How fast is your HD?

How fast is your HD?

Post by Sergio Brandan » Sat, 27 Mar 1999 04:00:00



 Hi!

This is my HD:

Quote:> QUANTUM FIREBALL EX6.4A, 6149MB w/418kB Cache, CHS=784/255/63, UDMA

<root> [/]
--> hdparm -t /dev/hda

/dev/hda:
 Timing buffered disk reads:  32 MB in  2.55 seconds =12.55 MB/sec
<root> [/]
-->

 and I would like to replace it with an U2W SCSI one.
 Whow fast is yours?

 Sergio

 
 
 

How fast is your HD?

Post by Mark Hah » Sun, 28 Mar 1999 04:00:00


Quote:>> QUANTUM FIREBALL EX6.4A, 6149MB w/418kB Cache, CHS=784/255/63, UDMA

17.2G Maxtor DM 4320, 14.87 MB/s

 
 
 

How fast is your HD?

Post by morga » Sun, 28 Mar 1999 04:00:00


Seagate 2gig UW:  
 Timing buffered disk reads:  16 MB in  0.49 seconds =32.65 MB/sec

7200 rpm 9 gig IBM UW:
 Timing buffered disk reads:  32 MB in  2.27 seconds =14.10 MB/sec

10k rpm 9 gig IBM UW:
 Timing buffered disk reads:  32 MB in  1.66 seconds =19.28 MB/sec

All are linux with 2940uw cards.  
Why is the seagate so much faster than the IBM drives?  The IBM drives
are
just a copula months old while the seagate is approximately 2 years
old?  
Morgan

 
 
 

How fast is your HD?

Post by Bengt Richt » Sun, 28 Mar 1999 04:00:00



>Seagate 2gig UW:  
> Timing buffered disk reads:  16 MB in  0.49 seconds =32.65 MB/sec

>7200 rpm 9 gig IBM UW:
> Timing buffered disk reads:  32 MB in  2.27 seconds =14.10 MB/sec

>10k rpm 9 gig IBM UW:
> Timing buffered disk reads:  32 MB in  1.66 seconds =19.28 MB/sec

>All are linux with 2940uw cards.  
>Why is the seagate so much faster than the IBM drives?  The IBM drives
>are
>just a copula months old while the seagate is approximately 2 years
>old?  
>Morgan

<disclaimer>I haven't installed my Linux yet, just speaking
OTTOMH from ancient low level experience </disclaimer>

It would be interesting if you could mount the partitions on your
disks raw (read-only!!) and read the first 32 MB of each, and also
the last 32MB of each, reading all sequentially, with nothing else
going on. Also if you would mention how your disks are partitioned.

Otherwise you are partly measuring the state of your file system and
the performance of the file access software. Which distribution/kernel
are you running? (BTW, are all the disks on the same box/controller?!)
And what is the size of your read blocks?

Not having delved yet into Linux diskio and file system code, I'd say
also - depending on how/when things are queued in the file system and
in the scsi controller, it might be interesting to try two threads
with one reading even buffers and the other reading the odd - if you
can arrange direct DMA into your buffers. What I'm getting at is that
a slower disk can appear faster if the software time from completion
of one read to the initiation of the next is more than the rotation
time between blocks, so you lose a whole revolution on a fast disk
but catch the next immediate block on a slower one, and lose no time.
This should be more apparent if you read small blocks on a controller
that doesn't cache tracks by default anyway. An interesting number
is revolutions/bufferful_read actual vs theoretical - remembering
that data/rev is denser at the outer edge of the disk.

Anyway, a lot of factors. I'm sure others will come up with more.

Regards,
Bengt Richter

 
 
 

How fast is your HD?

Post by Keith Oldh » Mon, 29 Mar 1999 04:00:00



> Hi!

>This is my HD:

>> QUANTUM FIREBALL EX6.4A, 6149MB w/418kB Cache, CHS=784/255/63, UDMA

><root> [/]
>--> hdparm -t /dev/hda

>/dev/hda:
> Timing buffered disk reads:  32 MB in  2.55 seconds =12.55 MB/sec
><root> [/]
>-->

> and I would like to replace it with an U2W SCSI one.
> Whow fast is yours?

> Sergio

I was under the impression that the raw data  speed of drives was
nowhere near as fast as 32Mbytes/sec.

I'm quite jealous ( although sceptical) !

hdparm for my PII350 system shows ~10MBytes/sec for a Quantum 6.4 SCSI
narrow (and ~400K /sec  for par. ZIP ext2).

--
Keith Oldham

 
 
 

How fast is your HD?

Post by Thomas Dorri » Tue, 30 Mar 1999 04:00:00



> Seagate 2gig UW:
>  Timing buffered disk reads:  16 MB in  0.49 seconds =32.65 MB/sec
...
> Why is the seagate so much faster than the IBM drives?

Holy smoking hard drive, Bat Man!  How much memory does each system have
and what tool(s) did you use to perform the tests?  I'm sure you know
this already, but most disk IO benchmarks need to be run with a
read/write size twice that of the real memory.  Otherwise, you'll just
be testing the in memory transfer rates of the OS.

