S.Chang wrote:

> CJT wrote:
>
>> Keg wrote:
>>
>>> Bruce D Porter <nospam@ytc1.co.uk> wrote in message
>>> news:<pan.2004.07.09.16.11.30.821176@ytc1.co.uk>.. .
>>>
>>>> On Fri, 09 Jul 2004 08:11:53 -0700, Keg wrote:
>>>>
>>>>
>>>>> For starters I have been a Sun guy as far back as the old 100-lb
>>>>> Sparc IPX
>>>>
>>>>
>>>>
>>>> <snip>
>>>>
>>>> Troll !
>>>
>>>
>>>
>>>
>>> lmao. Anyone that has used, still currently uses, or wver will use the
>>> word troll is.. well ..hmm nuff said. layes noob. Tune in next week
>>> for the next set of buzzwords....
>>>
>>> welcome to my killfile.
>>
>>
>>
>> So how much does an IPX weigh again?
>>
> 4 or 5KG?
>
It was a rhetorical question. But since you responded seriously,
I weighed one -- 10.5 lb (4.8 Kg) including hard disk drive. So
your estimate was a good one.

--
The e-mail address in our reply-to line is reversed in an attempt to
minimize spam. Our true address is of the form che...@prodigy.net.

At Fri, 09 Jul 2004 15:56:57 GMT, Rich Teer <rich.teer@rite-group.com> writes:

> Please explain how the availability of commodity PC parts is anything
> to do with Linux. What is preventing you from running Solaris x86 on
> that same x86 hardware?

Well, to be fair, Solaris x86 doesn't support nearly the same range of
x86 hardware that GNU/Linux does. Quite possibly he's got hardware
that GNU/Linux runs on, but Solaris x86 doesn't.

> IME x86 is notoriously HARD to diagnose. Unless your idea of
> diagnosis is "just replace the box".

There is a reasonably large class of applications where "just replace
the box" is by far the simplest and cheapest approach to maintenance.
Google does this, for example. If your application is in this
category, it's hard to avoid the argument for commodity hardware.

> Linux is NOT the future

No, actually, the future of GNU/Linux looks pretty good. It may not
take over the world, but it will do quite well. At this point, I'm
afraid that Solaris is playing catchup by many measures.

Paul Eggert <eggert@twinsun.com> writes:

>No, actually, the future of GNU/Linux looks pretty good. It may not
>take over the world, but it will do quite well. At this point, I'm
>afraid that Solaris is playing catchup by many measures.

Which measures would those be?

(I find that Linux is still trying to play catchup by many measures
as well, so I wouldn't mind comparing notes)

Casper
--
Expressed in this posting are my opinions. They are in no way related
to opinions held by my employer, Sun Microsystems.
Statements on Sun products included here are not gospel and may
be fiction rather than truth.

At 10 Jul 2004 18:01:33 GMT, Casper H.S. Dik <Casper.Dik@Sun.COM> writes:

> Paul Eggert <eggert@twinsun.com> writes:
>
>>No, actually, the future of GNU/Linux looks pretty good. It may not
>>take over the world, but it will do quite well. At this point, I'm
>>afraid that Solaris is playing catchup by many measures.
>
> Which measures would those be?

Here's one example off the top of my head, since it's fresh in my mind.
Yesterday I happened to be talking to representatives of a systems
manufacturer with a worldwide presence whose name you'd recognize
instantly. They're not an OS vendor -- they don't have axes to grind
in the OS world. They've built a communications-oriented server with
lots of hardware support for various networking primitives. This box
is meant for ISPs. We're not talking a cheap box here -- we're
talking something more in the $100k range.

It would have been a good fit for Solaris, technically. And my
impression is that it was a greenfields project so they could have
chosen any OS to do it. They have a reasonable amount of Unix
experience inhouse. But they chose GNU/Linux. And they will get it
to work. Reliably. That's what they do.

The main problem here isn't technical: it's political. For a variety
of reasons, the GNU/Linux world is now larger than the Solaris world.
You can find more devices, more (and cheaper) experts and support, and
more applications. GNU/Linux has some technical advantages over
Solaris (it supports more hardware, it's a bit faster for some apps,
etc.) but that's not the main point here.

> (I find that Linux is still trying to play catchup by many measures
> as well, so I wouldn't mind comparing notes)

Yes, that's also quite true. For example, I far prefer the Solaris
kernel to the Linux kernel technically. And some of the new Solaris
stuff like dtrace is quite good: I wish GNU/Linux had it.

