Implementing a RAID System

[kony]

Well I think he did elude to some problems:

"It should also speed up the entire server a little because the hard
drive doesn't write to multiple places at the same time so much."

That sounds theoretical to me. It may or may not pan out in
the application. The lan is often the bottleneck.

"it takes a good 8 hours to back up the system every night"

If the system isn't being used at that point, the cause
should be related to the tape. Certainly there is no 40GB
drive that is THIS slow, they're all capable of well over
20MB/s.

SQL & exchange can be pretty IO intensive. So can preparing thousands
of tiny files for backup. Putting the OS/Apps on a different physical
disk or array than the data is, as he thinks, a step in the right
direction. Some raid levels might help this also as well as faster
drives. When he starts talking about upgrades providing no real
expected performance benefit then we know all the bottlenecks are
gone.

That's one way to look at it. Another is that when we hear
the upgrade didn'tt provide any noticable difference, that
the bottleneck is either gone or not addressed. It could
help to put the apps and data on another physical disk, that
is a good idea. Not knowing the specs of the box nor the
load it could be as useful to upgrade the memory.

Changes aside, the primary issue was adding space so I'd not
jump to conclusions about someone
wanting/needing/being-willing-to-spend more for something
not expressly _needed_. It could be that in a few years of
company growth, the servers' load is heavier. Could be the
company shrinks and it gets lighter. Could be it's upgraded
to handle the future but then before that upgrade bears any
real fruit the whole box is replaced as a regular rotation
for reliability reasons... another server replaced when
convenient could be far less expensive than downtime.

 

[Curious George]

That sounds theoretical to me. It may or may not pan out in
the application.

It's not theoretical at all. It's simply what happens when there is
too much contention between OS, multiple applications, & multiple
applications' data stores all for the same physical drive.

The lan is often the bottleneck.

not in a small workgroup making primarily IO intensive requests i.e.
SQL, etc. For all we know he has a quad gigabit connection anyway.
Do you know the specs of a "supped up desktop" or his network
infrastructure? I don't.

If the system isn't being used at that point, the cause
should be related to the tape. Certainly there is no 40GB
drive that is THIS slow, they're all capable of well over
20MB/s.

No. If the system can't deliver a sufficient steady stream of data
the tape will have to backup up and do the resultant shoeshine dance.
A normal 2 hour stream can take ten times that time or more in a worst
case scenario. Forget about STR. It is unrealistic to expect
commodity non-raid drives to deliver anywhere near 20MB/s sustained
for hours on end. The smaller the files and the more fragmented the
file system or the smaller the caching space or poorer the
software/software config the worse the problem will be.

The tape works so it is not to blame. If the tape is shoeshining that
is a software configuration and/or disk subsystem problem

That's one way to look at it. Another is that when we hear
the upgrade didn'tt provide any noticable difference, that
the bottleneck is either gone or not addressed. It could
help to put the apps and data on another physical disk, that
is a good idea. Not knowing the specs of the box nor the
load it could be as useful to upgrade the memory.

Oh sure. We could talk about upgrading any or all of the components.
Bottom line is if there is too much contention for disk resources, as
seems to be inferred, that is a problem ram won't fix.

Changes aside, the primary issue was adding space so I'd not
jump to conclusions about someone
wanting/needing/being-willing-to-spend more for something
not expressly _needed_. It could be that in a few years of
company growth, the servers' load is heavier. Could be the
company shrinks and it gets lighter. Could be it's upgraded
to handle the future but then before that upgrade bears any
real fruit the whole box is replaced as a regular rotation
for reliability reasons... another server replaced when
convenient could be far less expensive than downtime.

Since he is already looking at upgrading the storage subsystem and is
inferring some possible performance issues with the current storage,
lets make sure he picks the right storage upgrade first, shall we?
Pure speculation without interest in understanding the OP's situation
neither helps him nor counters my suggestions/observations.

 

[kony]

It's not theoretical at all. It's simply what happens when there is
too much contention between OS, multiple applications, & multiple
applications' data stores all for the same physical drive.

Actually it is entirely theoretical. The theory would be
that the overhead from running the OS and apps, plus the
data stores, is significant. On any decent server, it
isn't. The reason is that such "decent" server has enough
memory to cache the apps and OS files it's running. It is
not generally going to be a scenario like a desktop system
where there is continual reading of HDD for OS or apps.

In other words, usually the data is the vast majority of the
IO. It may not make even the slightest difference in use.
It "could" matter but there is a difference between
theoretical speed and realized need.

not in a small workgroup making primarily IO intensive requests i.e.
SQL, etc. For all we know he has a quad gigabit connection anyway.
Do you know the specs of a "supped up desktop" or his network
infrastructure? I don't.

Hmm. It's running off a desktop-turned server with a single
40GB HDD in it. Sure, I could be wrong but I don't think it
has quad gigabit connections. Let's be realistic. The lan
_IS_ often the bottleneck. Always? No... but often. The
idea of throwing money at a problem is wasteful when there's
no reason to believe it will matter.

Note that earlier there was NO mention of performance
problems other than the length of time it takes to do the
tape backup.

No. If the system can't deliver a sufficient steady stream of data
the tape will have to backup up and do the resultant shoeshine dance.
A normal 2 hour stream can take ten times that time or more in a worst
case scenario. Forget about STR. It is unrealistic to expect
commodity non-raid drives to deliver anywhere near 20MB/s sustained
for hours on end.

Believe it or not, tape has been around a long time.
Somehow contrary to your theory, it works. Imagine that.

The smaller the files and the more fragmented the
file system or the smaller the caching space or poorer the
software/software config the worse the problem will be.

Sure, try to paint a worst-case scenario as an argument. If
it's all that bad, the last thing focused on should be RAID.
The system should be reconfigured to work properly before
any new additions and their related changes are made.

The tape works so it is not to blame. If the tape is shoeshining that
is a software configuration and/or disk subsystem problem

Could be. We have insufficient info.

Oh sure. We could talk about upgrading any or all of the components.
Bottom line is if there is too much contention for disk resources, as
seems to be inferred, that is a problem ram won't fix.

yes, _IF_.
Hundreds of dollars later the problem may still remain.

Since he is already looking at upgrading the storage subsystem and is
inferring some possible performance issues with the current storage,
lets make sure he picks the right storage upgrade first, shall we?
Pure speculation without interest in understanding the OP's situation
neither helps him nor counters my suggestions/observations.

So he should buy a new tape drive instead?

 

[Curious George]

Actually it is entirely theoretical. The theory would be
that the overhead from running the OS and apps, plus the
data stores, is significant. On any decent server, it
isn't. The reason is that such "decent" server has enough
memory to cache the apps and OS files it's running. It is
not generally going to be a scenario like a desktop system
where there is continual reading of HDD for OS or apps.

In other words, usually the data is the vast majority of the
IO. It may not make even the slightest difference in use.
It "could" matter but there is a difference between
theoretical speed and realized need.

