Re: Now here's something for FreeBSD advocacy

From: Bill Vermillion (bv_at_wjv.com)
Date: 08/06/04 

Date: Fri, 06 Aug 2004 13:15:01 GMT

In article <2ngk90FljtoU1@uni-berlin.de>,
Jean-Yves Avenard <me@privacy.net> wrote:
>Bill Vermillion wrote:
>
>> The drive shows that buffer to disk speed is 72MB sec.

>> The 15K SCSI drives from Seagate show about 630 - 840 Mbit sec -
>> which is going to come out well over 72MB second - probably
>> 90-100.

>YEs, but that's totally incorrect when it comes to reality. In
>spec what they are referring to is the speed of the interface. In
>reality you will never get over 40MB/s SCSI or not. The only time
>I saw faster write/read speed were on RAID system where it can
>read/write from more than one disk at a time.

>The interesting bit as that most review of SCSI I read showed the
>10,000 rpm disks usually as fast even faster than the 15K ones.

Rotational speed as a measure of performance is not what determines
speed. Back in the days when the Seagate Baracudda was the hot
drive [in more ways that one] I was getting measured performance
from IBM SCSI drives that surpassed the 'cudas.

The reason was they have a superior head design and therefore
were able to achieve a higher data transfer rate with a lower
rotaional speed.

Two main measures of performance are how fast you can get the
data to/from the platters, and then how fast you can deliver
it out the interface.

>About SATA system, the RAID 5 System from 3ware (9500 cards) got
>far better review and benchmark than RAID SCSI system costing 3
>times as much.

What brand of SCSI controllers.

I've seen some pigs out there in the past - and when I used to
handle a lot of SCSI system I made sure that I was not running the
Adaptec devices - with their driver of the month approach.

Without giving a brand to compare with 3ware it's not researchable.

>> The SATA has a maximum disk to computer interface of 150MB while
>> the 15K SCSI is 320MB.

>Again, those are theorical speed.
>ATA was only 100MB, still most disk never get more than 35MB/s

I thought you said 40MB/s above?

And 'most' disks isn't what you compare. It's
fastest-of-the-breed.

>> From my observations everytime [S]ATA gets faster, SCSI gets
>> faster. It may be over-priced in your opinion but it does perform
>> better.

>Well, you would expect so for something costing so much more than
>a SATA system. On the other hand, it's just a value for money.
>Paying 3 times more for something running at best 20% faster is a
>bit of a worry

If you need the 20% then it is important. It's all a value
judgement.

>> So you have to look at the price in this manner - Speed costs
>> money. How fast do you want to go.

>> Drop 3 or more SCSI drives onto that 320MB/second interface in a
>> RAID configuration and you will far exceed the SATA speeds.

>And put 3 or more SATA disk with a good hardware RAID card (BTW, 3ware
>provides a SCSI interface for the SATA system, so you get all the
>advantages of SCSI control comands with cheaper disks)

Bill 

-- 
Bill Vermillion - bv @ wjv . com

Relevant Pages

  • Re: libata in 2.4.24?
    ... As a result SCSI disks are reliable for database ... if the OS driver has not issued a FLUSH CACHE ... I hope you mean the drives don't report completion until the data is on ...
    (Linux-Kernel)
  • Re: File system performance, hardware performance, ext3, 3ware RAID1, etc.
    ... > it isn't necessarilly an artifact of the drives themselves. ... since read-ahead and track buffers are trivial ways to get to true platter ... speeds for a lot of reasonable loads. ... Can somebody fill me in on what modern disks do, ...
    (Linux-Kernel)
  • Re: MicroVAX 3500 questions
    ... DSSI nor SCSI, ... It's an inconvenience but the system will boot without it. ... etc are disks ... The RA7x drives are about the size of an RD54. ...
    (comp.sys.dec)
  • Re: 3B2 Disks
    ... 2 MFM drives on a custom controller. ... But a friend had a 3B2 which had a SCSI interface, ... tablets around for erasing disks. ... But -- the 3B2 was the porting base for unix for some time. ...
    (comp.sys.3b1)
  • SUMMARY SCSI Disk Errors - sense key: not ready
    ... Possible reasons for the disk problems included high humidity, power supply ... problems, dying motor, poor cabling and a problem with the SCSI interface to ... I suspect the drives' motors were slowly dying; I replaced both disks one ...
    (SunManagers)