F/W Streaming RAID
SCSI-2 Fast/Wide Streaming RAID Adapter/A  "Cheetah" FRU 06H3059
   Function of NVSRAM
   Cable Parts
HD LED Does Not Work
   HD LED Hack
Cyrix/Non-SOD Type 1 Incompatibility?
Cache Size
   Cache Size Factors
   Cache Size and Diminishing Returns
Accessing the RAID Configuration
   Configuration Utility
FWSR Bios Flash Disk
Cheetah in a 85 / 95 /95A
Cheetah in a Server 500
   Getting CD Rom to WORK On Server 500
   Scientific Wild-Assed Guess Why it Worked
Linux on FWSR?
LVD Drives on Cheetah?
Specifications For FWSR

Chaos forever! These are related.
Create and Maintain your Array
Array technology, features, classifications(FWSR)
RAID Message Table
Hotswap bays for 95A
   Removing side panel from 3 bay cage
   Closeups of microswitches



SCSI-2 Fast/Wide Streaming RAID Adapter/A  "Cheetah" FRU 06H3059
                                                                                                  Sidecard FRU 06H3060
194-170 19940517 IBM SCSI-2 Fast/Wide Streaming-RAID Adapter/A
FWSR Features

J1 Channel 1 68 pin edgecard
J2 Channel 2 68 pin edgecard
J3 Not connected. Or used.
J4 DASD Status Connector
J5 DASD Status Connector
J6 Possible serial port. Unused.
J9 Channel 2 external port.
Y1 50 MHz Oscillator
Y2 40MHz Oscillator
F1 Channel 2 PTC resistor
F2 Channel 1 PTC resistor

Channels
   The Cheetah has two channels. Each channel is controlled by an NCR53C720. The header J1 is Channel 1. It usually is attached to an internal array, but with the addition of a side card, it can controll an external array. The second channel uses J2 OR the external port, J9. This is still one channel, so one can use either the internal port, OR the external port. Do NOT try to use both J2 and J9 at once.

Notes:
NVSRAM is a Benchmarq 28 pin 8Kx8  bq4010YMA-200, Spec sheet
  Another equivalent is a Dallas DS1225Y-200, spec sheet

NVSRAM Functions
  Each NV SRAM has a self–contained lithium energy source and control circuitry which constantly monitors VCC for an out–of–tolerance condition. When such a condition occurs, the lithium energy source is automatically switched on and write protection is unconditionally enabled to prevent data corruption.

Cable Parts
   The mini C68 for the Channel edgecard connectors is the Molex 71660i, part# 15-92-3068, called a half pitch Centronics, or a VESA Media Connector. Suprise! AMP makes a similar part (mini-C68) AMP Part 1-557089-2 Any cable with a .025 pitch, 28 to 30 AWG will work with either connector.
  Don't ask if you can get this cable for $9.99! I have priced out the black sheathing, twist 'n loose 68 wire cable, and the molex connectors. If you bought the cable by the 100' reel (if you can get it shorter, tell me) and the sheathing the same way, the parts alone cost upwards of $50 for a replica of the original FW RAID cable. Can't run with the big dogs if you are whining with the pups.
   No, I do NOT have a pile of them in my closet! No, I do NOT know where to get them at $5 each. No, I do NOT know how to make a replica for $10. It's your data. If you want to cobble a cheap cable up, I suggest buying SCSI-III flat cable from Dalco and crimping on the needed Molex/AMP connectors. But don't blame me if you loose something.

A Better Cable Hack?
Allen Brandt wrote:
> A small, shotty attempt to get something uploaded concerning the PS/2. HERE

My Take on it:
   I am starting to have neurons fire. Actually, Allen provided the push. Al went and slit the conductors for better flexibility (in pairs).
   Could you slit the flat cable up towards the controller and get the very flexible cable bundle of the IBM original? The black sheathing is available from Jameco for about $1 a foot. Well worth it, IMHO. (Start the slit with an X-Acto and use the reverse of the blade to finish parting the conductors???)
   The sheathing is Techflex Cable Sleave, looks to be the 3/8" size. Sold in a 25' spool. Part #162157, Product # CCPT2X per spool $14.95  Techflex is HERE
What kind of signal degredation might occur? Each signal pair hopefully cancels it's noise out.
   If the Brandt manuever can be done from the top drive connector to theadapter, it might be a close match to the real thing



HD LED Doesn't Work
>Is i a fact that the HD LED does not work on a 9595A with a PassPlay RaidAdp.?
   The fixed disk light is non-functional with both the Server 95 A "Passplay" and Streaming-RAID "Cheetah" MCA RAID adapter.  I suspect this is also the case with other OEM'ed Mylex RAID adapters.