No. He has noticed drive contention. That's why he expects
improvement by separating the data stores - so it's not theoretical.
(reread his choice of words.) Your explanation attempts to prioritize
the sources of contention but doesn't contradict Chris' ostensible
observation. We can't just make believe he didn't notice or elude to
noticing anything.

RAM won't deal with that drive contention and it can't cache
everything. Only Chris can tell us how much contention there is or
how "decent" his "server" is. From his comments, though, this appears
to not be the total non-issue you'd like to imagine.

Hmm. It's running off a desktop-turned server with a single
40GB HDD in it. Sure, I could be wrong but I don't think it
has quad gigabit connections. Let's be realistic. The lan
_IS_ often the bottleneck. Always? No... but often. The
idea of throwing money at a problem is wasteful when there's
no reason to believe it will matter.

Yes & no. I don't think that is a responsible generalization or guess
re the lan. Esp given the nature of SQL & exchange access/usage
patterns & what that does to a server. Hyperbole aside there's
certainly no reason why a "souped up desktop" couldn't have a single
gigabit connection or dual fast ethernet. We also don't know what's
going on on the lan like roaming profiles, Qos, etc. or even how these
employees work. Also many of your tenuous theoretical suggestions
haven't been nearly as fiscally responsible as you might imagine.
Your guesses re server behavior can't come from field experience.
(BTW contesting hyperbole makes for a comical argument.)

Note that earlier there was NO mention of performance
problems other than the length of time it takes to do the
tape backup.

No. You just missed it & refuse to acknowledge it when I showed it to
you.

Disk contention was THE reported performance issue. It may not
presently be of a crippling severity but was noticeable nonetheless.

We don't know what the cause of the tape backup issue is. It could be
disk or software or both. It could also very well be (if my math is
correct) a DLT4000 (autoloader?) running a complete backup every
night. In which case all we have to suggest is to run incremental or
differentials nightly & full backups less often. A DLT7000 or better
may also help (esp for recovery times) without breaking the bank. A
higher capacity tape backup may also be warranted to keep up with
expanding data (Chris - don't bother with the "value" line of dlt's).

Believe it or not, tape has been around a long time.
Somehow contrary to your theory, it works. Imagine that.

You haven't ever used industrial tape before, have you? Yes it works
when it is set up right. But most tapes can't slow up to compensate
for poor system thruput (which may or may not be purely hardware
related). (I only know of newer HP ltos which can do this
compensation). That's why I'm endorsing troubleshooting it or at
least finding out more about the setup. geez!

Sure, try to paint a worst-case scenario as an argument. If
it's all that bad, the last thing focused on should be RAID.
The system should be reconfigured to work properly before
any new additions and their related changes are made.

His backup basically _is_ a worst case scenario. 8 hours to backup or
restore a days worth or changes on a 40 gig drive holding mainly a
small store of SQL, Exchange, & fax is _insane_.

He is going to add raid protected storage regardless of what we say
here _and_ there are some configuration issues to deal with. Get with
the program already.

Could be. We have insufficient info.

Then how about holding your theoretically based tenuous arguments for
one minute and allow Chris to answer my questions?

yes, _IF_.
Hundreds of dollars later the problem may still remain.

If there is unreasonable disk contention good planning & inexpensive
upgrades (which he already plans to do anyway) will solve it. You're
the one _guessing_ about upgrades & suggesting potentially thousands
wasted in white elephant spare computers. (Duplicate servers may be
good as a part of business continuity/disaster recovery planning - not
your "reliability" mumbo jumbo. If the HW is good & well built you
should get years of hard use out of it where your only real concern is
the software management, security, & very infrequent, long foreseen
upgrades to match growth.)

Lets say he was already planning on a 4 or 6 drive raid 5 or 10
volume. 2 or 3 raid 1 volumes may sound less sexy but might work
better in a jack of all trades server. The decreased latency and
actuator contention could very well be preferable & also compensate
for potential shared pci 32/33 bus issues. Think about it or
experiment with such a setup before responding. It's more efficient &
cost-effective & may be necessary if his hands are tied budget wise &
there is pressure from above for raid protection.

So he should buy a new tape drive instead?

Maybe. He'd have to answer my initial questions on the backup to know
for sure. (how 'bout resisting the urge to blow more hot air and wait
for his response so the facts of the situation are on the table
instead of your guesses?)

He needs space though so he's probably going to buy raid storage
first/anyway. I just want him to look at everything & make sure
everything new & old plays nice so he can find the most logical path
to his goals. I'm loosing interest in your speculation.

 

[kony]

No. He has noticed drive contention. That's why he expects
improvement by separating the data stores - so it's not theoretical.
(reread his choice of words.)

Do you mean where he wrote:

"Basically, I have the one server that does everything I
need it to do, and currently, it works great. Its just that
the users' files and the email system are growing
so the server is running out of room. " ?

Your explanation attempts to prioritize
the sources of contention but doesn't contradict Chris' ostensible
observation. We can't just make believe he didn't notice or elude to
noticing anything.

You're just trying to sell SCSI RAID again, regardless of
the need (or lack thereof). Who do you think you're trying
to kid?

RAM won't deal with that drive contention and it can't cache
everything.

WRONG!
A server's drives can completely spin-down if:

1) Not logging anything
2) Not running windows which is a pig of an OS that
constants wastes resources. Regardless, even then it can
have very minimal contention.

Only Chris can tell us how much contention there is or
how "decent" his "server" is. From his comments, though, this appears
to not be the total non-issue you'd like to imagine.

From the description of the situation, that's it's running
fine for the most part from a single 40GB drive, I'd pretty
darn confident that given enough memory and a single 2nd
drive, it'll have the space he needed, which was the whole
point of this thread least we forget.

Yes & no. I don't think that is a responsible generalization or guess
re the lan.

I don't think your misleading examples are a responsible way
to rationalize a suggestion either, so let's just stick to
the facts. The FACT was the need is additional storage, nor
disk subsystem performance. Even so, adding another
drive(s) will do so any way we look at it.

Esp given the nature of SQL & exchange access/usage
patterns & what that does to a server. Hyperbole aside there's
certainly no reason why a "souped up desktop" couldn't have a single
gigabit connection or dual fast ethernet.

Yes even an old Pentium 500 can handle Gigabit, at least
enough to make substantial difference over 100Mb.
Regardless, the issue here is NOT the performance, it's
adding storage space.

We also don't know what's
going on on the lan like roaming profiles, Qos, etc. or even how these
employees work. Also many of your tenuous theoretical suggestions
haven't been nearly as fiscally responsible as you might imagine.
Your guesses re server behavior can't come from field experience.
(BTW contesting hyperbole makes for a comical argument.)

