[luau] INFO: Heavy duty storage needs
MonMotha
monmotha at indy.rr.com
Thu May 2 19:40:20 PDT 2002
I'm glad someone liked my explanation :)
I recently built a nice dualie AMD 1.4GHz, 2GB DDR, with 2xU160 36GB
15kRPM hard drives. It's (among other things) a proxy server and
mailer. Proxying, especially when also doing mail, involves some
bandwidth (but not excessive, it's only a 1gbit uplink to the LAN), but
it also involves a lot of seeking. The tagged command queuing and
reduced seek times of the SCSI drives are a big advantage on this puppy.
--MonMotha
R. Scott Belford wrote:
> I have searched for such a concise and practical explanation of why SCSI
> is accepted as the better storage solution for high-intensive seeks.
> Now I understand. Thanks for explaining this so clearly.
>
> On Thursday, May 2, 2002, at 12:46 PM, MonMotha wrote:
>
>> This is true. In terms of raw I/O speed, IDE drives have caught up
>> with SCSI. The reason? The bus is no longer the bottleneck. Even
>> the best of SCSI drives would have trouble saturating a 100MB/sec bus
>> (though SCSI is 160). If all you need is raw I/O speed (which is a
>> good chunk of what your average single user desktop will be working
>> with, especially if they're doing a lot of multimedia), IDE drives are
>> a great choice (especialy combined with a software RAID solution, such
>> as the Linux software RAID or some of the software "RAID" cards that
>> are now out for the 'dozers).
>>
>> SCSI shows it's strength in Tagged Command Queueing. Unlike IDE
>> devices, SCSI drives can work on more than one command at the same
>> time, queueing them up in the best way possible. For example, an IDE
>> system where a person wants to simultaneously (quicker than a single
>> seek on the drive) pull information from 10 places on the disk will
>> have to do them in the order the system says. With SCSI, the system
>> can send all 10 (or usually 9) requests in quick succession, and the
>> drive will service them in the best way possible, minimizing redundant
>> seeking, serving out of cache when possible (often even during a seek
>> for the next command in the queue), etc. In multiuser environments,
>> this can give SCSI a HUGE advantage over IDE. SCSI drives also tend
>> to have lower seek times, often due to smaller platter sizes rotating
>> at significantly faster speeds (15,000 RPM vs. 7200 or even 5400 RPM).
>> This is also the reason SCSI drives tend to have smaller capacity as
>> compared to IDE drives while still being expensive. The same
>> "technology" has to be used to pack all those bits into a small space,
>> but the overall space is smaller (many SCSI drives could almost fit
>> their platters into 2.5" laptop HDD enclosures, see fujisu specs on
>> platter size). These lower seek times also help performance in
>> multiuser environments.
>>
>> Also, though this is not as much of a deal with UDMA IDE, IDE uses the
>> CPU for some of it's processing. The same things are done on SCSI on
>> the dedicated controller and onboard the drive itself.
>>
>> Conclusion: Don't assume SCSI is better, but don't assume raw I/O
>> bandwidth is the only measure of a hard drive.
>>
>> --MonMotha
>>
>> R. Scott Belford wrote:
>>
>>> If you intend to record most or all of the traffic moving over your
>>> network, you need to spend as much time thinking about your disk
>>> subsystem as your processor and Ethernet card. Last year Sandstorm
>>> spent several months comparing IDE drives with the UDMA100 interface
>>> to SCSI LVD-160 drives. We also explored a variety of RAID systems.
>>> The conclusion: today's IDE drives are significantly faster than SCSI
>>> drives costing two or three times more per gigabyte stored.
>>> This is not the result we were expecting, and it goes directly
>>> against the conventional wisdom that says SCSI is inherently better
>>> than IDE. Nevertheless, it does seem to be the ugly truth, at least
>>> for straightforward read/write tests in a single-user environment.
>>> Although we saw the highest performance with a hardware-based RAID 5
>>> system manufactured by _Advanced Computer & Network Corporation_, we
>>> saw nearly the same performance with a RAID 5 system based on the
>>> _3Ware Escalade 7000_ RAID controller.
>>> from an article about network data capture at
>>> http://www.oreillynet.com/lpt/a//network/2002/04/26/nettap.html
>>
>>
>>
>> _______________________________________________
>> LUAU mailing list
>> LUAU at videl.ics.hawaii.edu
>> http://videl.ics.hawaii.edu/mailman/listinfo/luau
>>
>
> _______________________________________________
> LUAU mailing list
> LUAU at videl.ics.hawaii.edu
> http://videl.ics.hawaii.edu/mailman/listinfo/luau
>
More information about the LUAU
mailing list