On Sat, 10 Jul 2004, Paul Eggert wrote:

> Well, to be fair, Solaris x86 doesn't support nearly the same range of
> x86 hardware that GNU/Linux does. Quite possibly he's got hardware
> that GNU/Linux runs on, but Solaris x86 doesn't.

Yes, I'll grant you that. In fact, off the top of my head, that's
the only advantage I'd concede easily to Linux.

> There is a reasonably large class of applications where "just replace
> the box" is by far the simplest and cheapest approach to maintenance.
> Google does this, for example. If your application is in this
> category, it's hard to avoid the argument for commodity hardware.

True, but replacing the box doesn't count as diagnosis, in my
opinion.

> No, actually, the future of GNU/Linux looks pretty good. It may not
> take over the world, but it will do quite well. At this point, I'm
> afraid that Solaris is playing catchup by many measures.

Perhaps, but Linux also has its fair share of catching up to do.

--
Rich Teer, SCNA, SCSA

President,
Rite Online Inc.

Voice: +1 (250) 979-1638
URL: http://www.rite-online.net

> + On 09-Jul-04 16:11:53
+Keg <rhugga@yahoo.com> wrote

[ Cute ]

>As an IT Director, I have no choice but to start moving towards Linux.
>Why?
>1) Support is orders of magnitude cheaper.

...And a magnitude less competent...

>2) Replacement parts are usually a 10 minute drive away.

Are you talking about systems without a servicecontract here ?

>3) Replacement parts are cheap enough to keep on hand. (hell keep
>entire spare systems on standby)

That goes for the smaller SUN systems too (V120/210/240 for example)
and the larger ones.. well.. large x86 system isnt cheap either,
especially if we talking many CPU's.. over 4 for example..

>4) You don't need another 100k of Veritas software & support.

Never used that on my sunmachines either, but we probably use them on
different applications.

But from curiosity, what do you use it for, and how do you solve
the same problem on linuxmachines free ?


>5) x86 based hardware is much easier and faster to troubleshoot and
>repair. (and thus employed skillsets are cheaper and easier to find)

STRONGLY disagree..

>You can't ignore that the "real-world" TCO is much much lower with
>Linux.

Disagree.

The reason that I (as a sysadmin) usally opt for SUN hardware is
that they are so much easier to manage, you can buy a servicecontract
that really is a servicecontract and not only words on a sheet
of paper, you can actually call or mail when you have problem and
someone try to solve your problem instead of trying to find other
things to blame.. and if it's a hardwareproblem you get things repaired
directly, not "Well we have to wait for this x part 3 weeks" or "just
send in your machine to us and we'll see what to do" or maybe they
send out a "tech" that knows far less than me and just stands there
and have no idea how to solve the problem, only trying stuff I
already tried..


We have x86/linux machines from dell and a company that build "noname"
systems, I think the support for both of the *SUCKS*, also there
only is support for the hardware, not the OS.. not that I call
suns support for solaris questions very often, but since the contract
covers both the hard and software I dont have to listen to shit
like "Its not our hardware, its a problem with the OS/software you
run on it"


>I know this message seems anti-Sun but it is not. I am devote Sun
>user. Moving to Linux for me would be like an NFL veteran saying
>goodbye to football after 15 years. I am very concerned and I don't
>want Sun to fall to the wayside as IBM becomes the predominant Linux
>player, which they ultimately will. It is likely that Linux will
>become IBM pending the outcome of this SCO thing.

I partly agree with you, however I think linux is 100% wrong way
for sun to go.. I think Solaris is Suns strongest card right now,
ESPECIALLY since it's the last real UNIX that is being actively
developed.


>Right now Sun's only answer to Linux is to install a pre-existing
>distribution onto a standard x86 hardware platform (which appears to
>be the Cobalt stuff they bought back when) Why would anyone spend the
>money for this hardware and support, when you can build a
>top-of-the-line dual-Xeon RAID 0+1 system for under $2k with no extra
>software (ie filesystem, software raid support/vol mgmt, system
>management) or support needed?

Well, Suns x86 machines are IMHO totally crap, I have NO idea why they
sell them and I only know ONE company that bought that crap, and their
reason is that they only want to have supportpeople from one company
running around.. (They also have a VERY large system consisting of
big sunmachines running databases and extremly much datastorage.)



>I justed replaced some very heavily hit E450 NFS servers with dual 1.7
>Ghz Xeon systems, 2GB RAM, and Adaptec SCSI cards for local system
>disk mirroring. I needed no additional support for this config and

In wich closet have you lived the last years ? for mirroing you dont
need to buy veritas SW.. sun have shipped a lightversion of that free
for many years now (at least for sol8/9) under different names.

