• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Ok my $2000 EX2K server is ready to rumble any suggestions?

H

Hardware

Golden Member
MD: Abit BD7-II Raid
CPU: P4-2.4B
RAM: 3*512 DDR PC266 ECC
GFX: TNT2 M64
HDD: 4*WD 120G 7200RPM 8MB (SCSI 15K class) as 2*Raid1 (120G)

BOOT = Raid1 MASTER0 + MASTER1
DATA = Raid1 SLAVE0 + SLAVE1

LAN onboard 100 (Realtek)

big tower (no server) with 400W PFC

roomtemp is 21C

Sound / USB / SER / Par DISABLE

OS W2KS SP3
 
Wow! That's some server you got there.

It must go right next to your terrabyte FBI fileserver.

I must say though! This is the *first* time i've seen Abit used in a server! OMG!!

(Besides the sole exception of the BP6)

But, I must say, the WD1200JB is in no way 15K scsi class. and is each WD on it's own seperate channel?

You see, that's the thing about SCSI that makes it *SO* good. You can have multiple drives on the same chain in raid without having to worry about them contending for access of the channel.

Plus, your WD harddrives will suffocate in an enviornment when you're constantly transfering 3 or 4K files because of it's abysmal seek time in relation to 15K SCSI drives.

For game servers though, it really doesn't matter at all.

All in all, a wonderful system. Quite a ways off in terms of bandwidth in comparison to your PC1200+ machine.. though..
 
Suggestions about what?

If it's for running Exchange2K, my best suggestions would be to install Exchange for starters 😉

Oh and I agree, a 1200JB is nowhere near a 15K SCSI disk, especially not in a typical server environment.
Nice IDE disks nevertheless 🙂
 
Wow! That's some server you got there.

It must go right next to your terrabyte FBI fileserver.

--- hehe no this is at work for my company

I must say though! This is the *first* time i've seen Abit used in a server! OMG!!

--- hmm intel chipset seems quite solid for a server

(Besides the sole exception of the BP6)

But, I must say, the WD1200JB is in no way 15K scsi class. and is each WD on it's own seperate channel?

I got 7 WD1200JB retail and you can read there 15K scsi class 🙂


You see, that's the thing about SCSI that makes it *SO* good. You can have multiple drives on the same chain in raid without having to worry about them contending for access of the channel.
--- hmm the seek time is 13ms hdtach

Plus, your WD harddrives will suffocate in an enviornment when you're constantly transfering 3 or 4K files because of it's abysmal seek time in relation to 15K SCSI drives.

--- LOL on a exchange server the complete database is in ram !!!!

For game servers though, it really doesn't matter at all.

All in all, a wonderful system. Quite a ways off in terms of bandwidth in comparison to your PC1200+ machine.. though..

--- my home server is quite slow and I dont know how to fix it (cpu usage 80% all the time for nothing) maybe its the price for realtime file encryption

 
Have a look at Storagereview's database, you can do a head-to-head comparison between drives, a very useful tool.

If you look at their webserver performance database, the top 13 disks are a bunch of SCSI disks(10K and 15K), after those, the first IDE disks start showing up.
The avarge seek time for Fujitsu's and Seagate's 15K disks is ~6 ms, while the WD 1200JB comes in at 13.5 ms.
Write times are fairly similar at 6.3, 6.5, and 14.3 ms times in the same order.

In the ServerSuite benchmarks, the SCSI drives completely annihilate the JB1200(and any other IDE drive for that matter), pretty much constantly scoring between 2 and 3 times as high.

Make some queries in their database, it's extremely useful for getting performance data on various disks, and you know the benches are done by people who know what they're doing.
 
Well, as hardware says for his applications the discs will be idle almost all of the time because, as he said, the data base is small enough to be cached into RAM.

And, I wasn't knocking ABit in a server, just... a first. 😀

But, you see, for desktop usage, WDJB1200's are fine. For server envionrments which don't demand massive amounts of Disc I/O, they're fine.

So basicly you're fine with the WDJB1200's.

As for the high CPU utilization on your server, encryption/decryption is a massive pain in the ass for any CPU.

