I was just informed that RackShack uses 40GB Maxtor 5400 RPM hard drives in their AMD Duron and XP 1700 systems, while Tranxactglobal.com / DV2 uses Western Digital 7200 RPM Hard Drives. I believe both the RackShack 5400 RPM drives and the TranxactGlobal 7200 RPM drives are connected via ATA/100 and have a 2MB buffer.

How much difference is there between a 7200 RPM Western Digital and a 5400 RPM Maxtor (http://www.maxtor.com/products/DiamondMax/DiamondMax/QuickSpecs/42088.htm) drive for hosting?

I was really disappointed that RackShack doesn't offer a faster hard drive as an option since I'm sure they could for an extra $50 setup fee for something, and I *think* that it would make a difference.

Big difference. I'd hate to host off of a 7200 RPM drive (instead of a 10k SCSI one) -- I can't imagine how bad a 5400 RPM would perform on a loaded server. I'd go for the 7200 without thinking. Performance matters too much to save a buck or two.

I see on pricewatch that there is only a difference of a few dollars between the Maxtor 5400 drives and the Western Digital 7200 drives, yet the 7200 drives have a 9 MS seek time where as the Maxtor drives are at 12 MS seek time, which would seem significantly slower :(

Pretty much every 'hardware review' website indicates that the performance of 7200RPM drives is much better than 5400RPM's.

When you look at many of these budget web 'servers', they can be upgraded to ample RAM, the CPU's are nearly the fastest on the market ... so often you'll find that the bottleneck in your system's performance will be the HDD's - if you can't afford SCSI, then you'll want to get the fastest performing drives you can afford.

i thought rackshack had 7200s but i checked up and turns out the 20 gigs are 4K020H1 which are 5400 but can manage a reasonable rate.

[root@www /sbin]# hdparm -t /dev/hda

/dev/hda:
Timing buffered disk reads: 64 MB in 1.93 seconds = 33.16 MB/sec

[root@www /sbin]# hdparm -i /dev/hda

/dev/hda:

Model=MAXTOR 4K020H1, FwRev=A08.1500, SerialNo=371120721359
Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs }
RawCHS=16383/16/63, TrkSize=32256, SectSize=21298, ECCbytes=4
BuffType=DualPortCache, BuffSize=2000kB, MaxMultSect=16, MultSect=16
CurCHS=16383/16/63, CurSects=-66060037, LBA=yes, LBAsects=39876480
IORDY=on/off, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120}
PIO modes: pio0 pio1 pio2 pio3 pio4
DMA modes: mdma0 mdma1 mdma2 udma0 udma1 udma2 udma3 udma4 *udma5

Cant speak for any of the other drives they offer only the one in this server. Its not great but its not shocking. anyone got a raq there and willing to post stats for those drives.

Originally posted by Wolfy
Pretty much every 'hardware review' website indicates that the performance of 7200RPM drives is much better than 5400RPM's.

When you look at many of these budget web 'servers', they can be upgraded to ample RAM, the CPU's are nearly the fastest on the market ... so often you'll find that the bottleneck in your system's performance will be the HDD's - if you can't afford SCSI, then you'll want to get the fastest performing drives you can afford.

I can't agree more! 7200rpm drives are definitely faster than the 5400rpm drives. I'm also a bit disappointed to see Rackshack wouldn't spend a few more dollars (out of the high setup fee) to use 7200rpm drive in their servers. The performance should be noticeable under heavy server load.

Raq 4r (RAID):
/dev/hda:
Timing buffered disk reads: 64 MB in 3.67 seconds = 17.44 MB/sec

/dev/hda:

Model=ST320413A, FwRev=3.05, SerialNo=7ED1JXGT
Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs RotSpdTol>.5% }
RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=0
BuffType=0(?), BuffSize=1024kB, MaxMultSect=16, MultSect=16
DblWordIO=no, OldPIO=2, DMA=yes, OldDMA=2
CurCHS=16383/16/63, CurSects=-66060037, LBA=yes, LBAsects=39102336
tDMA={min:120,rec:120}, DMA modes: mword0 mword1 mword2
IORDY=on/off, tPIO={min:240,w/IORDY:120}, PIO modes: mode3 mode4
UDMA modes: mode0 mode1 *mode2 mode3 mode4 mode5


Raq4:
/dev/hda:
Timing buffered disk reads: 64 MB in 3.56 seconds = 17.98 MB/sec

Model=ST330621A, FwRev=3.05, SerialNo=7EE04RR6
Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs RotSpdTol>.5% }
RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=0
BuffType=0(?), BuffSize=1024kB, MaxMultSect=16, MultSect=16
DblWordIO=no, OldPIO=2, DMA=yes, OldDMA=2
CurCHS=16383/16/63, CurSects=-66060037, LBA=yes, LBAsects=58633344
tDMA={min:120,rec:120}, DMA modes: mword0 mword1 mword2
IORDY=on/off, tPIO={min:240,w/IORDY:120}, PIO modes: mode3 mode4
UDMA modes: mode0 mode1 *mode2 mode3 mode4 mode5


What is the command for timing SCSI hd's? I'll post the results then.