>these systems can be maintained by any college intern. If I need a new
>CPU or something I send one of my IT guys down the street to Comp-USA
>or wherever.

...Just to find that that particular CPU isn't made anymore, and if you
buy another one the stepping is wrong so you have to replace both, and when
you do you realise that the BIOS software didnt support that CPU.. and
when you upgrade that you get strange problems with lockups since your
combination of cards isnt tested, and that fix in the new firmware at
the same time broke a interupthandler that makes the machine unusable..

Come on.. replacing parts is one of the nightmares when running x86
on production systems.

>Please Sun, get more proactive in the Linux space, it is the future,
>whether anyone likes it or not. You can't survive on Java alone.

Personally I think the linuxtrack will lead into suns death.

Supporting the existing linux developers with developer information and
maybe hardware is the way, and then concentrate on Solaris instead..

I say: Look at Commodore.. Look at SGI.. Look at all the things
they did wrong, dont do the same mistakes.

Glenn <glenn@IREPORTEVERYSPAMMER.canit.se> writes:

>I partly agree with you, however I think linux is 100% wrong way
>for sun to go.. I think Solaris is Suns strongest card right now,
>ESPECIALLY since it's the last real UNIX that is being actively
>developed.

Thanks; I think that both Linux and Solaris have their place in this
world; Sun's management is realizing more than ever how important
Solaris is for Sun "we're a Workstation, err, Server, err, perhaps
Solaris is an important part of our portfolio too".

Looking at Sun's history, it is clear that our roots are in the
desktop, the workstation. That, I think, is something that should
not have been neglected; how did Sun get into the data center,
through the desktop, how did MS Windows get there? Through the
desktop! How did Linux get there?

Solaris 10 has more new stuff and, more importantly, more
really innovative, exciting, new stuff than any Solaris release
before or even all of them combined, at least, that is how it
feels like. And we're much more serious about Solaris/x86 than ever
before; look at the Opteron hardware Sun is rolling out.

While much of the inovation is on server side and "big" computing,
the desktop is getting its share of attention; scroll mice
(about time, too!), integrated Xfree86 support.
(I can't be too enthused about GNOME, being a die-hard tvtwm user, but
I guess there are people who like that too)

>Well, Suns x86 machines are IMHO totally crap, I have NO idea why they
>sell them and I only know ONE company that bought that crap, and their
>reason is that they only want to have supportpeople from one company
>running around.. (They also have a VERY large system consisting of
>big sunmachines running databases and extremly much datastorage.)

Which ones in particular do you find crap, and if so why?
And, relative to Sun/SPARC gear or relative to other PCs?
I've never been too impressed when opening PC boxes and many but
not all of our SPARC boxes look very well done from the inside.

>In wich closet have you lived the last years ? for mirroing you dont
>need to buy veritas SW.. sun have shipped a lightversion of that free
>for many years now (at least for sol8/9) under different names.

I'm not sure when the last time was that DiskSuite was not
part of the OS bundle; I think it was part of the "server" kit
for somewhat longer but it was bundled with the OS on the same CD
starting with Solaris 8 (4 years ago) and integrated with the OS
(i.e., it's always installed) in Solaris 9 (2 years running now)

>..Just to find that that particular CPU isn't made anymore, and if you
>buy another one the stepping is wrong so you have to replace both, and when
>you do you realise that the BIOS software didnt support that CPU.. and
>when you upgrade that you get strange problems with lockups since your
>combination of cards isnt tested, and that fix in the new firmware at
>the same time broke a interupthandler that makes the machine unusable..

>Come on.. replacing parts is one of the nightmares when running x86
>on production systems.

That's why you just replace the whole system and hopes it isn't one
of the systems that just happens to run poorly under Linux or Solaris.

>Supporting the existing linux developers with developer information and
>maybe hardware is the way, and then concentrate on Solaris instead..

We're not planning to abandone Solaris anytime soon; we believe
strongly in it, but then, I am first and foremost a Solaris engineer.

We have put a lot of new stuff in Solaris 10 and you can testdrive
it, just look at:

http://www.sun.com/solarisexpress/

and some of our customers like some new bits a lot:

http://www.theregister.co.uk/2004/07...ace_user_take/


Casper

--
Expressed in this posting are my opinions. They are in no way related
to opinions held by my employer, Sun Microsystems.
Statements on Sun products included here are not gospel and may
be fiction rather than truth.