Why not check the task manager and see what on earth is causing such High CPU utilization? My guess is that it's your encryption program..

Are you using PGP?
 
If this is for an actual organization, I strongly suggest you get more RAM and look at SCSI disks. I'm an Exchange admin, and believe me, if it's a high load server you're going to see some major performance problems. The information store service alone on my machine takes up more than a gig of RAM. The WD drives are nice (I have 6 in RAID 5 in my personal server) but really aren't suited for a production environment. My $.02.
 
hdtach shows me 50mb/s average qith 6% cpu usage

1.5G is the way because 1G ECC is too much $$$ (i think)

As far I see now the WD 8MB rocks overall its so sweet so I will use it over time for all places.

 
even my home pc rocks now
the setup of windows 2000 or xp is a dream with the wd

so for $750 I get 240G usable space with full mirror and 50mb avearge read speed (13.8ms)
not to shabby can someone tell me what I can get with scsi for $750
 
Originally posted by: Hardware
even my home pc rocks now
the setup of windows 2000 or xp is a dream with the wd

so for $750 I get 240G usable space with full mirror and 50mb avearge read speed (13.8ms)
not to shabby can someone tell me what I can get with scsi for $750
Click to expand...

I'm not here to argue the cost savings of IDE over SCSI, but if this is for an organization that you are responsible for sometimes you're less of a bad guy for spending an extra $5000 than saving it but having the server go down constantly. However, I am just now remembering a discussion we had not too long ago about using IDE in a production environment, and I'm not here to pick a fight.
 
--- LOL on a exchange server the complete database is in ram !!!!
Click to expand...

You have a UPS I hope! Not some $120 desktop junk either! A real online enterprise class UPS such as Best, Liebert, Deltech, APC Matrix, etc. 🙂

No ECC memory? Shame on you! Get a flipped bit and lose the boss' calendar in a public folder you're gonna cry wishing you supplied a real solution! :Q

If you want reliability you MUST:

Use a fault tolerant SCSI based disk storage system. RAID 5 minimum requirement. Global hotspare pool of no less than one drive per logical defined array is a must!

Use a mainboard with a chipset that supports registered ECC memory that can fully supports correct and scrub functions!

SMP is very nice as well. After all if a process runs away loading the system to 100% (50% for 2 CPU) you still want snappy response!

What about backups? You certainly need a full backup solution as well.

Your WDC drives aren't SCSI (any RPM) class! No ATA disk interface has ANSI specification. This is a no no in a server (true server) environment.

Forget HD Tach, forget Sandra. You want to use IO Meter to stress the array(s).

I can tell you that if you pull a molex (power) connector off one of those drives while the machine is running that it will stop working. Sure the data (on the other drive) is still there, but it will require YOU to go there and fix it, unlike a true server RAID5 solution! Even without a hot spare, the RAID5 can still run in degraded mode until a replacement drive is plugged in. (which you can do without even shutting the system down!) Try doing THAT with IDE!

Cheers!
 
You have a UPS I hope! Not some $120 desktop junk either! A real online enterprise class UPS such as Best, Liebert, Deltech, APC Matrix, etc. <IMG src="i/expressions/face-icon-small-smile.gif" border=0>
------ hmm its apc and junk in my opinion

No ECC memory? Shame on you! Get a flipped bit and lose the boss' calendar in a public folder you're gonna cry wishing you supplied a real solution! <IMG src="i/expressions/face-icon-small-shocked.gif" border=0>
---- its 1.5G ECC !
If you want reliability you MUST:

Use a fault tolerant SCSI based disk storage system. RAID 5 minimum requirement. Global hotspare pool of no less than one drive per logical defined array is a must!

i------------ts raid 1 so its mirror!


Use a mainboard with a chipset that supports registered ECC memory that can fully supports correct and scrub functions!

------ I845E !

SMP is very nice as well. After all if a process runs away loading the system to 100% (50% for 2 CPU) you still want snappy response!
--- how can you bring 2.4g to 100% ?

What about backups? You certainly need a full backup solution as well.
--- hmm


Your WDC drives aren't SCSI (any RPM) class! No ATA disk interface has ANSI specification. This is a no no in a server (true server) environment.
--- i cant see your point !