Anyway, here's a preview of the scsi drive:

<<REMOVED. OOps, that wasn't the drive>>

Maybe i can get some tuning done?

now im not going to be held responsible for this and you have to do a little research but the following command could improve your performance.
If your drives are capable of udma 5 i.e. ata100 then the following command may speed it up considerably

hdparm -c1 -X69 -u1 -d1 /dev/hda

have a look on the web for the drive, or even give cobalt a ring. i reckon theres also gotta be a few people on the board way more clued up on this then me so they might want to comment :)

guess for scsi drives you just change the hda bit to the id of the scsi drive. thanks for the info on raqs tho, often wondered what they could do. not too bad for a amd k62. seagate ide drives hmmm. prefer maxtors myself.

Maybe i can get some tuning done?


On SCSI, I don't think so ... SCSI drives need no tuning. CMIIW.

cheers,
:beer:

panopticon, heres the lowdown. 5400 and 7200 drives have the same transfer speeds (so nothing moves faster), but 7200 drives typically have a 7-9 ms seek time, whereas 5400 rpm drives have a 11-14ms seek time. The 7200 drives seek faster, it's similar to the difference between a 56x cdrom and a 44x cdrom, ever notice they perform the same? Well they pretty much do, just one seeks faster, and it's only noticable if you have millions of tiny files.

The only reason i'd consider using 7200 drives would be if 1. they were scsi 7200 (which transfer much faster), or 2, if i knew there was a major quality difference (which in some makes there is, but i've never had a problem with maxtor or wd 5400's). If you want the performance, look for 10k and 15k scsi drives, or if you're cheap, raid 0, or 1+0 with 2-4 ide drives (comparable to scsi IMHO).

Hope that helps :)

and it's only noticable if you have millions of tiny files.
Millions of tiny files - like the dozens of jpg and gif files which make up the dozens of html, php, and data files which are delivered when lots of people load a web site?

Because a website is made of so many little files, I would think the seek time would come into play here much more than on a desktop system. 100 mbps is great for both drives, but with the thousands of >1kb - 50 kb files called every hour, I would think the seek times might really add up.

Not likely, you'll need litterally millions of them to make it add up to much. If you want performance, spend the extra $200 and get a second hdd and a raid card to do raid level 0 (speading data over two drives, 2x the throughput, etc). 3ms (in the grand play of things), would take over 330 files to make a 1 second difference in access time. :)

Hey,
This doesn't much relate to the original question but it can't hurt. IDE raid in Linux works very nicely. It's stable, reliable, and as demonstrated, very fast.

RAID 0 2x40 gb IBM 60gxp on hpt370 (kt7a-raid) under heavy load:

Timing buffered disk reads: 64 MB in 0.87 seconds = 73.56 MB/sec

Good god.. the length some of you guys go to save $3-$10 (the difference between many 5400 and 7200 drives) is beyond me. :) (and surprising, too -- i generally concur with Myles and tend to value his opinion highly on most other issues)

Ever hear of 'best bang for the buck'? 5400 ain't it.

Although, depending on the drive, it might not be the most horrible thing in the world, either (I'll get to that in a second). It's just that going 5400, when better drives are available for very little more money, isn't exactly smart. The hard drive is the slowest part of the system. When you go on a walk with a dog on a leash, the harddrive is the kind of dog that will sit there and resist you ever step of the way. And it's a big dog. It WILL slow your journey down...

Throw a 5400 in your own home system, then a 7200. Compare notes. Regardless of theory and ms calculations, 7200 is NOTICABLY faster for just ONE user, isn't it?

Speeding the hard drive up is the first and cheapest thing you can do to make every single part of that system faster.

$3-10 more is not the difference between slow IDE and SCSI (which could cost "a ton" more), but it's the difference between slow IDE and fast IDE. To not spend it is ultimately foolish, a misappropriation of funds (what better do you plan to divert that $3-$10 to????), especially on a server... where TONS of people are using that system, and disparate scripts and daemons are SEEKING files every which way and place on the hard drive.