For example, on my IBM 9ES 4.5 gig U2 drives, if I specify a 128meg file
size to Bonnie, the results are 40 MB/sec block reads.  However, when I
specify 256meg, I get a more realistic number around 12.5 MB/sec.

Thomas Dorris

 
 
 

How fast is your HD?

Post by Sergio Brandan » Tue, 30 Mar 1999 04:00:00




> > Seagate 2gig UW:
> >  Timing buffered disk reads:  16 MB in  0.49 seconds =32.65 MB/sec
> ...
> > Why is the seagate so much faster than the IBM drives?

> Holy smoking hard drive, Bat Man!  How much memory does each system have
> and what tool(s) did you use to perform the tests?  I'm sure you know
> this already, but most disk IO benchmarks need to be run with a
> read/write size twice that of the real memory.  Otherwise, you'll just
> be testing the in memory transfer rates of the OS.

> For example, on my IBM 9ES 4.5 gig U2 drives, if I specify a 128meg file
> size to Bonnie, the results are 40 MB/sec block reads.  However, when I
> specify 256meg, I get a more realistic number around 12.5 MB/sec.

> Thomas Dorris

 I see. I have infact 128Mb of RAM.
How do you specify 256Mb in hdparm? (I could not find the flag)

 Sergio

 
 
 

How fast is your HD?

Post by Sergio Brandan » Tue, 30 Mar 1999 04:00:00




> > Seagate 2gig UW:
> >  Timing buffered disk reads:  16 MB in  0.49 seconds =32.65 MB/sec
> ...
> > Why is the seagate so much faster than the IBM drives?

> Holy smoking hard drive, Bat Man!  How much memory does each system have
> and what tool(s) did you use to perform the tests?  I'm sure you know
> this already, but most disk IO benchmarks need to be run with a
> read/write size twice that of the real memory.  Otherwise, you'll just
> be testing the in memory transfer rates of the OS.

> For example, on my IBM 9ES 4.5 gig U2 drives, if I specify a 128meg file
> size to Bonnie, the results are 40 MB/sec block reads.  However, when I
> specify 256meg, I get a more realistic number around 12.5 MB/sec.

> Thomas Dorris

 I see. I have infact 128Mb of RAM.
How do you specify 256Mb in hdparm? (I could not find the flag)

 Sergio

 
 
 

How fast is your HD?

Post by Thomas Dorri » Wed, 31 Mar 1999 04:00:00



> I see. I have infact 128Mb of RAM.
> How do you specify 256Mb in hdparm? (I could not find the flag)

Sorry.  Upon further investigation, it seems hdparm takes steps to get
around the buffering issue.  If you're using hdparm, then I suspect the
numbers you get back are fairly accurate.  I'm very curious what Seagate
harddrive Morgan was using when he got these numbers, though:

Quote:> Seagate 2gig UW:
>  Timing buffered disk reads:  16 MB in  0.49 seconds =32.65 MB/sec

ThomasD
 
 
 

How fast is your HD?

Post by Thomas Dorri » Wed, 31 Mar 1999 04:00:00



>  and I would like to replace it with an U2W SCSI one.
>  Whow fast is yours?

Here's hdparm on am IBM 9ES Ultrastar 4.5gig Ultra2(Wide?) SCSI.  It's
16-bit, so I believe that makes it a "wide" device.  Although, if that's
the case, then why is there even an Ultra2 denomination?  Anyway, here's
what I get back from hdparm -t /dev/sda

/dev/sda:
 Timing buffered disk reads:  32 MB in  2.53 seconds =12.65 MB/sec

I'm not particularly impressed by this drive, but as Mike Hahn pointed
out in email, there's probably some other interaction going on in my
system that's * things up.

BTW, here's the results for a couple Western Digital drives I got in my
system as well: Timing buffered disk reads:

hda: WDC AC23200L, 3098MB w/256kB Cache, CHS=787/128/63
 Timing buffered disk reads:  32 MB in  6.43 seconds = 4.98 MB/sec

hdc: WDC AC36400L, 6149MB w/256kB Cache, CHS=13328/15/63
 Timing buffered disk reads:  32 MB in  6.41 seconds = 4.99 MB/sec

Thomas Dorris

 
 
 

How fast is your HD?

Post by karl » Wed, 31 Mar 1999 04:00:00


Quote:> what I get back from hdparm -t /dev/sda

This seems like a cool little util.

thanks
karlo

 
 
 

How fast is your HD?

Post by Sergio Brandan » Wed, 31 Mar 1999 04:00:00


Hello,

thank you for your data. I hope I'll have more and more of them.
I am astonished by the results: how can my EIDE be faster than
an UW SCSI? In particular I was wondering about U2W because
Adaptec claimes it for 80Mb/sec (burst) and I really like to see it.
 The price for those U2W HD is in fact more than three times the
price for an EIDE, to not mention the controller.

 Is somebody using an U2W  HD and wants to report it's speed?

 Sergio

 
 
 

How fast is your HD?

