Adaptec
DuraStor Review
Review by
Raoul Callaghan
Posted: Jan. 16th, 2003
I had the
pleasure of evaluating one of Adaptec's Hardware RAID systems this week.
I received a rackmountable DuraStor 6200S Hardware Controller and a 12 bay
312R SCSI Chassis having four 147GB SCSI Ultra320 Drives with 8Mb cache to
play with. (It also came with two 65GB Ultra160 Drives, but I didn't worry
about them when I saw the Ultra320's)
The 312R Chassis and Controller are only Ultra160 capable so the drives are
a bit excessive.
Nonetheless, the unit retails around $20,000AUD without any drives so another
$10,000AUD would get you an entry level RAID.
This is a
very brief review as I don't have the time to write this properly,
but should give readers an idea of the system's capabilities.
Where to
start:
Very Impressive and Very Noisy comes to mind when describing the
DuraStor. Compared to the Xserve the DuraStor's Controller bearings whine
a lot more and at a higher pitch. the unit has great literature to get the
unit up and running quickly and the quality of the High Density 68pin SCSI
cables are good also.
The unit
can be configured in many ways ranging from multiple drive chassis connected
to the controller with dual SCSI channels connected to a server, to a single
drive chassis using dual channels to the controller that can support two servers
for server redundancy purposes. As I only had one Atto adapter and two HD68
pin cables I could not connect the unit to the Xserve using both channels
of the Atto PCI card nor our spare Xserve, so the benchmarking scores I attained
for RAID1 could be improved.
The setup I opted for, or should I say was left with was a single channel
from the Drive Chassis to the RAID Controller, and a single Channel from the
RAID Controller to the Xserve. Ideally, I needed 6 HD68pin cables to test
this unit thoroughly. I wanted to test two Xserves with this device acting
in IP failover mode to see how receptive the unit would be to a server failure.
Unfortunately I could only get one Atto PCI adapter so I had to leave this
out of testing.
I used Xbench
to attain a quick score for the setup. Additionally, I have four drives in
the Xserve, two 120GB drives configured as RAID0, and two 60GB drives configured
as RAID1 to compare with the DuraStor.
Below are the xbench scores attained for each RAID Array.
| Xbench
Results |
| Two
120GB Drives ATA100
RAID1
(Mirror) |
Two
60GB Drives ATA100
RAID0
(Stripe) |
Four
120GB Ultra320 SCSI
RAID0
(Stripe) |
| Results
117.39
System Info
Xbench Version 1.0b3
System Version 10.2.2
Physical RAM 1024 MB
Processor PowerPC,G4@0 [1000 MHz]
L1 Cache 32K (instruction), 32K (data)
Bus Frequency 133 MHz
Processor PowerPC,G4@1 [1000 MHz]
L1 Cache 32K (instruction), 32K (data)
Bus Frequency 133 MHz
Disk Test 117.39
Sequential 124.73
Uncached Write
121.85 45.36
MB/sec [4K blocks]
126.00 44.13 MB/sec [256K blocks]
Uncached Read
129.17 19.09 MB/sec [4K blocks]
121.89 45.70 MB/sec [256K blocks]
Random 110.05
Uncached Write
133.03 2.22
MB/sec [4K blocks]
118.30 24.58 MB/sec [256K blocks]
Uncached Read
76.35 0.82 MB/sec [4K blocks]
112.54 21.66 MB/sec [256K blocks] |
Results 193.59
System Info
Xbench Version 1.0b3
System Version 10.2.2
Physical RAM 1024 MB
Processor PowerPC,G4@0 [1000 MHz]
L1 Cache 32K (instruction), 32K (data)
Bus Frequency 133 MHz
Processor PowerPC,G4@1 [1000 MHz]
L1 Cache 32K (instruction), 32K (data)
Bus Frequency 133 MHz
Disk Test 193.59
Sequential 210.02
Uncached Write
210.60 78.41 MB/sec [4K blocks]
258.78 90.64 MB/sec [256K blocks]
Uncached Read
130.53 19.29 MB/sec [4K blocks]
240.18 90.06 MB/sec [256K blocks]
Random 177.15