Now, a quick check of the Rackshack forums show that some of their servers use the 4D040H2 from Maxtor. What a mouthful. 4D040H2. That's their 40gig d540x drive. Probably the fastest of the 5400rpm drives, so in this case, the $3-$10 price cutting measure WON'T kill you. (the brain virus that caused the people who speced this drive, though.. now that might kill you... being penny wise and pound foolish tends to kill businesses equally well as personal lives)

Check out the StorageReview write up of this drive.

http://www.storagereview.com/welcome.pl/articles/200112/200112204G160J8_1.html

Bear in mind that StorageReview went ahead and did what many hosts won't. They turned off Accoustic Management, which makes the seek faster, at the expense of being a little louder. Many drives have this on by default. The drive you get in your server might have this on -- saving the non-existant ears of nearby servers a bit, and slowing you down. They also turned off write verification. I don't know if this comes on as default on the Maxtor drives, but this would speed things up, too -- maybe even double, depending upon the method of write verification the drive used (or maybe no gains if it verifies near simultaneously with the write).

Not just that, but they tested a WHOPPER version of this drive. A 160 gig monster. I think they use 40 gig platters, so that makes for a 4 platter drive (and the 40 gigger, a 1 platter drive). And, if I remember correctly, the bigger drives are these days, the faster they tend to be... because they can try to read and write from all platters simultaneously. 4 platters being faster than 1. It's like a simple form of Raid 0 inside the drive. Now, I'm not saying you'll get a 400% gain due this drive having 4 platters. In fact, I can't say what the gain will be. The 40 might only be 10% slower, in fact. It might be LIKE a simple form of raid inside the drive, but certainly don't get raid like gains -- just mild (but still noticable!) ones.

In any case, the 160gig d540x turned out to be about 18% slower than it's 80 gig (dual platter) d740x (7200rpm) cousin in a web server benchmark.

If you look at the logs of a web server running vbulletin, you'll definitely see what a drag the hard drive is (at any speed) during busy times. An 18% improvement is a welcome one, especially at $3-$10.

Oh, sure, more money can get you a nice SCSI RAID setup, and give you mega-performance for mega-bucks. But 18% $3-$10 is pretty darned compelling, too.

Keep in mind that not all RS servers even use the d540x. When searching the archives, I saw some other brands mentioned. So depending on what they bought and what's available, you might not even be lucky enough to get this particular model of drive.

Ultimately, be careful of being penny wise, pound foolish -- and be careful of the people you work with being that way, too. Your life is short enough without that kind of unwisdom, within yourself or those you work/associate with, reducing the quality of that already short life.

i generally concur with Myles and tend to value his opinion highly on most other issues

:blush: thanks :).

Thats a good point that i forgot to note, the multiple users analysing the system. Usually i've found it to be about a $20 difference, which is about 25% of the drives cost, so it has made a difference, the $20 can be diverted to get higher quality motherboards, ram, etc. I might start using 7200 drives later, but not as of yet, i love my caviars :D.

you eat those with capers or what? :)

hey, myles, you're in canada, aren'tchya? i don't remember the computer deals being all that good. then again, i don't know the best sources up there. but it seems like quite a few of my canadian friends end up ordering from the u.s.

I get many prices for comparison purposes from pricewatch.com. For ordering purposes, I usually play it safe and go to NewEgg.com. They're very competitive on everything, have a nice amazon-like customer review system, use cheap FedEx shipping, and are just very good all around. (I've gone to very slightly cheaper places and been unhappy/scammed) Price difference there between the d540x and d740x 40 gig is $10. Price difference on pricewatch from multiple vendors even lower. (and ever changing)

speaking of server hds... storagereview.com, right on their front page where they talk about the crackass IBM 120gxp drives '8 hour on' specification, they mention the Maxtor (Quantum designed or influenced) d740s as the ideal IDE server drive now. So Rackshack is 5/7th of the way there to using a good (or the best, if you like the storagereview guys) drive, you could say. :)

jmars, for our facility i order almost everything through pricewatch. The deals up here in canada are just plain bad with the exception of my wholesaler for CRT and LCD monitors. Theres a complete lack of competition, because nobody buys computer components online up here, because nobody will ship. I grabbed everything i could from pricewatch, i even attempted an order from newegg though, but they made my "crap list" because they refused to call my cc company and confirm my billing/shipping address because it was a canadian credit card. If they want to throw away a huge order because my billing address is in canada, their loss, but what a waste of time (when you've got 75% of vendors who do that). Around here i usually go to summit-direct.com (i think thats summits new webpage), and that reflects the pricing we canadians get (remember our currency is 1:1.6 now though =\).