An oscillation like that can occur with 2 HDDs in RAID 0, particularly if they are older drives,
because the controller ends up waiting for each to spin around to the correct sector,
and then the transfer occurs. So, the pattern can become WAIT-READ-WAIT-READ
where "WAIT" means no transfer occurring and "READ" means actually transferring data.
This pattern can be amplified if the controller is an old PCI-style card,
because the PCI bus itself may be causing additional WAIT time.
Similarly, during WRITE tests, the controller may be programmed to VERIFY each sector
after it's written. That can sync the controller into a pattern that becomes
WRITE-WRITE-WAIT-WAIT-READ-READ.
Moreover, the first WRITE happens to HDD 1, then the second WRITE to HDD 2,
and the third WRITE cannot happen to HDD 1 until HDD 1 has finished the first WRITE.
During this normal sequence, each HDD can end up causing delays for the other,
and vice versa.
I wouldn't worry about it too much: these phenomena result from
the normal operations of low-level firmware in controllers and
in the HDDs themselves. Remember: "IDE" means integrated
drive electronics, so there is a lot of logic already built into
the circuitry that is internal to each HDD, and that logic is
not easily modified.
RAID 1 arrays operate differently with modern RAID controllers,
because the first READ is the one that is used, and the second READ
is usually discarded, because both READS are requesting "mirrored" or
identical data sectors.
Sincerely yours,
/s/ Paul Andrew Mitchell, Inventor and
Webmaster, Supreme Law Library
All Rights Reserved without Prejudice
Facebook
Twitter