the overwhelming silence is testimony enough. I'm
> sticking with Linux -- 200,000 articles a day is enough proof for me.
>
Good grief. talk about self-serving, self-fulfilling-prophesy-type
questions!
Yeah, you're not going to get SCO boxes running news servers. The linux s/w
is there, it's free, and any company that's running a news server has the
technical staff necessary to support it. Anyone who did this using SCO
would have to be out of their mind, for purely financial reasons,
regardless of any technical questions.
Now ask how many people are running their corporate databases (Oracle or
otherwise) on linux. Ask how many people who *don't* have tech staff are
running linux. Ask how many people who run canned packages are running
linux. Actually, you probably still won't get much more response, because
most people running a business just don't have time for this kind of
useless conversation (I'm just in a bad mood today :-). From this you will
no doubt conclude with massive smugness that nobody's using SCO. (Yet
somehow they manage to maintain a corporate presence with cash flow in the
hundreds of millions per year).
I'm your worst nightmare in this type of conversation. At my job, we use
*both* SCO and linux. I have a budget to pay attention to, so the less I
can spend on SCO licenses, the better, I'm a techie, so I like fooling with
stuff like linux. I have one staff member dedicated almost completely to
linux research and support. I have the authority and the control to make
the decisions about what is used. Everything says I should be using linux
as much as possible. I am the perfect practical test of a situation where
linux has the potential to completely sweep SCO into the dustbin.
And the current count for our site is: Linux... 3 boxes, SCO.... 6 boxes.
Why? because the particular requirements for a particular function in most
cases mandate SCO. And because when you phone up companies like Oracle to
ask about linux support, their response is that they do not nor do they
have any plans to support linux now or in the future. And because the
documentation, when you can find it, SUCKS. I have a bookshelf full of SCO
docs. Which I use (I *hate* online docs!).
And because the same people who claim that we should be using linux for
absolutely everything, curl their lips with contempt when someone brings up
business requirements. That pretty much sets the tone for linux.
From: bi...@bilver.oau.org (Bill Vermillion)
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/27
Message-ID: <1997May27.234413.441@bilver.oau.org>#1/1
X-Deja-AN: 244350371
References: <337be7f8.4877975@news.anc.net> <5lv3l9$lip@shady.shady.com>
<3385d345.1066429@news.anc.net> <01bc69ea$ba43cac0$0257c2cf@dennis>
Organization: W. J. Vermillion - Orlando / Winter Park, FL
Newsgroups: comp.unix.sco.misc
In article <01bc69ea$ba43cac0$0257c2cf@dennis>,
Dennis Taylor <spammers...@ipipelinet.net> wrote:
>I'm your worst nightmare in this type of conversation. At my job, we use
>*both* SCO and linux. I have a budget to pay attention to, so the less I
>can spend on SCO licenses, the better, I'm a techie, so I like fooling with
>stuff like linux. I have one staff member dedicated almost completely to
>linux research and support. I have the authority and the control to make
>the decisions about what is used. Everything says I should be using linux
>as much as possible. I am the perfect practical test of a situation where
>linux has the potential to completely sweep SCO into the dustbin.
>And the current count for our site is: Linux... 3 boxes, SCO.... 6 boxes.
Just a question - as something appears to be unsaid here.
One person supports 3 Linux machine and you support 6 SCO?
That would seem to imply that the Linux machines take twice as
much maintenance as SCO.
Or perhaps you have more people on SCO. What is the support
ratio of machines to support personel?
You started a thought but seemed not to finish it.
W
--
Bill Vermillion - bill.ve...@oau.org | bi...@bilver.com
From: "Dennis Taylor" <spammers...@ipipeline.net>
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/28
Message-ID: <01bc6b76$36322460$0257c2cf@dennis>#1/1
X-Deja-AN: 244492664
References: <337be7f8.4877975@news.anc.net> <5lv3l9$lip@shady.shady.com>
<3385d345.1066429@news.anc.net> <01bc69ea$ba43cac0$0257c2cf@dennis>
<1997May27.234413.441@bilver.oau.org>
Organization: Not Really
Reply-To: "Dennis Taylor" <spammers...@ipipelinet.net>
Newsgroups: comp.unix.sco.misc
> Just a question - as something appears to be unsaid here.
>
> One person supports 3 Linux machine and you support 6 SCO?