Uncached Write
328.39 5.48 MB/sec [4K blocks]
169.50 35.22 MB/sec [256K blocks]
Uncached Read
87.79 0.95 MB/sec [4K blocks]
122.94 23.66 MB/sec [256K blocks] |
Results
311.22
System Info
Xbench Version 1.0b3
System Version 10.2.2
Physical RAM 1024 MB
Processor PowerPC,G4@0 [1000 MHz]
L1 Cache 32K (instruction), 32K (data)
Bus Frequency 133 MHz
Processor PowerPC,G4@1 [1000 MHz]
L1 Cache 32K (instruction), 32K (data)
Bus Frequency 133 MHz
Disk Test 311.22
Sequential 150.58
Uncached Write
105.15 39.15 MB/sec [4K blocks]
162.61 56.96 MB/sec [256K blocks]
Uncached Read
82.61 12.21 MB/sec [4K blocks]
251.95 94.47 MB/sec [256K blocks]
Random 471.86
Uncached Write
491.62 8.21 MB/sec [4K blocks]
237.64 49.39 MB/sec [256K blocks]
Uncached Read
829.69 8.96 MB/sec [4K blocks]
328.48 63.23 MB/sec [256K blocks] |
The results
speak for themselves really. I did connect the Drive Chassis direct to the
Xserve and used software to stripe the four drives. Software RAID'ing drives
in OSX was noticably slower when I was conducting real world tests with
large (20GB) files. I wish I had more time as I would re-conduct an xbench score
for this setup to see the CPU overhead on data transfer.
Here are some jpeg pictures of the System and components.
Link |
Description |
|
Xserve, DuraStor
Controller & DuraStor Chassis on Desktop. |
|
MacOSX Server Desktop
showing each RAID array. |
|
Xserve used with
iMovie to capture images to drive. |
|
DuraStor Chassis
with Controller stacked on top. |
|
DuraStor Logo,
showing screws (half twist required only to remove). |
|
Size comparison
of Durastor with Xserve. |
|
LCD Interface. |
|
Live diagnostic
feedback on LCD. |
|
Temperature &
Date monitoring. |
|
Controller interface
panel removed. |
|
Controller PSU1
Ejected. |
|
Drive Chassis panel
removed also. |
|
Controller and
Chassis pivoted. |
|
Front Chassis Circuitry. |
|
ATTO Dual Channel
PCI adaptor in Xserve. |
|
Rear view of system
and cabling. |
|
Controller I/O
inputs and outputs. |
|
Controller with
spare channel not in use. |
|
I/O for redundant
Xserve option, int/ext is for LCD or software monitoring via serial
(PC only software). |
|
Dip switches for
single or dual channel configuration, auto spin up of drives etc. |
|
Redundant power
supplies for zero down time (unlike an Xserve) for Controller and Chassis. |
|
Ejected 3.5"
10,000rpm drives. |
|
Drive shelves are
very firm to insert and remove drives which is good to prevent accidentals. |
|
Note: no SCSI ID
terminationon the drive as this is controlled by the Chassis in conjunction
with the Controller. |
|
80pin Data and
voltage interface for the drive. |
|
Fujitsu label of
drive. 1,200,000 Mean-Time-Between-Failure (MTBF). |
Transferring data
at around 100MB/sec sustained certainly boots the Xserve a lot faster.
When creating the array, the controller gives you a chunk size option in kilobytes
to best suit your data requirements. 32KB for lots of small files etc.
Considering that the Xserve needs to be shutdown if the PSU fails or is about
to fail, the DuraStor is ideal as the Controller can support a redundant Xserve.
The Durastor is a very nice addition to any corporation who is very serious
about redundancy and 100% uptime.
We also have an Arena Tower here which is also an Ultra160 interface unit
but uses IDE drives. Unfortunately I could not test this item as it is listed
here as stock.
Here is a link to the drive mechanism drive
mechanism for those who are interested, and also for the 312
Chassis and the Controller.
Might I say that I
was rather sad when it was time to pack this up and not keep it with our Xserves.
-Raoul Callaghan
|
