|
Q&A
listing:
Q1:
What is a CDE
and what corrective action is recommended? . . . . .
A CDE
is a Controller Detected Error (CDE) and is characterized as a firmware
inconsistency. A CDE is a recoverable error and should be retried
by the host.
There
are two types of CDE;
- System
Fault
- Exception
Fault
Indications
that a CDE has occurred are;
- Right
most DOC character blinking - if CDECURS is enabled via the DOC
- SCSI
sense indicates 04-44/80-FF - or 06-44/80-36 (unit attention due
to CDE reset)
- DOC
flashes SELF-RECOVERY
A CDE
history log is maintained in NVRAM on the WOODI. The log
is 50 bytes deep. Each CDE entry is one word (2 bytes) in length.
So 25 CDEs will fit in the log. Each entry is the CDE reference
number in hex and can be looked up in the CDE lookup page. The CDE
log index points to the last word position recorded.
Example
output:
CDE
Log Index : 04
0000: 02470350035002470000000000000000
0010: 00000000000000000000000000000000
0020: 00000000000000000000000000000000
0030: 00000000000000000000000000000000
0040: 00000000000000000000000000000000
0050: 00000000000000000000000000000000
0060: 00000000
Q2:
A 6000 drive has intermittent failures, how can I tell if it is
running within specifications?. . . . .
See
the LD6000 performance specifications page. (TBD)
Q3:
How is the LM6000 media laid out?. . . . .
See
the LM6000 media page. (TBD)
Q4:
What is media 'striping'?. . . . .
Striping
is a term used to indicate the simultaneous use of 2 or more write/read
heads in rotating media. In LD6000 we stripe data to the LM6000
media by simultaneously writing/reading on both channel A and channel
B. The purpose of striping is to increase read/write rates. Striping
is transparent to the host system. Striping is a complex endeavor
by the drive designers and application developers. The LM6000 media
is organized in Logical Block Address (LBA) space. This space is
considered to be relatively contiguous to application developers
but is actually fragmented all over the disk. For example LBA 0
is physically located at the outermost radius of the B side of the
media while LBA 1 is physically located at the innermost radius
of the A side of the media, go figure. This allows the drive heads
to operate in an opposing fashion; channel B writes from outer to
inner radius while channel a writes from inner to outer radius.
To furthur complicate matters each side of the media is split into
2 zones, you guessed it, inner and outer. the outer zone has 112
sectors per track and the inner zone has 75 sectors per track. This
is due to the outer zone having more physical space than the inner
zone. This non-linear sector arrangement allows LD6000 to make maximum
use of the 12" disks' addressable storage areas.
Q5:
Whats the difference between SCSI single-ended & differential?.
. . . .
There
are several differences;
| |
Single Ended |
Differential |
| Noise Reduction |
No |
Yes |
| Max Bus Length |
6 Meters / 19.2
Feet |
25 Meters /
82 Feet |
| Fast Synchronous
supported |
Not recommended |
Yes |
| Ultra Fast Synchronous
supported |
Not recommended |
Yes |
| Termination |
Passive/Active/FPT |
Differential |
| Cost |
comparable to
Diff |
Comparable to
SE |
| Installation |
Same as Diff |
Same as SE |
| Maintenance |
Same as Diff |
Same as SE |
| Availability |
All platforms supported |
Not all platforms supported |
In
the end, differential SCSI busses are better.
Q6:
What is the 'map area' on the lm6000 media?. . . . .
A map
area is a place where the LD6000 drive performs real time write
retries on the LM6000 media. Due to the nature of WORM media we
are forced to perform write recovery in a place that is NOT in the
logical address space but in a physical address space. This way
we don't get in the way of the host file system. In this map area
there are two types of data, maps and alternates (alts). Maps are
pointers to the physical alts location from the Logical Block Address
location. Alts are the actual user data that are relocated to the
alternate area. The LD6000 writes 247 alts before generating a map.
There is room for 6000 decimal alts on each side of the media. This
is a max relocation rate of .10%. There are actually two map areas
on the LM6000 media, static maps and dynamic maps. Static maps are
generated at the time of manufacture and accomodate microscopic
flaws that occur in the glass and media substrate. Dynamic maps
are generated on-the-fly by the LD6000 drive.
Q7:
Does the LD6000 drive always perform a Write with Verify?. .
. . .
No.
The LD6000 must, according to ANSI, allow the use of the 2Ah write
(no verify) and 2Eh write verify. Philips LMS cannot guarantee the
data has been written unless a 2Eh write verify has been performed.
In order to preserve the use of the write verify command we made
a small change to the 2Ah write command. If the 2Ah command is sent
as shown in the ANSI SCSI spec. the LD6000 drive will perform a
write verify. Yes, a write verify. Reason is we included a DISVFY
bit in the command descriptor block in byte 9 bit 6. If this bit
is not set we will still perform a write verify.
Q8:
What is the maximum read and write sustained transfer rate?.
. . . .
On
its best day at the races the LD6000 can read LM6000 media at 2.3Mbs
sustained and write at 1.1Mbs sustained. The LD6000 can read LM4000
media at 700Kbs sustained. LD6000 does not write to LM4000 media.
Q9.
What can be done to improve the data transfer performance of the
LD6000?. . . . .
- Clean
the media
- Use
a differential SCSI bus
- Turn
on SCSI disconnects
- Turn
on SCSI fast negotiation (10Mbs burst)
Q10.
How often should the LM6000 media be cleaned?. . . . .
The
casual user can perform periodic visual
inspection of the media glass surface and clean as necessary.
The
super user / performance enthusiast can
be very picky about when to clean the media. Use this formula;
ReadRetryCount
/ BlocksRead = ReadRetry rate
These
counters are maintained in the LD6000 drive in NVRAM on the WOODI.
The information is available via SCSI (access event logs command)
or RS-232 (event6k.exe). Each channel, A &
B, are kept current.
Q11.
Why does read ahead sometimes degrade sustained data transfer performance?.
. . . .
The
Read Ahead function in LD6000 shares buffer space with the data
buffer. Because of this when read ahead is in process (reading ahead
and dumping data to the data buffer) and a command is issued to
the drive that does not satisfy the read ahead extent is received
the read ahead function must be shut down before the pending command
can be processed. This shut down process uses drive processor time
which in turn adds time to the pending command.
When
read ahead is used as recommended up to 600% read performance improvements
can be realized.
Q12.
Where can part numbers and part costs be found?. . . . .
Q13.
When are LD6000 training courses offered?. . . . .
Check
out our Technical Training web page
|