> That would seem to imply that the Linux machines take twice as
> much maintenance as SCO.
>
> Or perhaps you have more people on SCO. What is the support
> ratio of machines to support personel?
>
> You started a thought but seemed not to finish it.
Actually, what I was trying to get across was that, even in a situation
where I theoretically *could* replace all SCO with Linux (in terms of me
having the authority to do so), I *haven't*. Since the next obvious retort
by the linux accolytes would be to question my knowledge, competence, and
personal grooming habits, I was also trying to make it clear that we *have*
linux therefore we know linux at least that well.
To answer your question directly, though, we don't really have enough staff
to say "You are a sysadmin" and "You are an operator" etc. Most of the
time, most of the machines run unattended, so it's hard to quantify how
many person-hours are allocated to sysadmin-ing linux vs sco. And since the
two groups do entirely different things, it would be meaningless anyway.
From: bi...@bilver.oau.org (Bill Vermillion)
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/29
Message-ID: <1997May29.044204.9593@bilver.oau.org>#1/1
X-Deja-AN: 244653515
References: <337be7f8.4877975@news.anc.net>
<01bc69ea$ba43cac0$0257c2cf@dennis> <1997May27.234413.441@bilver.oau.org>
<01bc6b76$36322460$0257c2cf@dennis>
Organization: W. J. Vermillion - Orlando / Winter Park, FL
Newsgroups: comp.unix.sco.misc
In article <01bc6b76$36322460$0257c2cf@dennis>,
Dennis Taylor <spammers...@ipipelinet.net> wrote:
>> Just a question - as something appears to be unsaid here.
>> One person supports 3 Linux machine and you support 6 SCO?
>> That would seem to imply that the Linux machines take twice as
>> much maintenance as SCO.
>> Or perhaps you have more people on SCO. What is the support
>> ratio of machines to support personel?
>> You started a thought but seemed not to finish it.
>Actually, what I was trying to get across was that, even in a situation
>where I theoretically *could* replace all SCO with Linux (in terms of me
>having the authority to do so), I *haven't*. Since the next obvious retort
>by the linux accolytes would be to question my knowledge, competence, and
>personal grooming habits, I was also trying to make it clear that we *have*
>linux therefore we know linux at least that well.
Well I was wondering about your grooming habits :-)^32 (that's
about 4 billion smileys).
>To answer your question directly, though, we don't really have enough staff
>to say "You are a sysadmin" and "You are an operator" etc. Most of the
>time, most of the machines run unattended, so it's hard to quantify how
>many person-hours are allocated to sysadmin-ing linux vs sco. And since the
>two groups do entirely different things, it would be meaningless anyway.
So now another question. Since the two groups do different
things - what are you running on each that exploits the
advantages. (I'm sure others have the same questions).
My Linux experience hasn't been that good - pretty buggy when I
tried it over a year ago. But the new Caldera is due here
soon.
You have to do what works best. I don't see a lot of serial
boards in the Linux world - but I do have one SCO system with
160 serial ports on it - scattered everywhere. The ap is
supported in SCO.
OTOH I do some work on SGI's as Web Servers and have been
experimenting with FreeBSD. I haven't timed it - but it
'feels' faster on a 120MHz Pentium than the 200MHz MIPS 4400
SGI Challenge S. (The latter is about $10K more)
We'll probably run some sites on FreeBSD and keep the SGI's for
running the secure servers.
I find my early Xenix experience that was system 7 based is
coming back as I work with the FreeBSD. I'm remebering to type
ps -lax instead of ps -ef. Depending on it's success we might
also go to BSDI for some catalog aps.
The nice thing about having so many different choices, is that
they all do things slightly differently and you can pick the
one that will work best for the job intended (as opposed to
some other OS types where there is just one vendor)
I'm a believer in 'whatever works best' and if they are
commercial aps, use the OS of choice by the SW vendor if
possible, and if not, at least one of the supported ones.
--
Bill Vermillion - bill.ve...@oau.org | bi...@bilver.com
From: "Dennis Taylor" <spammers...@ipipeline.net>
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/29
Message-ID: <01bc6c44$73c99d20$0257c2cf@dennis>#1/1
X-Deja-AN: 244739613
References: <337be7f8.4877975@news.anc.net>
<01bc69ea$ba43cac0$0257c2cf@dennis> <1997May27.234413.441@bilver.oau.org>
<01bc6b76$36322460$0257c2cf@dennis> <1997May29.044204.9593@bilver.oau.org>
Organization: Not Really
Reply-To: "Dennis Taylor" <spammers...@ipipelinet.net>
Newsgroups: comp.unix.sco.misc
> So now another question. Since the two groups do different
> things - what are you running on each that exploits the
> advantages. (I'm sure others have the same questions).