Sorry but you're again grasping at straws. We don't need to
know what's going on. We are not the administrator. Simple
solution that isn't worth all this nonsense would be to just
buy the drives and be done. SCSI is NOT a bad choice here,
but not necessary either. Since the topic of external
drives popped up it might even be the best choice, except
that it's not at all clear that this server needs external
drives with 4 empty bays and only a need for 80GB.
An entire new mATx emachines with a Gigabit nic would
probably do the job and cost less than a SCSI enclosure
alone, sans the drives or controller.

No. You just missed it & refuse to acknowledge it when I showed it to
you.

Whatever, it's not my dime. Anyone can just try to sell the
most expensive solution then dream of potential scenarios
where it might help. In reality, most businesses don't feel
they're not getting adequate performance from modern gear,
their primary concerns are overpaying for tech they don't
need... at least that's what the last few polls have
determined.

Disk contention was THE reported performance issue. It may not
presently be of a crippling severity but was noticeable nonetheless.

.... so what?
Unless we argue that he should use drive compression to gain
space, any other alternative is going to be adding drive(s).

Actually, let's get this straight.

DISK CONTENTION WAS NOT _THE_ REPORTED PERFORMANCE ISSUE.
Where do you get these crazy ideas?

ANOTHER quote:
" it works very well and should continue to work that way.
The only thing is now we need more drive space."

Even after mentioning it multiple times, you're still hoping
that you'll magically change the facts just by going on and
on about irrelevant things?

We don't know what the cause of the tape backup issue is. It could be
disk or software or both. It could also very well be (if my math is
correct) a DLT4000 (autoloader?) running a complete backup every
night. In which case all we have to suggest is to run incremental or
differentials nightly & full backups less often. A DLT7000 or better
may also help (esp for recovery times) without breaking the bank. A
higher capacity tape backup may also be warranted to keep up with
expanding data (Chris - don't bother with the "value" line of dlt's).

Well at least we're coming closer to same mindset on some
issues.

You haven't ever used industrial tape before, have you? Yes it works
when it is set up right. But most tapes can't slow up to compensate
for poor system thruput (which may or may not be purely hardware
related). (I only know of newer HP ltos which can do this
compensation). That's why I'm endorsing troubleshooting it or at
least finding out more about the setup. geez!

Yes, but you didn't endorse that, you argued that as a
reason for SCSI RAID. If he decides on SCSI RAID, so be it,
but it's not a (this) therefore (that) scenario. The tape
issue should be looked at independant of the type of disk
expansion.

Then how about holding your theoretically based tenuous arguments for
one minute and allow Chris to answer my questions?

Pot calling kettle black?

Rest of post snipped.
It's all a waste going on about this.
Better to weigh the pros and cons of the external storage
option and leave it there. Either way the storage is
needed.

 

[Curious George]

Do you mean where he wrote:

"Basically, I have the one server that does everything I
need it to do, and currently, it works great. Its just that
the users' files and the email system are growing
so the server is running out of room. " ?

No stupid. It's where he said:

"It should also speed up the entire server a little because the hard
drive doesn't write to multiple places at the same time so much."

key words being "so much" as in "as much as it does now"

just as it stared you in the face before.


"works great" or "works fine for the most part" likely means
"sufficient reliability" & "sufficient performance" & "does what he
needs." That's very different from meaning there are absolutely no
bottlenecks of any kind or any significance whatsoever. It means
things are within tolerable limits, working, & meeting the
organization's needs.

I noticed that he eluded to a bottleneck (which you're blowing way out
of proportion because you can't accept that you missed something even
though it happens all the time) on a subsystem that is already slated
for reexamination & upgrade. Even though things have been OK it's
something to be aware of NOW as opposed to 6 months after an upgrade
is rolled out and then scampering around trying to get more budget
approved & more upgrade time scheduled with the technical &
nontechnical bosses scowling at the miscalculation. It would be seen
to them that they need to now fix his upgrade because he didn't choose
correctly & might raise flags that they should give him less control
of the computers or at least upgrade process.

Of course you wouldn't know anything about this. Whenever you want a
new toy mommy caves when you cry.

You're just trying to sell SCSI RAID again, regardless of
the need (or lack thereof). Who do you think you're trying
to kid?

No stupid petulant child. SATA, Firewire, & PATA are all still on the
table in this thread. The solution offered for drive contention was
really load-balancing-like configuration. SCSI Raid in general is
certainly a viable option for a number or reasons but not necessarily
a total or very specific solution to deal with this - which is why I
didn't discuss it that way.

What exactly is your paranoia & obsession about scsi raid anyway?
Must you despise everything you don't understand?

A multi-role mission-critical company database server is the supreme
example for scsi or scsi raid being necessary or at least warranting
consideration. _Especially_ if there is a preference or a need to
house the databases in external DAS.

I still can't figure you out. Are you so stupid that you don't know
how stupid you are or sound- or do you deliberately troll by the most
pathetic technique imaginable; taking the position of a know-it-all
ignoramus?


If I'm trying to sell scsi raid solutions, where's my corporate
signature or email or URL? When did I offer an email quote? What
company exactly do you think I work for or own that makes or sells
exclusively scsi products and that has such a large market share of
scsi raid in total that a lone person occasionally endorsing scsi in
general for textbook scenarios everyone already knows (including
recommending some eol products of multiple manufacturers) would impact
their bottom line? (surely you can't honestly believe these illogical
paranoid delusions. Really stop it! Don't listen to the voices!)

I'd like to once make a penny off of the free help I give in usenet.
Forget it. Having ignorant bigot know-it-all punk kid delusional
trolls like you stay away from me would be payment enough.

WRONG!
A server's drives can completely spin-down if:

1) Not logging anything
2) Not running windows which is a pig of an OS that
constants wastes resources. Regardless, even then it can
have very minimal contention.

I don't know what the hell you think you are talking about. It makes
absolutely no sense!!! If there is no activity than a machine's specs
incl ram size are of no import. Drive contention doesn't even exist &
the ram might as well be caching nothing! Who the F**K cares that the
machine is doing nothing & can spin down & conserve power when no one
is using or wants to use it? What _matters_ is when 15 or 20 ppl want
to log in to the domain and run SQL reports & audits, post data, read
email, access data for customers, & make & review faxes & electronic
documents on the server all at the same time (which isn't really a
full fledged server of unknown age- but probably at least several
years old already) and all while email comes in from the outside & the
machine does its IIS or ISA stuff & encryption & software upgrade
management or whatever the hell else he has going on!

& how the hell exactly is he supposed to run his MS SQL & Exchange
(yes it's all MS that's the whole point of him mentioning SBS) in a
realistic way that makes sense to his bosses on his x86 pc in some
*nix or vms or dos or CP/M?

You should really quit while you're behind. You're just coming across
as ignorant & posturing for the last word at all costs without being
able to draw upon any logic or knowledge.


O, why the hell am I still humoring this absurd, clueless drivel?


One final thought for this thread though:
I think it's very funny and telling that you would endorse a "mATX
emachine" for the corporate server described. Really funny. Thanks,
I needed a good laugh. No wonder you have no concept of reliability
or hard use & you're accustomed to system boards & systems dying in no
time and think the solution is to buy extra computers and rotate them.
All really choice stuff.