In article <44uve0h8u1n6f5fgdc4ukf3ue1a1a7i2sd@4ax.com>,
Josh Mckee <jtmckee@rmac.net> wrote:

> On 9 Jul 2004 08:11:53 -0700, rhugga@yahoo.com (Keg) wrote:
>
> >For starters I have been a Sun guy as far back as the old 100-lb Sparc
> >IPX systems and I still believe they make the best hardware on the
> >planet. However, I have been getting increasingly concerned with the
> >way they are operating lately.
> >
> >Example 1: We were told that some of the mid-range equipment we bought
> >3 years ago would "Meet our database needs for 5 years or more". Now
> >our 3-year support contracts have expired and support for some of our
> >machines nearly tripled and they are now telling us the equipment is
> >old and we need newer gear. So now I am forced to get 3rd party
> >support, which generally sucks.
> >
> >The Linux Issue:
> >
> >I think Sun is dropping the ball badly. The only thing holding Linux
> >back as a larger player in the data center is the lack of enterprise
> >quality x86 hardware and the scheduling issues with the pre-2.6 linux
> >kernels. Both of those issues are quickly changing and Linux, in my
> >opiniion, is gonna be a larger player in the data center over the next
> >5 years. (our data center is already 50% Linux, excluding the 2
> >250-node linux clusters we have)
> >
> >As an IT Director, I have no choice but to start moving towards Linux.
> >Why?
> >1) Support is orders of magnitude cheaper.
> >2) Replacement parts are usually a 10 minute drive away.
> >3) Replacement parts are cheap enough to keep on hand. (hell keep
> >entire spare systems on standby)
> >4) You don't need another 100k of Veritas software & support.
> >5) x86 based hardware is much easier and faster to troubleshoot and
> >repair. (and thus employed skillsets are cheaper and easier to find)
> >
> >You can't ignore that the "real-world" TCO is much much lower with
> >Linux.
> >
> >I know this message seems anti-Sun but it is not. I am devote Sun
> >user. Moving to Linux for me would be like an NFL veteran saying
> >goodbye to football after 15 years. I am very concerned and I don't
> >want Sun to fall to the wayside as IBM becomes the predominant Linux
> >player, which they ultimately will. It is likely that Linux will
> >become IBM pending the outcome of this SCO thing.
> >
> >Right now Sun's only answer to Linux is to install a pre-existing
> >distribution onto a standard x86 hardware platform (which appears to
> >be the Cobalt stuff they bought back when) Why would anyone spend the
> >money for this hardware and support, when you can build a
> >top-of-the-line dual-Xeon RAID 0+1 system for under $2k with no extra
> >software (ie filesystem, software raid support/vol mgmt, system
> >management) or support needed?
> >
> >I justed replaced some very heavily hit E450 NFS servers with dual 1.7
> >Ghz Xeon systems, 2GB RAM, and Adaptec SCSI cards for local system
> >disk mirroring. I needed no additional support for this config and
> >these systems can be maintained by any college intern. If I need a new
> >CPU or something I send one of my IT guys down the street to Comp-USA
> >or wherever.
> >
> >Please Sun, get more proactive in the Linux space, it is the future,
> >whether anyone likes it or not. You can't survive on Java alone.
> >
> >From a persoanl career standpoint I have to start focusing my energies
> >on Linux so that I can keep myself employed in the next 5-10 years. I
> >honestly think, unless things change drastically, that Sun's future is
> >very bleak. I mean, if Sun can't sell gear to the techies that love
> >the stuff, what can they do?
> >
> >I had budgeted for 3 4800's for Oracle database repacements for next
> >year, I am now testing viable Linux alternatives and so far it looks
> >promising. (and this is based on the 2.4 kernel with the poor
> >scheduler)
> >
> >Just curious how everyone else felt on the subject. Recent discussions
> >with some Sr. Sun engineers here in San Diego also opened my eyes a
> >bit. Even some of Sun's own people feel the same way.
> >
> >Just ranting I guess, I really like Sun's equipment but it is becoming
> >more and more difficult to put Sun gear into the Data Center,
> >especially with IT budgets today. To me, tripling support costs to
> >force customers into newer gear is un-ethical and I will take no part
> >in it.
>
> I faced a similar issue a few years ago. I had to replace a couple of
> E250 servers as their lease was expiring. Replacement options from Sun
> at the time were very expensive compared to IBM's Linux offerings. I
> ended up replacing the pair of E250's with a pair of IBM xseries 345
> servers running RedHat 7.3. The IBM systems were much better equipped
> and cost considerably less than the E250's. Today those systems have
> performed flawlessly...not a single problem in the couple of years
> that they've been in operation.
>
> I don't know if I would choose this same path today. Not that I am
> unhappy with the choice I made a couple of years ago. At the time
> Sun's low end models were lacking features, such as redundant power
> supplies, that I felt any good server should have. Sun has improved
> the competitiveness (as in capabilties and price) of their low end
> offerings since then making the decision closer than before.
>
> Josh