Post by Thomas Dorri » Thu, 01 Apr 1999 04:00:00



> In particular I was wondering about U2W because Adaptec claimes it
> for 80Mb/sec (burst) and I really like to see it.

That's only the bus speed...it doesn't mean the physical device is going
to be that fast, just that the bus is capable of supporting that speed.
I don't believe you'll find any hard drive that can pump that much data
out that fast.  At least not in our combined budgets...

That's where RAID striping comes in (in particular RAID0).  The idea is
to take a collection of fast drives, relatively inexpensive drives,
combine them into what appears to be one big drive, then distribute the
data across them.  On a good SCSI system, under a good multi-threaded
OS, the results should be almost equal to adding up the throughput of
the individual drives (assuming the SCSI bus and memory bus and PCI bus,
etc., can support that rate).

My success has been very limited in that regard, but I barely know what
I'm doing, so that's probably to be expected...  ;-)  That's the theory
anyway. A SCSI bus capable of 80MB/sec transfers is great, but you gotta
have the device(s) to drive it that fast and no drive that I know of
does this all by itself.

ThomasD

 
 
 

How fast is your HD?

Post by David Ript » Sat, 03 Apr 1999 04:00:00



Quote:>thank you for your data. I hope I'll have more and more of them.
>I am astonished by the results: how can my EIDE be faster than
>an UW SCSI? In particular I was wondering about U2W because
>Adaptec claimes it for 80Mb/sec (burst) and I really like to see it.

That's the max that a U2W controller can handle.  No current drive
can saturate a UW controller, let alone U2W.  But it's good for
the controller to be faster than the drive, in case you want to
run multiple drives, or you want to keep using your controller
with the superfast drive you buy in a few years, or your drive
is dumping from its cache.

Quote:> The price for those U2W HD is in fact more than three times the
>price for an EIDE, to not mention the controller.

That's why most people use IDE.

Quote:> Is somebody using an U2W  HD and wants to report it's speed?

No SCSI drives handy, but hdparm says that my laptop's HD
(reported as a Toshiba MK6409MAV - a 6.5 GB 2.5" IDE drive)
sustains 8.00 MB/s, and does 44.44 MB/s out of cache.

Check out www.storagereview.com, for lots of drive benchmark
comparisons.  (No hdparm, though.)

--

spamgard(tm): To email me, put "geek" in your Subject line.

 
 
 

How fast is your HD?

Post by 2349 » Sat, 03 Apr 1999 04:00:00


Any RAID array worth it's salt can swamp a UW/U2W controller. Been there,
done that. The trick is to increase the cache on the controller to math the
maximum demand that the system will place on it. My config:

Mylex DAC960PG w 4 MB/Cache
3 Seagate 4.5 GB Cheetahs (UW)

Burst data speed tests indicate that I get 39.3MB/sec from my array
(configured as a RAID 0)

Granted, this isn't a configuration you run into every day, but with IDE
RAID arriving on the scene, you're going to see a LOT more of these.

Carlos




> >thank you for your data. I hope I'll have more and more of them.
> >I am astonished by the results: how can my EIDE be faster than
> >an UW SCSI? In particular I was wondering about U2W because
> >Adaptec claimes it for 80Mb/sec (burst) and I really like to see it.

> That's the max that a U2W controller can handle.  No current drive
> can saturate a UW controller, let alone U2W.  But it's good for
> the controller to be faster than the drive, in case you want to
> run multiple drives, or you want to keep using your controller
> with the superfast drive you buy in a few years, or your drive
> is dumping from its cache.

> > The price for those U2W HD is in fact more than three times the
> >price for an EIDE, to not mention the controller.

> That's why most people use IDE.

> > Is somebody using an U2W  HD and wants to report it's speed?

> No SCSI drives handy, but hdparm says that my laptop's HD
> (reported as a Toshiba MK6409MAV - a 6.5 GB 2.5" IDE drive)
> sustains 8.00 MB/s, and does 44.44 MB/s out of cache.

> Check out www.storagereview.com, for lots of drive benchmark
> comparisons.  (No hdparm, though.)

> --

> spamgard(tm): To email me, put "geek" in your Subject line.

 
 
 

1. New/fast HD with old/slow HD

I want to buy an additional IDE HD for my Redhat 7.3
machine.  A friend advised me that if I add the
older drive back with the fast new drive, that
the new drive will only perform as fast as the
older drive.....something about "lowest common
denominator" the oldest/slowest drive will pull
down the performance of the new drive to its level.

Is this true?

How does the new IDE drive get choked by the old/slow drive?

2. Processor Required for Firewall

3. New/fast IDE HD with older HD

4. Parsing strings via shell vars?

5. How fast am I connected (PPP)?

6. Modifying Boot Disk

7. <<NEED HELP FAST, I am a newbie>>

8. Writing cd-roms on Solaris.

9. How FAST am I going? (waddle waddle)

10. Am I the only person around with a 40 gig HD?

11. Why am I plagued with these HD problems?

12. Adaptec 2940W with Quantum 4.3GB fast-wide HD

13. FAST small HD for swap space