(BTW buying a PeeCee with a plastic window & a light in it doesn't
qualify you as a "computer guru". That stuff might of turned heads of
a few kids in the early or mid '90's but you were little more than a
glint in the mailman's eye back then.)


Well, you won. Go ahead and take a victory lap. Do your little
dance. You've wore me down & your stupidity got under my skin. I
just don't have any more time or energy to school you; correcting your
misinformation and exposing your posturing. I really don't know why I
have been teaching you so much as a reward for being an ignorant snot.
I guess I was too distracted by the idea of Chris or someone else in
the group learning something or chiming in or challenging me.

Savor this grand victory. Enjoy it. You must really feel big to
hijack another thread with stupidity. You obviously invested a lot of
time on this. (Boy that ass of yours must really be sore - with all
that you pulled out of it in the last couple of days). 8-0


Good Riddance.

 

[kony]

No stupid. It's where he said:

"It should also speed up the entire server a little because the hard
drive doesn't write to multiple places at the same time so much."

key words being "so much" as in "as much as it does now"

just as it stared you in the face before.

See? This is how low you stoop.
That was not an issue prompting replacement, it was a theory
about a secondary benefit, but one NOT needed as multiple
times it was already CLEARLY stated that beyond the disk
space and tape backup times, it was doing fine for their
needs.

"works great" or "works fine for the most part" likely means
"sufficient reliability" & "sufficient performance" & "does what he
needs." That's very different from meaning there are absolutely no
bottlenecks of any kind or any significance whatsoever. It means
things are within tolerable limits, working, & meeting the
organization's needs.

Wow you really are an idiot.
"Works great" certainly does not mean any-random-thing you'd
like to tack on to suit your argument.

I noticed that he eluded to a bottleneck (which you're blowing way out
of proportion because you can't accept that you missed something even
though it happens all the time)

Again, more nonsense.
I read it, nothing was missed.
The thread was about adding storage space. Perhaps you
should re-read it when you get done babbling.

...on a subsystem that is already slated
for reexamination & upgrade. Even though things have been OK it's
something to be aware of NOW as opposed to 6 months after an upgrade
is rolled out and then scampering around trying to get more budget
approved & more upgrade time scheduled with the technical &
nontechnical bosses scowling at the miscalculation. It would be seen
to them that they need to now fix his upgrade because he didn't choose
correctly & might raise flags that they should give him less control
of the computers or at least upgrade process.

.... or when they find that all that was needed was a $60 ATA
hard drive and he spent $500-1200, he's not looking so good
either.

Of course you wouldn't know anything about this. Whenever you want a
new toy mommy caves when you cry.

I now realize that you're just a troll that acts like a baby
when others don't agree with you. My mistake was giving you
the benefit of the doubt, that you at least THOUGHT you knew
what you were talking about.

No stupid petulant child. SATA, Firewire, & PATA are all still on the
table in this thread. The solution offered for drive contention

You mean the "drive contention" issue that you pulled out of
your ass, right? It could be that type of problem but we
still haven't determined this factually.

... was
really load-balancing-like configuration. SCSI Raid in general is
certainly a viable option for a number or reasons but not necessarily
a total or very specific solution to deal with this - which is why I
didn't discuss it that way.

What exactly is your paranoia & obsession about scsi raid anyway?
Must you despise everything you don't understand?

Reread my last couple posts, I"m quite sure I mentioned SCSI
RAID as being one alternative. The difference was that I
was recommending it if/when for known benfits, not
hypothetical scenarios when the whole topic started out as
how to add some storage space.

_IF_ the external storage enclosure is chosen, SCSI RAID is
a good choice. This is because it's a bus rather than a
port, not because of any of the drifting speculation and
self-serving arguments you've been making.

A multi-role mission-critical company database server is the supreme
example for scsi or scsi raid being necessary or at least warranting
consideration. _Especially_ if there is a preference or a need to
house the databases in external DAS.

Warranting consideration, yes. Supreme example? Certainly
not.

I still can't figure you out. Are you so stupid that you don't know
how stupid you are or sound- or do you deliberately troll by the most
pathetic technique imaginable; taking the position of a know-it-all
ignoramus?

So when your half-baked argument fail you feel attempted
insults will help? Considering the logic you've used when
TRYING to stay on topic, this is not surprising at all.

So basically you feel that when someone doesn't agree, you
can stop around and whine such that, suddenly they WILL
agree? LOL.

If I'm trying to sell scsi raid solutions, where's my corporate
signature or email or URL? When did I offer an email quote? What
company exactly do you think I work for or own that makes or sells
exclusively scsi products and that has such a large market share of
scsi raid in total that a lone person occasionally endorsing scsi in
general for textbook scenarios everyone already knows (including
recommending some eol products of multiple manufacturers) would impact
their bottom line? (surely you can't honestly believe these illogical
paranoid delusions. Really stop it! Don't listen to the voices!)

Paranoia seems more like a description for someone who just
wrote that whole paragraph. Verbose is good when there's a
lot of content, but blowing off hot air is better done at
the gym.

I'd like to once make a penny off of the free help I give in usenet.
Forget it. Having ignorant bigot know-it-all punk kid delusional
trolls like you stay away from me would be payment enough.

I'd like it if you were charged every penny that you've cost
people by suggesting hardware they don't need. In this
thread, it may actually be that SCSI RAID is the solution,
but not because of the arguments you've been making.

I don't know what the hell you think you are talking about. It makes
absolutely no sense!!!

You really mean that you deny anything you don't understand,
right?

If there is no activity than a machine's specs
incl ram size are of no import.

Where do you come up with this crap?
There can be MASSIVE activity.
It can be so busy the CPU and network is constantly pegged
at 100%, and be using memory the whole time.

Perhaps you don't know anything about servers, and have
starved them for memory so much that it just seems like SCSI
RAID helps because they're always hitting the pagefile.
This is starting to drift off topic though, I never claimed
this was the case with the OP's specific server, only to
attempt (but fail) to educate you about this "contention"
hypothetic you want taken as gospel.

Drive contention doesn't even exist &
the ram might as well be caching nothing!

Truely cluess.
Boot up any modern OS and the ram is caching EVERYTHING.
The ram continues to cache EVERYTHING until I/O exceeds size
of available memory, then cache is flushed. You'd have us
believe servers generally have significant memory installed
so they can just leave most of it unused?

Who the F**K cares that the
machine is doing nothing & can spin down & conserve power when no one
is using or wants to use it? What _matters_ is when 15 or 20 ppl want
to log in to the domain and run SQL reports & audits, post data, read
email, access data for customers, & make & review faxes & electronic
documents on the server all at the same time (which isn't really a
full fledged server of unknown age- but probably at least several
years old already) and all while email comes in from the outside & the
machine does its IIS or ISA stuff & encryption & software upgrade
management or whatever the hell else he has going on!