My 3 linux boxes are a dialup server, a web/mail server, and one that isn't
particularly doing anything. It was a samba server until I got that going
on SCO - the only reason I switched samba to SCO was administrative. Less
hassle configuring printers, and the samba exports can be backed up on the
same tape as the regular stuff.
My SCO boxes are: the corporate database (currently old cobol stuff, being
converted to oracle), an onsite backup system, an offsite backup system (we
are paranoid), two development systems, and my home system.
The advantages of the linux boxes is that they are free (linux, not the
h/w), the internet server s/w is free, there's lots of people using it for
that (critical mass, I guess) so getting answers is relatively easy. As far
as your comment about serial ports, we don't have anything like your
requirements (160 ports???!!?!?!), but for our range (up to 16) there's
lots of choice for hardware.
For the SCO boxes, we have a lot of purchased software and hardware that
AFAIK doesn't (or didn't when we bought it) run on linux. There's the cobol
compiler and runtime (Acucobol), a 3780 modem, Digi Portservers, FAXIMUM,
Uniplex, and of course Oracle.
Regarding the comment that you or someone else made (can't keep track)
which was something to the effect of "How long do you think Oracle will
keep ignoring the linux community?":
My answer would be: pretty much forever. For two reasons: The official one
is that the linux o/s isn't controlled, in that there's no guarantee that a
particular installation has any particular standard 'untweaked' o/s.
Support is hard enough with a rigidly controlled o/s like for instance SCO
- imagine trying to give phone support for something where some guy has
been fooling with the scsi drivers...
The second (unofficial) reason is the kicker, I think. It has to do with
the type of people who use linux. Generally speaking they can be described
as 1) techies, and 2) frugal. Otherwise why go with linux? There will be
specific situations like internet servers, but by and large it just doesn't
sound like a profitable market segment to be targeting.
Now I'm discussing Oracle, but the same logic could be applied to any
package with non-trivial pricing and support requirements.
>
> My Linux experience hasn't been that good - pretty buggy when I
> tried it over a year ago. But the new Caldera is due here
> soon.
Yeah, I've always found the linux installation on a new system takes about
an SCO-license-cost worth of my time.
From: Mike Jagdis <mi...@roan.co.uk>
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/30
Message-ID: <EAzp5J.CnE@roan.co.uk>#1/1
X-Deja-AN: 244939398
References: <337be7f8.4877975@news.anc.net>
<01bc6b76$36322460$0257c2cf@dennis> <1997May29.044204.9593@bilver.oau.org>
<01bc6c44$73c99d20$0257c2cf@dennis>
X-Mail2News-User: ne...@toaster.roan.co.uk
X-Comment-To: "Dennis Taylor" <spammers...@ipipelinet.net>
X-Mail2News-Path: punt-2.mail.demon.net!roan.demon.co.uk!toaster.roan.co.uk
Organization: Roan Technology Ltd.
Newsgroups: comp.unix.sco.misc
Dennis Taylor said
>For the SCO boxes, we have a lot of purchased software and hardware that
>AFAIK doesn't (or didn't when we bought it) run on linux. There's the cobol
>compiler and runtime (Acucobol), a 3780 modem, Digi Portservers, FAXIMUM,
>Uniplex, and of course Oracle.
Strangely enough most of that _does_ run on Linux even where there
aren't actually native versions :-).
>Regarding the comment that you or someone else made (can't keep track)
>which was something to the effect of "How long do you think Oracle will
>keep ignoring the linux community?":
>
>My answer would be: pretty much forever. For two reasons: The official one
>is that the linux o/s isn't controlled, in that there's no guarantee that a
>particular installation has any particular standard 'untweaked' o/s.
>Support is hard enough with a rigidly controlled o/s like for instance SCO
>- imagine trying to give phone support for something where some guy has
>been fooling with the scsi drivers...
SCO rigidly controlled? With Linux the patches tend to be pretty
much sequential so when someone phones all I need is "uname -r".
