Ok enough enough enough.

Sustained burst performance figures are WORTHLESS. WORTHLESS. WORTHLESS. 
Nobody uses or needs 100MB/sec sustained transfer rate.

What is important is SEEKS PER SECOND.

Now consider: 4x ATA100 drives on an ATA100 controller. Each drive can do
about 100 I/O's per second (1000ms, 10ms random access speed). POP QUIZ:  
running RAID0, how many seeks per second can you do with 4 ATA100 drives?

Is it 400? NO.
It's 100. Because only one IDE drive can be active on the chain.

Now consider the same situation with SCSI. In THAT case, it IS 400 seeks 
per second. Because the SCSI bus is designed for contention, and seek 
requests to individual drives can be interleaved.

Now consider a chain with 15 drives on it. You can't even do this with 
IDE.

Now why is I/O's per second important? simple. In general, on a UNIX
system, disk access is done ONE BLOCK AT A TIME. You almost NEVER see a
sustained burst disk access (unless you're doing "dd if=/dev/cdrom
of=myfile" or something).

So: one block on a UNIX filesystem is typically 8k. If you can do 100
I/O's per second, that's (100 x 8k) = 800kbytes per second! WOW!! amazing
throughput!!

For applications like mail servers and database servers, the I/O's per 
second is MUCH more important than any contrived figure of raw bandwidth.

All the people on this list who love to parrot "25MB/second burst rate for 
an ATA100 drive!!!" should go read Brian Wong's "Configuration and 
Capacity Planning for Solaris Servers" as well as Adrian Cockcroft's 
articles on the same topic. VERY illuminating.

Andre, you're an SCSA. I'm ashamed of you.  ;)


---
Orlando Andico <[EMAIL PROTECTED]>
Mosaic Communications, Inc.


_
Philippine Linux Users Group. Web site and archives at http://plug.linux.org.ph
To leave: send "unsubscribe" in the body to [EMAIL PROTECTED]

To subscribe to the Linux Newbies' List: send "subscribe" in the body to 
[EMAIL PROTECTED]

Reply via email to