Yes, Josh. This sounds like a great fit. But you didn't by commodity
hardware that would be obsolete in 1 year. Bet you can still buy parts
and such for those systems (but not a the corner PC store).

But what do you do when these systems aren't big enough to handle the
load (either CPU, memory, or storage)? Can they be expanded by adding
more memory, CPUs, and storage? Will the I/O throughput scale with
expansion of storage and CPU or plateau out? If so, great.

The E250 is a small to mid-range system. Can this be done with systems
that are more in the E4500-E5500 range (I haven't seen the newer 4800 or
newer models)?

The Google/Hotmail many-small-systems model works great for a
distributed load. But what of a single system that's supposed to do one
thing like MRP or Engineering documentation or corporate email for 3000+
people. When there's only so much floor space in the datacenter and
only a limited number of sysadmins who can care, feed, and backup
systems, isn't fewer systems easier to manage than 100+ smaller systems?

--
DeeDee, don't press that button! DeeDee! NO! Dee...

In article <vilain-140D21.19243710072004@comcast.dca.giganews.com>,
"Michael Vilain <vilain@spamcop.net>" wrote:

> In article <44uve0h8u1n6f5fgdc4ukf3ue1a1a7i2sd@4ax.com>,
> Josh Mckee <jtmckee@rmac.net> wrote:
>
> > On 9 Jul 2004 08:11:53 -0700, rhugga@yahoo.com (Keg) wrote:
> >
> > >For starters I have been a Sun guy as far back as the old 100-lb Sparc
> > >IPX systems and I still believe they make the best hardware on the
> > >planet. However, I have been getting increasingly concerned with the
> > >way they are operating lately.
> > >
> > >Example 1: We were told that some of the mid-range equipment we bought
> > >3 years ago would "Meet our database needs for 5 years or more". Now
> > >our 3-year support contracts have expired and support for some of our
> > >machines nearly tripled and they are now telling us the equipment is
> > >old and we need newer gear. So now I am forced to get 3rd party
> > >support, which generally sucks.
> > >
> > >The Linux Issue:
> > >
> > >I think Sun is dropping the ball badly. The only thing holding Linux
> > >back as a larger player in the data center is the lack of enterprise
> > >quality x86 hardware and the scheduling issues with the pre-2.6 linux
> > >kernels. Both of those issues are quickly changing and Linux, in my
> > >opiniion, is gonna be a larger player in the data center over the next
> > >5 years. (our data center is already 50% Linux, excluding the 2
> > >250-node linux clusters we have)
> > >
> > >As an IT Director, I have no choice but to start moving towards Linux.
> > >Why?
> > >1) Support is orders of magnitude cheaper.
> > >2) Replacement parts are usually a 10 minute drive away.
> > >3) Replacement parts are cheap enough to keep on hand. (hell keep
> > >entire spare systems on standby)
> > >4) You don't need another 100k of Veritas software & support.
> > >5) x86 based hardware is much easier and faster to troubleshoot and
> > >repair. (and thus employed skillsets are cheaper and easier to find)
> > >
> > >You can't ignore that the "real-world" TCO is much much lower with
> > >Linux.
> > >
> > >I know this message seems anti-Sun but it is not. I am devote Sun
> > >user. Moving to Linux for me would be like an NFL veteran saying
> > >goodbye to football after 15 years. I am very concerned and I don't
> > >want Sun to fall to the wayside as IBM becomes the predominant Linux
> > >player, which they ultimately will. It is likely that Linux will
> > >become IBM pending the outcome of this SCO thing.
> > >
> > >Right now Sun's only answer to Linux is to install a pre-existing
> > >distribution onto a standard x86 hardware platform (which appears to
> > >be the Cobalt stuff they bought back when) Why would anyone spend the
> > >money for this hardware and support, when you can build a
> > >top-of-the-line dual-Xeon RAID 0+1 system for under $2k with no extra
> > >software (ie filesystem, software raid support/vol mgmt, system
> > >management) or support needed?
> > >
> > >I justed replaced some very heavily hit E450 NFS servers with dual 1.7
> > >Ghz Xeon systems, 2GB RAM, and Adaptec SCSI cards for local system
> > >disk mirroring. I needed no additional support for this config and
> > >these systems can be maintained by any college intern. If I need a new
> > >CPU or something I send one of my IT guys down the street to Comp-USA
> > >or wherever.
> > >
> > >Please Sun, get more proactive in the Linux space, it is the future,
> > >whether anyone likes it or not. You can't survive on Java alone.
> > >
> > >From a persoanl career standpoint I have to start focusing my energies
> > >on Linux so that I can keep myself employed in the next 5-10 years. I
> > >honestly think, unless things change drastically, that Sun's future is
> > >very bleak. I mean, if Sun can't sell gear to the techies that love
> > >the stuff, what can they do?
> > >
> > >I had budgeted for 3 4800's for Oracle database repacements for next
> > >year, I am now testing viable Linux alternatives and so far it looks
> > >promising. (and this is based on the 2.4 kernel with the poor
> > >scheduler)
> > >
> > >Just curious how everyone else felt on the subject. Recent discussions
> > >with some Sr. Sun engineers here in San Diego also opened my eyes a
> > >bit. Even some of Sun's own people feel the same way.
> > >
> > >Just ranting I guess, I really like Sun's equipment but it is becoming
> > >more and more difficult to put Sun gear into the Data Center,
> > >especially with IT budgets today. To me, tripling support costs to
> > >force customers into newer gear is un-ethical and I will take no part
> > >in it.
> >
> > I faced a similar issue a few years ago. I had to replace a couple of
> > E250 servers as their lease was expiring. Replacement options from Sun
> > at the time were very expensive compared to IBM's Linux offerings. I
> > ended up replacing the pair of E250's with a pair of IBM xseries 345
> > servers running RedHat 7.3. The IBM systems were much better equipped
> > and cost considerably less than the E250's. Today those systems have
> > performed flawlessly...not a single problem in the couple of years
> > that they've been in operation.
> >
> > I don't know if I would choose this same path today. Not that I am
> > unhappy with the choice I made a couple of years ago. At the time
> > Sun's low end models were lacking features, such as redundant power
> > supplies, that I felt any good server should have. Sun has improved
> > the competitiveness (as in capabilties and price) of their low end
> > offerings since then making the decision closer than before.
> >
> > Josh
>
> Yes, Josh. This sounds like a great fit. But you didn't by commodity
> hardware that would be obsolete in 1 year. Bet you can still buy parts
> and such for those systems (but not a the corner PC store).