When someone phones with a SCO question I generally need to know
which combination of patches are loaded, which device driver is
in use and whether it is an official SCO version or a vendor
release. Personally I have no problem in supporting either of
them - although if someone wanted a full identify-and-fix contract
the cost of the SCO source license might be tricky :-).
>The second (unofficial) reason is the kicker, I think. It has to do with
>the type of people who use linux. Generally speaking they can be described
>as 1) techies, and 2) frugal.
On what do base that generalization? It's as misleading as declaring
that Windows is used to play games because the majority of Windows
shipments are on home PCs.
>Otherwise why go with linux?
Because people determined a business requirement and found that Linux
delivered an efective solution? There are many large companies
running Linux (other than as net servers and gateways). Why not ask
them? (Some of them actually *depend* on running mission critical
non-Linux-native programs on Linux. Their choice.)
>Now I'm discussing Oracle, but the same logic could be applied to any
>package with non-trivial pricing and support requirements.
It could. But your logic is based on flawed assumptions :-).
>Yeah, I've always found the linux installation on a new system takes about
>an SCO-license-cost worth of my time.
Takes me 5 or 10 minutes in either case - then just leave it
running. I wish I got paid as much as you :-).
Mike
--
.----------------------------------------------------------------------.
| Mike Jagdis | Internet: mailto:mi...@roan.co.uk |
| Roan Technology Ltd. | |
| 54A Peach Street, Wokingham | Telephone: +44 118 989 0403 |
From: "Dennis Taylor" <spammers...@ipipeline.net>
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/30
Message-ID: <01bc6d09$e8e3aea0$0257c2cf@dennis>#1/1
X-Deja-AN: 244979529
References: <337be7f8.4877975@news.anc.net>
<01bc6b76$36322460$0257c2cf@dennis> <1997May29.044204.9593@bilver.oau.org>
<01bc6c44$73c99d20$0257c2cf@dennis> <EAzp5J.CnE@roan.co.uk>
Organization: Not Really
Reply-To: "Dennis Taylor" <spammers...@ipipelinet.net>
Newsgroups: comp.unix.sco.misc
> >The second (unofficial) reason is the kicker, I think. It has to do with
> >the type of people who use linux. Generally speaking they can be
described
> >as 1) techies, and 2) frugal.
>
> On what do base that generalization? It's as misleading as declaring
> that Windows is used to play games because the majority of Windows
> shipments are on home PCs.
And yet whenever someone gets on a kick like the one that started this
thread, the usual reasons declared for preferring linux are "It's free!"
and "You get source code!" which appeal to 1) the frugal, and 2) techies.
>
> >Otherwise why go with linux?
>
> Because people determined a business requirement and found that Linux
> delivered an efective solution? There are many large companies
> running Linux (other than as net servers and gateways). Why not ask
> them? (Some of them actually *depend* on running mission critical
> non-Linux-native programs on Linux. Their choice.)
Well, it *would* be interesting to hear from some of these companies
(seriously. not sarcasm). Unfortunately, although every time I get into
this argument, I get hit with the statement that "lots of companies are
using linux for business apps....", no-one ever seems to come up with
concrete examples. I've just about relegated this statement to the same bin
as the one where creationists say "lots of evolutionists have converted to
creationism".
>
> >Now I'm discussing Oracle, but the same logic could be applied to any
> >package with non-trivial pricing and support requirements.
>
> It could. But your logic is based on flawed assumptions :-).
The statement is easy to make, but I see no support for it.
>
> >Yeah, I've always found the linux installation on a new system takes
about
> >an SCO-license-cost worth of my time.
>
> Takes me 5 or 10 minutes in either case - then just leave it
> running. I wish I got paid as much as you :-).
>
Another statement I keep running into. Easy to make, unprovable, and more
importantly unfalsifiable. I can only go by my own experiences with linux.
And to get back to another statement you made (I think it was you - might
have been another posting) to the effect that "strange - these packages all
run under linux - although the implementation is SCO" or something like
that.....
I don't doubt that a lot of SCO products run under linux - in fact I've
proven it to myself on several things. But the point that we keep dancing
around with these arguments is that if I buy (just to use a ferinstance)
Acucobol for SCO, I don't have to *prove* anything! I get a sco box, get
Acucobol, and it runs. If I want to try it on linux, I am A) taking a
chance, B) using up more of my time, possibly a *lot* more of my time. If I
run into problems, I phone tech support. If I am on linux, they say some
variation of "gee tough shit buddy".