You're the only one suggesting the machine "is doing
nothing". Apparently you believe people should spend $1000
on their disk stoarge system just to have a box only 1/4 as
fast as if it had ample memory.

& how the hell exactly is he supposed to run his MS SQL & Exchange
(yes it's all MS that's the whole point of him mentioning SBS) in a
realistic way that makes sense to his bosses on his x86 pc in some
*nix or vms or dos or CP/M?

Even with Windows, the "contention" can be very light.
Non-MS OS was a remark regardling complete spin-down of
drives.

You should really quit while you're behind. You're just coming across
as ignorant & posturing for the last word at all costs without being
able to draw upon any logic or knowledge.


O, why the hell am I still humoring this absurd, clueless drivel?

Because your argument was poor and you still don't grasp how
sticking-on-topic and focusing on the facts presented, is
the way to a desired result.

One final thought for this thread though:
I think it's very funny and telling that you would endorse a "mATX
emachine" for the corporate server described. Really funny. Thanks,
I needed a good laugh. No wonder you have no concept of reliability
or hard use & you're accustomed to system boards & systems dying in no
time and think the solution is to buy extra computers and rotate them.
All really choice stuff.

You posed one extreme, I countered with the other.
Truth be told, there is nothing particularly bad about using
an eMachine for light serving, if one had ECC memory, an
upgraded power supply, and either very regular backups or
the RAID, to be determined by the admin as that is who knows
the real need.

As for "system boards and systems dying", are you aware that
it's pretty obvious when you pull crap out of your ass with
no factual backing? I NEVER suggested buying extra
computers just to rotate one when another dies. I DO
advocate redundancy though, partly because of blundering
idiots like you who can't stick to facts and freakout when
facing even very simply issues... which ultimately results
in downtime, extra expense, and wasted time.

Again, I'm not opposed to SCSI RAID for this thread's
purpose. I'm against all the BS you spew out to suit your
arguments. When you spew the BS, I comment, and you start
name-calling. Truely a waste of your time.

(BTW buying a PeeCee with a plastic window & a light in it doesn't
qualify you as a "computer guru". That stuff might of turned heads of
a few kids in the early or mid '90's but you were little more than a
glint in the mailman's eye back then.)

So again you have no argument so you desperately hope I am
some young kid with no computer experience. Neither is true
but honestly it doesn't make a damn bit of difference, what
matters is whether you can manage to pull your head out of
your ass and keep on-topic and deal with supplied facts
instead of making up your own as you go along.

Well, you won. Go ahead and take a victory lap. Do your little
dance. You've wore me down & your stupidity got under my skin. I
just don't have any more time or energy to school you; correcting your
misinformation and exposing your posturing. I really don't know why I
have been teaching you so much as a reward for being an ignorant snot.
I guess I was too distracted by the idea of Chris or someone else in
the group learning something or chiming in or challenging me.

Translation - Some imagination but that's now reached it's
limits.

Savor this grand victory. Enjoy it. You must really feel big to
hijack another thread with stupidity. You obviously invested a lot of
time on this. (Boy that ass of yours must really be sore - with all
that you pulled out of it in the last couple of days). 8-0

I do hope you feel better now... otherwise you just wasted
your time again.

 

[Curious George]

very creative back peddling Kony, I'll grant you. But still you're
grasping at straws, inventing things, & making less sense with every
post.

Yes the more ram the better for sql just as the higher the disk
subsystem IO the better. Yes ram often brings more bang for the buck
than disk IO BUT when the database is too large it can't fit all the
tables regardless. There are limits to what ram can do & believe it
or not, disk IO (including writes) does occur even with gigs of ram.
Imagine that! Would you also believe there are other HW components to
consider in SQL or Exchange HW performance tuning?

YES we have to know details of the server to do any performance
tuning. YES we have to now details of the network & how its used to
know whether the network is indeed a bottleneck. etc. It's awfully
arrogant to think your guesses about guesses are fact or even
relevant.

Your ideas & misinterpretations about the meaning of reliability,
uptime, MTBF, TCO, & ROI are indeed troubling. Your blatant lies
about your previous statements on these topics here and elsewhere are
even more disturbing.

You are indeed correct _I've_ been wasted _MY_ time humoring you, as I
doubt yours is really worth anything to anyone.


Chris, the fact of the matter (& where this all came from) is that
upgrades (even seemingly small ones) are a good opportunity to
reassess the entire situation and recalculate your life cycle
expectations, TCO & ROI assessment, as well as conduct performance
tuning. Don't listen to Kony. You don't just throw in ram or
anything else blindly & you don't just wait for a significant or
crippling problem before you reevaluate performance or any other
issues (I'm not saying you, Chris, in actuality, do)

Check out for starters:
http://www.sql-server-performance.com/hardware_tuning.asp#I/O
http://www.sql-server-performance.com/sql_server_performance_audit.asp
http://www.webpronews.com/it/database/wpn-20-20041210SQLServerHardwarePerformanceChecklist.html
http://support.microsoft.com/default.aspx?scid=kb;en-us;322883
http://support.microsoft.com/default.aspx?scid=kb;en-us;298475

& other articles or books you can find on the topic. Maybe you
already know all this. Regardless I am adamant that these types of
considerations should be included in upgrade planning (even if you
think the end result will be, & ends up actually being, essentially
staying with most of what you already have). Because databases grow &
need to uninterrupted you don't wait for an obvious & crippling
problem either performance, security, or availability wise. Upgrades
& life cycle NEED to be carefully planned. & no Mr. cavalier Kony,
planning & evaluation is not expensive & doesn't _require_ expensive
external consultants. Believe it or not, Kony, being accurate,
prepared, thorough, & generally ontop of things is a good thing & is
not expensive. Being ill-prepared, negligent, & cavalier is.

The cheapest storage you can find without thought & the most ram you
can afford is not always the ideal route or the most comprehensive
approach. As far as what is good enough; well to thine own network be
true. Just a summation of my position for you to chew on. I have
nothing more to add here. The hypotheses about your actual situation
have been covered ad nauseam. I'm sure by now you are either
thoroughly disgusted or scared away from follow up posting.

 

[kony]

On Sun, 30 Jan 2005 09:04:35 GMT, Curious George

Believe it or not, Kony, being accurate,
prepared, thorough, & generally ontop of things is a good thing & is
not expensive. Being ill-prepared, negligent, & cavalier is.

... and this has to do with the price of tea in China?

I haven't lost any data in years, don't have machines
breaking down like you imply either... nor do most people,
but apparently you do unless you spend a premium on only one
narrow type of hardware?

Maybe if you were a little more proactive, doing more than
obsessing over SCSI, you'd have prevented all these problems
you feel are imminent. Instead you just suggest one
solution and leave out the part about it NOT magically
solving anything.