That's correct. Even though I didn't buy commodity hardware I still
saved quite a bit of money, approximately $70K, by going with Linux on
x86.

> But what do you do when these systems aren't big enough to handle the
> load (either CPU, memory, or storage)?

They were much more capable than the "equivalent" Sun equipment.

> Can they be expanded by adding more memory, CPUs, and storage?

I'm not sure what you're asking about. These systems can be configured
with a lot more memory. But they can only be configured with dual
processors...just like the E250. The difference is that the CPU's were
2.0GHz instead of 400MHz like the E250's. I could probably upgrade them
to 3.4GHz models as IBM is still selling the xseries 345 with this speed
of CPU. Storage...these systems came very well equipped with 5 x 36GB
Ultra320 10K RPM HD's. More than sufficient for their intended purpose.
Generally speaking the IBM system significantly outperformed the closest
Sun system...at a significant cost savings.

> Will the I/O throughput scale with expansion of storage and CPU or plateau out?
>If so, great.

These systems are more than adequate for their intended use.

> The E250 is a small to mid-range system. Can this be done with systems
> that are more in the E4500-E5500 range (I haven't seen the newer 4800 or
> newer models)?

It could...at a more higher cost. The E250 was chosen because it was the
lowest cost Sun system (at the time) that had redundant power supplies.
The E250 was more than sufficient for the intended task...it just cost
considerably more than an "equivalent" IBM x86 system running Linux.

> The Google/Hotmail many-small-systems model works great for a
> distributed load. But what of a single system that's supposed to do one
> thing like MRP or Engineering documentation or corporate email for 3000+
> people. When there's only so much floor space in the datacenter and
> only a limited number of sysadmins who can care, feed, and backup
> systems, isn't fewer systems easier to manage than 100+ smaller systems?

The systems were sized according to their intended task. They're not MRP
or Engineering systems.

