Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sample.eng.ohio-state.edu!purdue!haven.umd.edu!uvaarpa!murdoch!usenet
From: sdm7g@Virginia.EDU (Steven D. Majewski)
Newsgroups: comp.periphs.scsi
Subject: Re: SLOW WRITE ON OPTICAL ERASABLES
Summary: also ECC & Defect mgmt. overhead
Keywords: ANSI format, Magneto-Optical Media
Message-ID: <1991Jun25.160859.16313@murdoch.acc.Virginia.EDU>
Date: 25 Jun 91 16:08:59 GMT
References: <1080@camco.Celestial.COM> <1991Jun15.173019.27914@scuzzy.in-berlin.de>
Sender: usenet@murdoch.acc.Virginia.EDU
Reply-To: sdm7g@Virginia.EDU
Followup-To: comp.periphs.scsi
Organization: University of Virginia
Lines: 69

In article <1991Jun15.173019.27914@scuzzy.in-berlin.de> src@scuzzy.in-berlin.de
(Heiko Blume) writes:
>bill@camco.Celestial.COM (Bill Campbell) writes:
>>I'm in the process of evaluating erasable optical drives.  The
>>first drive I've installed is a Ricoh RO-5030E2, and I'm
>>surprised at how slow it is writing. 
>
>[benchmark]
>
>>Is this typical of erasable write speeds?  I expected them to be
>>slow, but not this slow!  Even removing a file from a directory
>>seems to take a long time while it updates the directory entry.
>
>magneto-optical recording has a problem: before writing data to
>a spot that spot must first be erased. that means if you want
>to write a block, it takes at least one full revolution + 1 block
>of the disk. on average 1.5 revolutions are needed, which causes
>the slow writing you experienced.
>

It also depends on whether the driver is doing a WRITE/VERIFY on 
all writes. ( This SHOULD be the DEFAULT, but it is slower. ) 

ANSI formatted Magneto-Optical disks have a Primary and a Secornday
defect list. The Primary defect list is built when you format the 
disk. If a write verify is done, then a read is performed using only
half of the error-correcting code capability. If it can be read, that
ensures that you still have some head-room in the ECC to protect you
from future data loss. If it gets an error on read, then a replacement
block is allocated and written, and the bad block is entered into the
Secondary Defect List.  

This is done on a SCSI WRITE/VERIFY command, or if the default for ALL
WRITE/COPY/etc. is set to WRITE/VERIFY with a SCSI SET MODE.

I don't know how may steps (reads/writes) are involved or if the
controller caches any of the lists, but it sounds like quite a bit 
of overhead. 
The erase before write is necessary because the laser switches 
faster & focuses finer than the magnetic field. When a spot 
is heated by the laser, and then cools, it adjusts its polarity to the 
magnetic field. If you needed to focus the magnetic field to write
succesive 1/0/'s, you would need the type of head distance that a 
Whichester has.  So instead, it writes a whole sector to zero, switches
the field, and selectively writes the ones. Some software can do 
pre-erasing and manage a list of available (erased) blocks. 
So worst case is: ERASE SECTOR, WRITE SECTOR, READ SECTOR, (ERROR),
 READ DEFECT LIST (to find next available replacement block), 
 WRITE REPLACEMENT BLOCK, READ SECTOR (verify again), WRITE (update)
 DEFECT LIST. ( The *really* bad news is the SEEKS implicit in the 
 above! Some drives MAY be able to do a read during write - that way you
 only pay a penalty when there actually is a write error - the best of 
 all possible worlds! ) 


If you are using the MO disk as a scratch device, you can probably turn
this off. If you intend it to be used for Archival Data Storage, you 
should be using the WRITE/VERIFY. 
Pinnacle's disks for MS_DOS DO NOT do WRITE/VERIFY. ( They were more 
interested in improving their performance specs, than in keeping your data! ) 
Their disks & drivers for MAC and UNIX ( I have been told ) have the 
mode as a user selectable option. ( I don't know what the default is.)

*********
 I am going to try to prepare a horror story (posting) about what can
 happen to you with improperly formatted MO media. The contents of my disk
 got up and moved 31 sectors to the right ( boot block, FAT, Directorys
 and all ! ) !
*********