I keep making this point and you keep missing it. MY SINGLE MOST VALUABLE
COMMODITY IS MY TIME. Unless I give up eating, sleeping, or shitting, I
have a limited supply of time. This is why we don't try to put oracle on
linux. This is why we only buy multiport boards for our linux machine that
are advertised to work on linux. For all your defense of linux, you have
not given me one single really good reason why I should stick my neck out.
"Just as good" is not good enough. As a hobbyist, I can screw around with
any new or obscure thing that tickles my fancy. As an I.S. Manager, I
*must* make conservative decisions, even when I find them to be personally
galling; otherwise I am not doing my job.
From: Mike Jagdis <mi...@roan.co.uk>
Subject: Re: Linux vs. SCO - SCO losing
Date: 1997/05/30
Message-ID: <EB057H.nAM@roan.co.uk>#1/1
X-Deja-AN: 245173459
References: <337be7f8.4877975@news.anc.net>
<01bc6c44$73c99d20$0257c2cf@dennis> <EAzp5J.CnE@roan.co.uk>
<01bc6d09$e8e3aea0$0257c2cf@dennis>
X-Mail2News-User: ne...@toaster.roan.co.uk
X-Comment-To: "Dennis Taylor" <spammers...@ipipelinet.net>
X-Mail2News-Path: punt-2.mail.demon.net!whthom.demon.co.uk!toaster.roan.co.uk
Organization: Roan Technology Ltd.
Newsgroups: comp.unix.sco.misc
Dennis Taylor said
>And yet whenever someone gets on a kick like the one that started this
>thread, the usual reasons declared for preferring linux are "It's free!"
>and "You get source code!" which appeal to 1) the frugal, and 2) techies.
What do you expect on Usenet? The OS groups are dominated by techies.
The free software groups are dominated by people that appreciate
free software. The sex and warez groups are, well... Why assume that
free OS groups are going to be a fair cross section of the community?
>Well, it *would* be interesting to hear from some of these companies
>(seriously. not sarcasm). Unfortunately, although every time I get into
>this argument, I get hit with the statement that "lots of companies are
>using linux for business apps....", no-one ever seems to come up with
>concrete examples. I've just about relegated this statement to the same bin
>as the one where creationists say "lots of evolutionists have converted to
>creationism".
Unfortunately I'm probably not at liberty to discuss most of those
that I know since without asking as they tend to consider their
Linux use as a competitive advantage (seriously!). However the
Linux Journal has run some articles about companies using Linux.
Two of the most public are Sixte in Germany and the Marie Curie
Cancer Research (Canada?).
>I don't doubt that a lot of SCO products run under linux - in fact I've
>proven it to myself on several things. But the point that we keep dancing
>around with these arguments is that if I buy (just to use a ferinstance)
>Acucobol for SCO, I don't have to *prove* anything! I get a sco box, get
>Acucobol, and it runs. If I want to try it on linux, I am A) taking a
>chance, B) using up more of my time, possibly a *lot* more of my time. If I
>run into problems, I phone tech support. If I am on linux, they say some
>variation of "gee tough shit buddy".
Dunno who you talk to then. If you want hot line support you can
buy some from me :-). If you have a problem running anything
non-Linux on Linux with iBCS let me know and I'll fix it when I
have time. If you need faster turnaround I'll happily quote for it.
>I keep making this point and you keep missing it. MY SINGLE MOST VALUABLE
>COMMODITY IS MY TIME. Unless I give up eating, sleeping, or shitting, I
>have a limited supply of time. This is why we don't try to put oracle on
>linux. This is why we only buy multiport boards for our linux machine that
>are advertised to work on linux. For all your defense of linux, you have
>not given me one single really good reason why I should stick my neck out.
I wasn't trying to give you reasons why you should use Linux. Just
pointing out that *your* reasons are not necessarily everyone's. Some
people think Linux is the bees knees. Some people think it's SCO. Some
even plump for Solaris or NT. Other people are needs driven and will
use whatever seems to best satisfy their needs.
Mike
--
.----------------------------------------------------------------------.
| Mike Jagdis | Internet: mailto:mi...@roan.co.uk |
| Roan Technology Ltd. | |
| 54A Peach Street, Wokingham | Telephone: +44 118 989 0403 |