LED For Cheetah
   BUT if you take an LED off of J6, pin 1 and 2, it will light when the drives are accessed. Just run a lead up to between the LED blocks in the display panel. Watch the polarity. If the LED doesn't light, switch the header around. You do not need a resistor for this. 
   I tried this, but the LED didn't have enough umph. Pretty dim through the LED Panel. Maybe some sort of a drive circuit?
  Just had a thought- twist the existing HD LED out of the Op Panel and put the LED that is connected to J6 in there....

Possible Cyrix-Cheetah Incompatibility?
Tim Clarke
Hi gang,
      just thought that I'd better warn you. After checking out the Cyrix 5x86 at 4x clocking (in Type-1 non-SOD w/cache) my PassPlay RAID adapter seems to have been "duffed up". I only get a part of the BIOS v1.05 initialisation/installation message and the machine hangs (with *any* CPU) at CP:96. Looks as though the Flash ROM has been partially overwritten (just a guess).



Cache Size
  I know that the Cheetah has 4MB soldered on. This is to help you understand WHY 4MB is probably all you'll need. 
    How important is the amount of cache ram on the PassPlay RAID adapter--4 MB, 16 MB, 64 MB?  Under what circumstances will a cache increase pay off? (The system in question is running NetWare 4.1, but I'm interested in general info on this subject.) 
    I notice that the more recent Cheetah RAID adapter has only 4 MB with no upgrade possible. It seems counterintuitive, but I seem to remember reading somewhere that large amounts of controller cache aren't really that useful with modern drives and operating systems. 

From Peter
   Having a large cache is only half the truth. Bigger cache means more damage if the controller chokes and cannot write data back to the drives. Large caches on Raid controllers make sense only if they are battery-backed  (Ed. I have seen battery-backed 72 pin SIMMs) and if there is a cache and restart the system to that point where the operation was stopped and write the cache data down to the drives and maintain the integrity of the data / array. 
    The older Raid-controllers (Server-95 Raid "Passplay", Fast/Wide Streaming Raid /A "Cheetah" and Fast/Wide Raid PCI "DAC960") don't have battery backed cache. Even 4MB of cache memory contains a large number of "data-stripes" (usually 8K blocks). 
      These data-stripes will be lost if the machine powers down for any reason, or the controller fails, or the operating system hangs. Recalculate how many sectors fit in 4MB - and the higher the  number of missing sectors the lower the chance that the Raid-Utility will be able to restore the missing data. 

Cache Size Depends On 
   a) Overall drive data-throughput (buffering x accesses while drives are in *mechanically* causes delay / dead zone / recalibration) 
   b) Data-stripe size (8K normally - 64K under WinNT might give better results) 
   c) Operating system (WinNT and OS/2 are very "swap active") 
   d) Structure of the RAID itself (Raid-5 uses the cache much more than Raid-1 ... because the mirroring is imminently fast with buffering the data).
   e) the nature of the data blocks. Consequently high internal redundancy of the data will cause higher "hit rates" within the cache than permanent data-streaming with new data, which void the cached data and only "pass through". 
       Like on all caches there is a limit where enlarging the cache any further makes no sense. And I think this limit is at around 4MB on a 5 drives Raid-5 system running under OS/2 or WinNT. The content-redundancy of the data is mostly not given - so the cache is mostly used to buffer the Raid data-overhead between the drives (during reading / writing / synchronizing the Raid structure) - on the transfer between drive-subsystem and processor the cache does not play a major role. 
      A larger cache here costs only money and bears the above mentioned risks to render the entire array useless if something crashes. 

Cache Size and Diminishing Returns
 From "Mark"
      Generally speaking, increasing the amount of cache will always improve performance. The  performance gain will be more for sequential access type applications than for random access type applications. Typically increasing the cache from 2 to 4MB will see a bigger % gain than 4 to 16 MB and that will see a bigger % gain than 16 to 32 MB and so on. 