Sun systems are great for big systems. But for entry level systems they
are not cost competitive. You may want to Google on this subject to
learn more about what I was looking for out of these servers. Low end
was what I was looking for.

Josh

In article <jtmckee-F63EE7.20522310072004@netnews.comcast.net>,
Josh McKee <jtmckee@rmac.know-spam-bogus.net> wrote:

> In article <vilain-140D21.19243710072004@comcast.dca.giganews.com>,
> "Michael Vilain <vilain@spamcop.net>" wrote:
>
> > In article <44uve0h8u1n6f5fgdc4ukf3ue1a1a7i2sd@4ax.com>,
> > Josh Mckee <jtmckee@rmac.net> wrote:
> >
> > > On 9 Jul 2004 08:11:53 -0700, rhugga@yahoo.com (Keg) wrote:
> > >
> > > >For starters I have been a Sun guy as far back as the old 100-lb Sparc
> > > >IPX systems and I still believe they make the best hardware on the
> > > >planet. However, I have been getting increasingly concerned with the
> > > >way they are operating lately.
> > > >
> > > >Example 1: We were told that some of the mid-range equipment we bought
> > > >3 years ago would "Meet our database needs for 5 years or more". Now
> > > >our 3-year support contracts have expired and support for some of our
> > > >machines nearly tripled and they are now telling us the equipment is
> > > >old and we need newer gear. So now I am forced to get 3rd party
> > > >support, which generally sucks.
> > > >
> > > >The Linux Issue:
> > > >
> > > >I think Sun is dropping the ball badly. The only thing holding Linux
> > > >back as a larger player in the data center is the lack of enterprise
> > > >quality x86 hardware and the scheduling issues with the pre-2.6 linux
> > > >kernels. Both of those issues are quickly changing and Linux, in my
> > > >opiniion, is gonna be a larger player in the data center over the next
> > > >5 years. (our data center is already 50% Linux, excluding the 2
> > > >250-node linux clusters we have)
> > > >
> > > >As an IT Director, I have no choice but to start moving towards Linux.
> > > >Why?
> > > >1) Support is orders of magnitude cheaper.
> > > >2) Replacement parts are usually a 10 minute drive away.
> > > >3) Replacement parts are cheap enough to keep on hand. (hell keep
> > > >entire spare systems on standby)
> > > >4) You don't need another 100k of Veritas software & support.
> > > >5) x86 based hardware is much easier and faster to troubleshoot and
> > > >repair. (and thus employed skillsets are cheaper and easier to find)
> > > >
> > > >You can't ignore that the "real-world" TCO is much much lower with
> > > >Linux.
> > > >
> > > >I know this message seems anti-Sun but it is not. I am devote Sun
> > > >user. Moving to Linux for me would be like an NFL veteran saying
> > > >goodbye to football after 15 years. I am very concerned and I don't
> > > >want Sun to fall to the wayside as IBM becomes the predominant Linux
> > > >player, which they ultimately will. It is likely that Linux will
> > > >become IBM pending the outcome of this SCO thing.
> > > >
> > > >Right now Sun's only answer to Linux is to install a pre-existing
> > > >distribution onto a standard x86 hardware platform (which appears to
> > > >be the Cobalt stuff they bought back when) Why would anyone spend the
> > > >money for this hardware and support, when you can build a
> > > >top-of-the-line dual-Xeon RAID 0+1 system for under $2k with no extra
> > > >software (ie filesystem, software raid support/vol mgmt, system
> > > >management) or support needed?
> > > >
> > > >I justed replaced some very heavily hit E450 NFS servers with dual 1.7
> > > >Ghz Xeon systems, 2GB RAM, and Adaptec SCSI cards for local system
> > > >disk mirroring. I needed no additional support for this config and
> > > >these systems can be maintained by any college intern. If I need a new
> > > >CPU or something I send one of my IT guys down the street to Comp-USA
> > > >or wherever.
> > > >
> > > >Please Sun, get more proactive in the Linux space, it is the future,
> > > >whether anyone likes it or not. You can't survive on Java alone.
> > > >
> > > >From a persoanl career standpoint I have to start focusing my energies
> > > >on Linux so that I can keep myself employed in the next 5-10 years. I
> > > >honestly think, unless things change drastically, that Sun's future is
> > > >very bleak. I mean, if Sun can't sell gear to the techies that love
> > > >the stuff, what can they do?
> > > >
> > > >I had budgeted for 3 4800's for Oracle database repacements for next
> > > >year, I am now testing viable Linux alternatives and so far it looks
> > > >promising. (and this is based on the 2.4 kernel with the poor
> > > >scheduler)
> > > >
> > > >Just curious how everyone else felt on the subject. Recent discussions
> > > >with some Sr. Sun engineers here in San Diego also opened my eyes a
> > > >bit. Even some of Sun's own people feel the same way.