Getting back on topic again, IF the OP wants a seperate
storage enclosure and RAID3 or 5, then SCSI RAID is the best
solution. HOWEVER, these criteria are not needed for this
application. What "Curious George" has been going on and on
about is apparently the only thing he knows, so he just
repeats it like a parrot and repels anyone who'd be able to
teach him something new to parrot.

 

[Curious George]

On Sun, 30 Jan 2005 09:04:35 GMT, Curious George



.. and this has to do with the price of tea in China?
clueless

I haven't lost any data in years, don't have machines
breaking down like you imply either...

Then you are either lying here or in other threads.

nor do most people,
but apparently you do unless you spend a premium on only one
narrow type of hardware?

still clueless & wrong

Maybe if you were a little more proactive

that what I'm talking about fool

, doing more than
obsessing over SCSI, you'd have prevented all these problems
you feel are imminent. Instead you just suggest one
solution and leave out the part about it NOT magically
solving anything.

clueless, wrong. blatant misrepresentation

Getting back on topic again,

oh puhleeze.

IF the OP wants a seperate
storage enclosure and RAID3 or 5, then SCSI RAID is the best
solution. HOWEVER, these criteria are not needed for this
application. What "Curious George" has been going on and on
about is apparently the only thing he knows, so he just
repeats it like a parrot and repels anyone who'd be able to
teach him something new to parrot.

actually raid 3 & most raid 5's are likely very poor choices for his
server. Since you either have no reading comprehension abilities or
are a deliberate liar I'll tell you again what I've been talking about
ad nauseum: load balancing storage configuration & taking a storage
upgrade as an opportunity for reassessment & performance tuning. You
just want to believe your delusions. You like them don't you?

 

[CBFalconer]

very creative back peddling Kony, I'll grant you. But still you're
grasping at straws, inventing things, & making less sense with every
post.

High time for a thread PLONK. Take the personalities to e-mail.

 

[Curious George]

High time for a thread PLONK. Take the personalities to e-mail.

You feel better now Mr. Dos IRQ's? Yeah that's a great idea lets
escalate a simple "LA LA LA I'M NOT LISTENING" scenario (you could
otherwise easily bypass) into attacking personal accounts. Saying
Plonk in a forum is an easy, safe strategy towards appearing to
intervene & act superior.


Despite all the stupidity here, fact is Chris isn't going to be able
to fit more than 2 or 4 gigs in his SBS desktop. There's no talk
about AWE, PAE just ram in generic, mystical, & conflicting terms.
Lets get real- it makes no sense to suggest Chris in this scenario be
flippant about storage or the upgrade planning. And no scsi in
general terms is not necessarily the whole answer.

I may have been overanalyzing & reading into what Chris said too much,
but then that would be the counter argument. Not the crap that's been
thrown around here. Yes there are real issues that merit discussion,
& are inline with the OP's questions, even if personality issues
obscured that.

 

[kony]

Then you are either lying here or in other threads.

Nope. I HAVE had HDDs fail, but no data lost. It's called
backups.

As for machines breaking down, I REPAIR other peoples boxes.
The elder 1/3 of the equipment I have is that which was
broken before it even came to me, and yet I STILL have it
working fine a large percentage of the time. So yes, I may
comment from time to time about hardware failure, but
because I SEE it happening in equipment that comes it with a
problem already.

Beyond that, I welcome you to obsess over this and attempt
to dig up anything to the contrary... so long as you can
stick to the facts presented instead of making things up as
you go along.

Regardless, CBFalconer is right, the thread is dying and
we've not been productive. I'm done unless you feel like
proving your comment about "lying".

 

[Chris Guimbellot]

To both kony and Curious George (and anyone else monitoring this thread):

I can definitely say that I have never had a posting with such a passionate
response. I do appreciate it very much though. I am learning (if nothing
else) that there are many more things that I need to consider before I do
anything. I also wanted to apologize for taking so long to respond to the
post. I was called away for a few days, and to be honest, I have been
sitting here for three hours just reading all of the replies so that I could
try to put some coherent thoughts in order so that I could compose this
response. There are so many lines of analysis here (and so many posts that I
missed), I am going to try to formulate my response in one post as opposed
to responding to each of you. I think that would better suit the flow. I
will try to snip when possible prefacing the snip with the
author. On other topics I will just refer to the topic since there are too
numerous an
amount of snips to include. I hope the two of you can bare with me through
this post. That said, here goes.

I figured I would start out with a little bit about the box itself and what
we use it for. Basically, we are a small office of about 20 people (only
about 15 of which are logged in at one time). The box is a PIII1GHz with
512MB of RAM. In regards to the SBS setup, of course Exchange is running all
the time and most of my users have Outlook open all the time. SQL, on the
other hand, has only one database (in addition to the built-in ones) and it
is
not transactional. It simply houses info and may get changed once or twice a
day. I use roaming user profiles, and send faxes, etc (of course I bought
fax boards to relieve pressure from the processor). The standard small
office server-type stuff.

CURIOUS GEORGE

Hey Chris. While you're rethinking this server, how are the cpu, ram,
and network interfaces doing during normal use?

In regards to where a bottleneck (or potential bottleneck) could be, I have
no idea. I dont know how to test those particular things such as the NIC,
RAM use, disk writes, etc. I would be happy to learn how though if either of
you knows any sites, tests, etc. Also, for what it's worth, the NIC is a
simple 100Mb/s (you were right kony) so do not have any gigabit equipment. I
guess I would need a new switch too if that was the case that the NIC was a
bottleneck. In essence, I do not know how to perform tests so that I could
effectively answer your question. Even if I could, I have no benchmarks to
compare against.

CURIOUS GEORGE

Also you mention having ATA inside & also the current drive is SCSI. What
exactly do you have as far as disks and how are they being used?

Inside the server, I have two disks: the first is my main drive. It is a
SCSI drive with 40Gb of capacity (this is the one running out of room). The
second drive is a 10Gb ATA drive that I put in it a long time ago for
temporary purposes. Actually, I need to pull it out because it is ancient.
It needs to go into the circular file.


CURIOUS GEORGE

We don't know what the cause of the tape backup issue is. It could be
disk or software or both. It could also very well be (if my math is
correct) a DLT4000 (autoloader?) running a complete backup every
night. In which case all we have to suggest is to run incremental or
differentials nightly & full backups less often. A DLT7000 or better
may also help (esp for recovery times) without breaking the bank. A
higher capacity tape backup may also be warranted to keep up with
expanding data (Chris - don't bother with the "value" line of dlt's).

Great guess. Actually, it is an HP SureStore 40, a DLT4000 mounted
externally and connected to the SCSI card in the server. It is not an
autoloader. I manually switch the tape every morning. I guess the 4000 part
would be
why my backup is going slow. In regards to the full versus differential
backup, all of the SBS MVPs told me to, as well as do themselves, a full
backup every night. They have always known exactly what was going on, so I
trust their opinion and will stick with the full backups every night.
Besides the length of time they take to complete, the full backups and
SureStore have never let me
down. By the way, I use the backup program in Windows to do the job. If I do
need a DLT7000, that is not a problem. I can get one relatively
inexpensively. I would however rather not buy one if I dont have to, but I
know that as my data growns, so will the backup time. Also, since we are
moving our Marketing department (only two uesrs) from Mac to PCs, I am about
to put a few more GB of files on the server, probably necessitating the need
for a higher capacity tape drive, and definately more storage be it RAID or
a simple HDD. Also, Curious...what are you referriong to by 'value line of
DLTs'? I have never heard of that.

