Dual Opteron – better SATA cables for the Raptors
When you need cables, you often need them right now. Fast shipping is great.
Overnight is fantastic. It’s even better when the vendor calls back and offers you
a slightly more expensive cable, for the same price, because the one you asked
for is out of stock and will take a few days to order in from the States.
This is the service I got from MyCableShop.
I found them via Google, and they had the cables I wanted, at a much cheaper price
than my local computer shop could source them. I wanted these
12″ SATA cables, and
with shipping, they’d be about $5 each. Just what I need to solve
a messy
cabling issue that arose from using 38″ cables in a small 2U case.
Not only did MyCableShop call me to recommand another cable shortly after
I placed my order, but the cables arrived the day after I ordered them.
I’m impressed.
The new cable layout
The new cable layout is much improved. The 12″ cables fit nicely between the card
and the backplane without any excess blocking the fans.
You will notice that I used two red 18″ cables. The 12″ cables are great for the
6 ports at the back of the 3Ware card. But for ports 6 and 7 at the front of the
card, I needed a slightly longer cable. I knew I had several of these cables
somewhere in the house, but I could find only one. I headed to my local computer
shop (Over The Top Computing) where
Chris gave me one he had in the back.
These 18″ cables were slightly too long, so I routed the excess to the side of
the RAID card. Before I ordered the smaller cables, someone mentioned I could do
this for the longer cables, but there was so much excess cable (nearly two feet
extra per cable) I thought the right thing to do was get new cables. I’m glad
I did.
There is one previous photo of the Raptors
but it shows only 6 of the 8 drives. The photo below shows all 8 of the Raptors
in their cages.
Are the new cables any slower?
Cables are cables. They should not matter. But I wanted to make sure. So
I did a very simple test. How fast is a disk copy with the old cables?
[dan@opti:~] $ time cp -r /usr/ports . real 1m49.440s user 0m0.454s sys 0m10.352s [dan@opti:~] $ time rm -rf ports real 0m16.398s user 0m0.157s sys 0m2.008s [dan@opti:~] $ time cp -r /usr/ports . real 1m47.370s user 0m0.352s sys 0m10.593s [dan@opti:~] $ time rm -rf ports real 0m17.365s user 0m0.137s sys 0m2.084s [dan@opti:~] $ time cp -r /usr/ports . real 1m47.481s user 0m0.439s sys 0m10.541s [dan@opti:~] $ time rm -rf ports real 0m16.547s user 0m0.133s sys 0m2.048s [dan@opti:~] $ time cp -r /usr/ports . real 1m46.543s user 0m0.479s sys 0m10.588s [dan@opti:~] $ time rm -rf ports real 0m17.276s user 0m0.105s sys 0m2.125s [dan@opti:~] $
OK, those are pretty consistent times.
Then I change the cables, and nearly die:
$ time cp -r /usr/ports . real 2m30.969s user 0m0.406s sys 0m10.693s
*COUGH*
That’s nearly 42% longer, with the only difference being the cables!
This can’t be right. And it isn’t. Let’s try again:
$ time cp -r /usr/ports . real 1m45.426s user 0m0.438s sys 0m10.645s [dan@opti:~] $ time cp -r /usr/ports . real 1m45.426s user 0m0.438s sys 0m10.645s [dan@opti:~] $ time rm -rf ports real 0m16.939s user 0m0.136s sys 0m2.133s [dan@opti:~] $ time cp -r /usr/ports . real 1m45.467s user 0m0.575s sys 0m10.710s [dan@opti:~] $ time rm -rf ports real 0m19.831s user 0m0.134s sys 0m2.300s [dan@opti:~] $
OK, that’s good. It was just a caching issue. Now the times are
back to what they were.
Monitoring fan speed/CPU temperature
I had planned to monitor CPU temperature and fan speed using something
such as healthd
or mbmon. Unfortunately,
neither port is of any use with my m/b. It uses the ADM1026 sensor chip
which the current version of mbmon does not support.
Unfortunately, it looks quite different from the already supported ADM1025/1027.
That’s a shame. It seems a pity not to monitor such vital readings.
Next stop!
Now that I have the right cables, the RAID has been set up, and NetSaint is
monitoring the array for me, it’s time to start install the software for my
services. I plan to do that over the next week.