> > > >
> > > >Just ranting I guess, I really like Sun's equipment but it is becoming
> > > >more and more difficult to put Sun gear into the Data Center,
> > > >especially with IT budgets today. To me, tripling support costs to
> > > >force customers into newer gear is un-ethical and I will take no part
> > > >in it.
> > >
> > > I faced a similar issue a few years ago. I had to replace a couple of
> > > E250 servers as their lease was expiring. Replacement options from Sun
> > > at the time were very expensive compared to IBM's Linux offerings. I
> > > ended up replacing the pair of E250's with a pair of IBM xseries 345
> > > servers running RedHat 7.3. The IBM systems were much better equipped
> > > and cost considerably less than the E250's. Today those systems have
> > > performed flawlessly...not a single problem in the couple of years
> > > that they've been in operation.
> > >
> > > I don't know if I would choose this same path today. Not that I am
> > > unhappy with the choice I made a couple of years ago. At the time
> > > Sun's low end models were lacking features, such as redundant power
> > > supplies, that I felt any good server should have. Sun has improved
> > > the competitiveness (as in capabilties and price) of their low end
> > > offerings since then making the decision closer than before.
> > >
> > > Josh
> >
> > Yes, Josh. This sounds like a great fit. But you didn't by commodity
> > hardware that would be obsolete in 1 year. Bet you can still buy parts
> > and such for those systems (but not a the corner PC store).
>
> That's correct. Even though I didn't buy commodity hardware I still
> saved quite a bit of money, approximately $70K, by going with Linux on
> x86.
>
> > But what do you do when these systems aren't big enough to handle the
> > load (either CPU, memory, or storage)?
>
> They were much more capable than the "equivalent" Sun equipment.
>
> > Can they be expanded by adding more memory, CPUs, and storage?
>
> I'm not sure what you're asking about. These systems can be configured
> with a lot more memory. But they can only be configured with dual
> processors...just like the E250. The difference is that the CPU's were
> 2.0GHz instead of 400MHz like the E250's. I could probably upgrade them
> to 3.4GHz models as IBM is still selling the xseries 345 with this speed
> of CPU. Storage...these systems came very well equipped with 5 x 36GB
> Ultra320 10K RPM HD's. More than sufficient for their intended purpose.
> Generally speaking the IBM system significantly outperformed the closest
> Sun system...at a significant cost savings.
>
> > Will the I/O throughput scale with expansion of storage and CPU or plateau
> > out?
> >If so, great.
>
> These systems are more than adequate for their intended use.
>
> > The E250 is a small to mid-range system. Can this be done with systems
> > that are more in the E4500-E5500 range (I haven't seen the newer 4800 or
> > newer models)?
>
> It could...at a more higher cost. The E250 was chosen because it was the
> lowest cost Sun system (at the time) that had redundant power supplies.
> The E250 was more than sufficient for the intended task...it just cost
> considerably more than an "equivalent" IBM x86 system running Linux.
>
> > The Google/Hotmail many-small-systems model works great for a
> > distributed load. But what of a single system that's supposed to do one
> > thing like MRP or Engineering documentation or corporate email for 3000+
> > people. When there's only so much floor space in the datacenter and
> > only a limited number of sysadmins who can care, feed, and backup
> > systems, isn't fewer systems easier to manage than 100+ smaller systems?
>
> The systems were sized according to their intended task. They're not MRP
> or Engineering systems.
>
> Sun systems are great for big systems. But for entry level systems they
> are not cost competitive. You may want to Google on this subject to
> learn more about what I was looking for out of these servers. Low end
> was what I was looking for.
>
> Josh

No, this was exactly the point I wanted to clarify and make. Sun's
loosing ground in the low-range space or perhaps just offering something
there to have a wider range of systems.

I have a feeling some marketing person in SUN might think that this is a
good thing--overhead for maintaining parts and support, etc for E250 vs.
larger systems. Software support costs the same for both systems,
doesn't it. It's based on responce time, not CPU. It may be that the
money for SUN is in the larger systems, so they're focusing their
attentions there. Wouldn't surprise me.

Seems like everyone wins. You get a system that works for you on
hardware that does the job. Sun gets to focus on what makes them
money--big systems.

--
DeeDee, don't press that button! DeeDee! NO! Dee...