I guess the thing to do now is to check for the existence of a bottleneck,
and if so, fix it. As I mentioned above, any help with that would be great.
Now that I have thought about the coming switch from the Macs and adding
those files to the server, I now think I should definately consider swithing
to a higher capacity tape drive in addition to any upgrades I make to the
hard drive system. There needs to be some upgrade done on storage ( I figure
we all can agree on that). I guess the only question now is what kind.

Once again, I appreciate the help and the insight and I look forward to your
responses. Thanks again,

Chris

 

[Curious George]

I thought we scared you away. I know I needed a break from here.
While I would have liked to never participate in or see this thread
again (& to try to recoup some of the time I wasted here by avoiding
usenet in general for a while) much of the time I've been talking
about my desire for you to answer my questions & respond to your
particular situation - so it's only right & not blow you off & respond
(esp as no one else has yet).

On Mon, 31 Jan 2005 11:54:21 -0500, "Chris Guimbellot"

I figured I would start out with a little bit about the box itself and what
we use it for. Basically, we are a small office of about 20 people (only
about 15 of which are logged in at one time). The box is a PIII1GHz with
512MB of RAM. In regards to the SBS setup, of course Exchange is running all
the time and most of my users have Outlook open all the time. SQL, on the
other hand, has only one database (in addition to the built-in ones) and it
is not transactional. It simply houses info and may get changed once or twice a
day.

Well then my inference was wrong. 40gigs & growing of mainly SQL,
Exchange & Fax sounded to me like a much larger more write intensive
SQL environment. That's because you didn't mention file service or
software management responsibilities (which can take up a lot of
space) & Exchange space is often contained through quotas & fax space
is often be limited through archival after a certain limited period of
time.

It's not necessarily common but there are some small businesses or
certain business types where even a handful of really good employees
or contracts can make a decent footprint on the SQL on a small
machine. If everything is so small that 512 ram is enough to hold the
database & everything else & everyone is always completely happy this
is an entirely different animal (is it?).

Problem for me was different ppl are looking for or value different
things so a statement like "works great" for one person can be a
mainly reliability & availability statement. For someone else it can
be a mainly performance statement. I can only therefore take it to
mean "meeting needs" in basic, nonspecific terms. If works great
means every one has instant, zippy access to everything at all times &
data & workstations are growing extremely slowly you can likely ignore
most of this thread's performance "discussion" or at least take it
with a grain of salt. (sorry)

I use roaming user profiles, and send faxes, etc (of course I bought
fax boards to relieve pressure from the processor). The standard small
office server-type stuff.

roaming profiles, software management, etc (infrastructure type stuff)
can bog down the network so were mentioned in the lan
performance/bottleneck generalization/hypothesis. My guess is these
activities should not really severely impact your environment- but you
have to tell us. Hmm, multiple fax boards on a small, entry level,
budget server. So faxes esp mass faxes play a pretty significant
role?

CURIOUS GEORGE

In regards to where a bottleneck (or potential bottleneck) could be, I have
no idea. I dont know how to test those particular things such as the NIC,
RAM use, disk writes, etc. I would be happy to learn how though if either of
you knows any sites, tests, etc. Also, for what it's worth, the NIC is a
simple 100Mb/s (you were right kony) so do not have any gigabit equipment. I
guess I would need a new switch too if that was the case that the NIC was a
bottleneck. In essence, I do not know how to perform tests so that I could
effectively answer your question. Even if I could, I have no benchmarks to
compare against.

You primarily want to identify something that is being hammered and
does, or has the potential in the near term with expected growth,
create a slowdown the users will notice. If everyone is happy and the
machine can withstand projected growth there is no benefit in dealing
with a benchmark number as it is mainly used for comparative purposes.

My exaggeration re the nic was simply to make the point we don't know
much about the computer & the network & it's important. I wasn't
really trying to guess it. For example a graphics design firm that's
constantly moving largeish files between the server & the workstations
(who actually manipulate the files) is going to have different needs &
different bottleneck areas to look at than say a billing service where
the workstations primarily send small sql requests the server has to
chew on. What about say a Radiology practice with both large images &
billing depts over both lan & wan. These could all easily be called
"small businesses" with 20 employees or less, all requiring mostly
file service, SQL, Exchange, & Shared Fax, & all with very different
needs & potential bottleneck sources.

I'm not telling you to upgrade the switch &/or nic just yet (unless
they're getting hammered & there's a problem). Just FYI the
inexpensive Intel Pro adapters can work together well and the adaptec
10/100 quad server nics can be inexpensive & solid. Esp with a good
switch, port aggregation or gigabit server link can be a simple cheap
upgrade that can help many Fast Ethernet users gain access to the
server. It's a possibility even on a budget server but there are
limits to what you can realistically expect here when so much is on
the same 32/33 pci bus or some el-cheapo switches.

How do you identify bottlenecks? by monitoring & studying the
server's load and the demands during the day, especially busy
"crunch-time" type periods. The bundled OS & SQL monitoring tools as
well as the websites I cited are a good place to start & digest the
concepts. Specifics & how you set up the monitoring are going to
depend largely on what you start noticing & specifics of use/needs but
will focus on the core HW components: CPU, RAM (& virtual memory),
Disk IO, Network Interface & may need to look at the core software
requests of said HW. Benchmarking, per se, may only play a smaller,
secondary role in all this; it will provide a basis for determining
the performance benefit of a specific upgrade.

CURIOUS GEORGE

Inside the server, I have two disks: the first is my main drive. It is a
SCSI drive with 40Gb of capacity (this is the one running out of room). The
second drive is a 10Gb ATA drive that I put in it a long time ago for
temporary purposes. Actually, I need to pull it out because it is ancient.
It needs to go into the circular file.

The confusion (& why I asked) was from calling it a 40gig drive (which
is an ata capacity) as opposed to what I guess now is really a 36gig
scsi drive. Problem is, from a server performance perspective, even
an older scsi drive with significantly less raw thruput, can often
deliver higher IOs than "faster" & newer ATA drives. It's still hard
to say therefore from this description what you need to be happy and
how that steers you away scsi or ata or ATA/firewire, etc. Also there
is some performance benefit with NTFS simply by moving to a volume
which is less than 80% full. I will agree somewhat with Kony here in
that 512MB ram is not a lot for a server & that adding ram may help
things (if needed). Whether it does & how much really depends on the
details.