Forget HD Tach, forget Sandra. You want to use IO Meter to stress the array(s).
----where can i find io meter?



I can tell you that if you pull a molex (power) connector off one of those drives while the machine is running that it will stop working. Sure the data (on the other drive) is still there, but it will require YOU to go there and fix it, unlike a true server RAID5 solution! Even without a hot spare, the RAID5 can still run in degraded mode until a replacement drive is plugged in. (which you can do without even shutting the system down!) Try doing THAT with IDE!

--- hmm i am not afraid of 10 minutes downtime every 3 years or who often do you thik the ide drives will fail?




 
Sorry about the SCSI comment, it seems to have started off a flame war.

Anyways, hardware, congrats on a nice server.

Sharkeeper, I doubt you really read his original specs. Your No ECC comment sorta damns you in that regard. 😀
 
The performance right now is awesome!


it can pull out 50 megs / s but I need 1000 ethernet and our switch is just 100 !
 
If you need good I/O bandwidth, than theres no escaping it unless you get a e7500 or greater. Nothing under the e7500 delivers huge I/O bandwidth (well depends on how much you need), not even the i860 or 762MPX. Then there are serverworks grand champions that can sustain 4.8GB/sec of I/O bandwidth.

As for a gigabit NIC, I suggset the Intel Pro/1000+, its pretty cheap, about $60 on newegg last I check and is backwards compatible with 100 and 10, plus its cheaper than most server 100 cards and is the cheapest yet best entry level gigabit card. I dont really trust any networking stuff (especially on a server) unless its enterprise level. There are reasons why ISPs buy $5000 Cisco switches instead of $500 Netgear/Linksys/D-link switches, and its not because the Ciscos have more ports.
 
Sharkeeper, I doubt you really read his original specs. Your No ECC comment sorta damns you in that regard.
Click to expand...

Yes but that's useless if it's not properly implemented. (lack of motherboard support for correct+scrub = ECC is useless in a server environment!)

Ditto on the comments regarding enterprise level peripherals. You can certainly find enterprise stuff for cheap on ebay and have great results providing you use common sense. 🙂

Cheers!
 
Disable the write cache in Windows 2000 Device manager for all disk drives, unless you have to do that in the Bios Raid setup in which case, do it there (basically, wherever you have to, and yes, it will slow you down quite a bit).


XADM: Exchange DB and Caching Hard Disks and Controllers

Use of a write-caching disk controller can seriously jeopardize the typically reliable Exchange database data integrity. Significant data corruption can result from a system failure when a write-caching controller without an extremely reliable battery backup is used. This type of controller can compromise the normally reliable Exchange database recovery mechanism....

Even if the server uses an uninterruptible power supply (UPS), this does not guarantee the security of the cached writes. Many types of system failures can occur that a UPS does not address. For example, a memory parity error, an operating system trap, or a hardware glitch that causes a system reset can produce an uncontrolled system interruption. A memory failure in the hardware write cache can also result in the loss of vital log information.

Another possible problem related to a write-caching controller may occur at system shutdown. It is not uncommon, if the operating system is taking a long time to shutdown gracefully, to become impatient and "cycle" the machine manually. When the power to the machine is turned off or the RESET button is pressed before the operating system has shutdown completely, cached writes can be discarded, potentially damaging the database.
Click to expand...
 
I'm 95% sure best practice for exchange is to split the log files and the actual mailbox database, might want to look into that at least for more redundancy.
 
It really depends on the situation.

I am part of the Exchange team here at work and e-mail is a highly visible application. When it goes down you hear it from everyone. Our Exchange servers are all clustered with dedicated front end servers and dedicated virus scanning servers. Storage is handled by SANs and we use a LTO tape library for backup. We also have 24 hour tech support with 2 hour on-site response time for repairs.

If something goes wrong with the system, you don't want to explain to your boss or your boss' boss that you didn't make the system fault tolerant enough. You want the reason that the system is down to be something that was totally out of your control.

Just a word of caution, this is my situation - like I said earlier, it depends on your situation.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top