Upgrade advice? Ok so get rid of the ancient ata drive if it doesn't
help. You have a choice now mainly between:

1. keep the 36GB scsi drive & add a small array for the data
2. do the same as 1 but also upgrade that 36GB drive to an array
3. copy over the present 36GB drive onto a larger capacity bootable
array.

It may not be necessary to separate everything over different
controllers/arrays (although ppl frequently recommend it for several
reasons). Some of this decision depends on whether I was over-parsing
your words & you did or did not in fact notice the drive thrashing
heavily & whether you feel there is some ease of
recovery/serviceability by separating OS & data stores & that the
extra SW & HW configuration changes & expense are worth it.

CURIOUS GEORGE

Great guess. Actually, it is an HP SureStore 40, a DLT4000 mounted
externally and connected to the SCSI card in the server. It is not an
autoloader. I manually switch the tape every morning. I guess the 4000 part
would be why my backup is going slow.

Really backup time is not that important. Restore time, ease, &
reliability is. If it takes a full business day or longer with you
babysitting the machine & twiddling your thumbs to do a complete
restore in the event of a disaster, that is a problem. The business
will loose a lot more money than the price of a new DLT. (FYI if you
are patient you should be able to sniff out a new SDLT1 or LTO1
autoloader in the $1000-$2000USD range on ebay. The VXA2 (stand alone
& autoloader) are also good budget options (that is to say the drive
NOT the tapes). Remember while DLT7000 is a good value, it is "only"
35gigs uncompressed/cartridge & you'll be buying new cartridges for it
anyway)

In regards to the full versus differential
backup, all of the SBS MVPs told me to, as well as do themselves, a full
backup every night. They have always known exactly what was going on, so I
trust their opinion and will stick with the full backups every night.
Besides the length of time they take to complete, the full backups and
SureStore have never let me
down.

Many ppl don't do full daily backups basically because they chew
through storage, time, & generate a lot of extra wear & tear. With
incrementals/differentials you are trading some convenience through
simplicity for convenience through cost & time & media esp for
retaining multiple backups.

However full backups still need to be done & incrementals need to be
coalesced into full ones at some point or some way anyway - so I'm not
going to say it's a wrong strategy. Full backups certainly work & are
straightforward. It's a good & safe recommendation. Whether the
strategy is wrong really depends on what you are doing with these
tapes. If every day or every week you are simply overwriting the same
tapes and loosing ALL those old snapshots without any media rotation
or offsite transport; that's not cool (yeah you're probably not doing
that)

By the way, I use the backup program in Windows to do the job. If I do
need a DLT7000, that is not a problem. I can get one relatively
inexpensively. I would however rather not buy one if I dont have to, but I
know that as my data growns, so will the backup time.

as will the restore time/downtime when the business needs it.

no more than one tape per disk volume is often not a bad strategy, esp
with no autoloader. With the DLT 4000 you are looking at up to 3
tapes & manual changes per day or per restore (in the near term).
This can be a problem. It also may be coming close to exceeding your
backup window in the near future. Incrementals/differentials will
help with ease of backup & shorten the backup windows to some degree,
but that "solution" still neglects the restore (which is really all
that matters in the end).

Also, since we are
moving our Marketing department (only two uesrs) from Mac to PCs, I am about
to put a few more GB of files on the server, probably necessitating the need
for a higher capacity tape drive, and definately more storage be it RAID or
a simple HDD. Also, Curious...what are you referriong to by 'value line of
DLTs'? I have never heard of that.

i.e. the DLT1, DLT VS80, DLT VS160 as well as the drives in this
roadmap listed as providing "Value" as opposed to "Performance":
http://www.quantum.com/am/products/dlt/technology_roadmap.htm#top

They're not necessarily bad or lesser quality (AFAIK) there are just
some nuisances that bother me like special cartridges that are harder
to get on discount, slower speed (esp the DLT1), & format
compatibility issues. The lower up-front cost tends to dissolve over
the lifecycle IMHO. Also if you are interested in the DLTIce feature
for compliance you may need to avoid these (I think).

I guess the thing to do now is to check for the existence of a bottleneck,
and if so, fix it.

That is _if_ there is a bottleneck that needs fixing now or that needs
to be planned for when future growth & lifecycle intersect. There was
much ado over a statement I made that began with: "Well I think he did
allude to some problems" & ended with questions to clarify your
situation. You've answered the backup issue & responded to these
questions. Now you still have to determine the disk needs & load
specifically.

Apparently this looks like more of a basic file & fax server than much
of a SQL or mail server; so the balls in your court again. I'm not
going to make any more inferences about allusions here & invite you to
fill in the blanks esp your potential confidence (or lack thereof) in
current performance & whether there is likelihood of performance boost
of a raid upgrade exceeding need or whether a whole server replacement
is forthcoming.

As I mentioned above, any help with that would be great.
Now that I have thought about the coming switch from the Macs and adding
those files to the server, I now think I should definately consider swithing
to a higher capacity tape drive in addition to any upgrades I make to the
hard drive system. There needs to be some upgrade done on storage ( I figure
we all can agree on that). I guess the only question now is what kind.

My vote is a smaller raid 1 boot array & a 74 or 80 gig or better raid
1 for data (not _necessarily_ a performance tip here) on a HW
controller that could handle online array expansion/migration. From
the looks of things, though, everything on a single array might be OK
too. RAID 3 or 4 might be OK (maybe Mylex U160 scsi with 128MB for
"budget") but I am not really endorsing it or raid 5.

It doesn't sound like you really need to high IO's of more smaller
drives for a particular capacity or the fastest 15K screamer.
Internal arrays could work unless you are sold on the advantages on
hotswap, which likely doesn't fit in your chassis. I'm not going to
touch a scsi vs ata suggestion. You doubtless already have some ideas
about their strengths based on current usage.

The storagereview.com's reliability database is an interesting project
you might want to look at before drive purchase. Give thought to the
3ware, LSI, & to some extent IBM controllers. Despite the big name
I'm not sure Adaptec is a great pick here. I don't recommend getting
the cheapest controller you can find or a hotswap backplane if it's of
mediocre quality.

Once again, I appreciate the help and the insight and I look forward to your
responses. Thanks again,

Chris

no prob. I hope this wasn't an all around long-winded time waster
that has sent you on a wild goose chase. Even if I totally
misinterpreted your setup, it still wouldn't hurt to project needs &
capacity & start thinking about the next major upgrade esp or even if
it just gives you that much more confidence in the current platform &
its expected longevity. "Obsessing" over every detail of a smaller
setup seems like overkill but it tends to not really be all that much
work. If it saves you from "mistakes" it is a worthwhile investment.

If other Mac->PC conversions & departments join this server in the
future you will have to reanalyze, but you will do so armed with
baseline measurements & projections that make life a lot easier. You
won't ever be starting from scratch again or caught with your pants
down & can respond or make suggestions to deal with changes quickly.
As long as no one feels threatened by your analysis/suggestions your
extra work here can pay off.