Access the RAID Configuration
   Both the FWR (Passplay) and FWSR (Cheetah) are only configurable through the RAID Utilities disk. You CANNOT see the SCSI Disks under "Set and View SCSI Devices" like normal SCSI drives. Boot with FWSR Option Disk, #1 ver. 2.31 in order to view or configure the array. 
   Both adapters use the same Utilities disk of the later IBM F/W Streaming RAID Adapter /A (Codename "Cheetah" - with external port) since both are based on Intel i960 / Mylex / NCR technology. There was a single-disk version 2.22, which should be unique for all /A-Raid adapters of that kind, but not the PCI-versions. The RAIDADM (manager) should work on both /A-adapters.

Configuration Utility version 2.31 consists out of two disks:
FWSR Option Disk, #1 ver. 2.31
FWSR Option Disk, #2 ver. 2.31
Readme for FWSR Option Disks

Not sure if this fits-
RAID Supplemental Diskette Version 2.0    And the Readme.txt RAIDSEND is a utility that provides an OS/2 ONLY command-line interface for performing various tasks on a IBM F/W Streaming RAID Adapter/A, the IBM SCSI-2 F/W PCI-Bus RAID Adapter, and the Mylex PL adapter for the IBM PC Server 704.



Fast/Wide Streaming RAID Flash Bios for "Cheetah"FRU 06H3059

CAUTION!!!
   The Passplay and the Cheetah differ in the microcode, which *may not* be interchanged. The Passplay (FWR) adapter uses a microcode-level 1.6x through 1.99, the Cheetah (FWSR) uses 2.xx levels. If you flash the one adapter with the code from the other you end up in non-functional adapters.

FWSR Flash BIOS 2.21   Only for RAID controller WITH an external port on it!
FWSR Flash BIOS Readme



Cheetah in a 95
  The RAID bay for the 85/95/95A does not have a place foe the status cable to attatch. The RAID bay has a 68 pin edgecard at the back where the molex style SCSI connector attaches to. The 95 RAID bays automatically terminate the SCSI drives inside. Do NOT enable termination on the individual drives!
   I installed a CD Rom in Bay 7. I used a 68 to 50 pin adapter from the RAID cable connector. I have installed both NT Workstation 4 and OS/2 on it. Both were able to detect and use the CD Rom during setup. FWIW, I had only one bay with three drives in it.
 

Cheetah in a Server 500

Setting the CD Rom ID in a Server 500
From Rich Nagle
  Following repeated failures of NT 4 Server setup to recognize the CD Rom connected to the passthrough connector on the top backplane, I noticed that the CD Rom was showing up as one SCSI ID# higher than it was when I checked it under the RAID Utility View Configuration.
  After checking the SCSI ID jumper on the backplane (set to LO for IDs 0 thru 5 on the backplane), a sudden flash of inspiration occured- I set the CD Rom to ID 5, went back under the RAID Utilities, and the CD Rom was now ID6. I then deleted, then recreated the array. Now when I ran NT Setup the CD Rom was recognized automatically. 

God-like SWAG of What Happened
From Us, the god-Emperor of Microchannel
   While Rich was single-handedly adding a dollar to each share of AT&Ts stock, I noticed that the SCSI IDs for each drive connector on the backplane went from 0 to 5. This wasn't the blinding flash that explained everything, but I think it's the reason. Read on to follow my logic.
   Peter said he set CD Roms to ID6 for installation on one contract. But Rich noticed the odd +1 to the SCSI ID of the CD Rom. Remember the SCSI IDs went from 0 to 5. We know 7 and 15 are used by the controller. But ID6 is NOT accounted for.
   Without the inspiration of Jolt, I was hobbled to mere human powers. Try this for a fit:
Any device (backplane or drive) attached to the passthrough connector automatically has 1 added to the SCSI ID. The SCSI ID jumper has something to do with setting a fourth bit for ID.
   For a CD actually jumpered to ID6, the backplane adds 1 (it's attached to the passthrough connector). The CD Rom becomes ID7, which conflicts with the SCSI controller. The SCSI controller asserts it's rank and shuts the CD Rom up. 

Linux on FWSR
From Peter
>> Is anyone running linux on one of these machines?

Not on machines with the IBM Raid controller with the old 2.43 firmware. No Linux driver available

The IBM Fast/Wide Streaming Raid Adapter PCI as used in the Server 320/520 MCA-PCI versions is derived from the Mylex DAC960PL - it only has 128K Flash ROM (one 28F010 chip) but a second open socket. Firmware 3.x requires 256K Flash. I'd tried to plug in a second 28F010 ... but I think the old software contained in that chip confused the adapter a bit ... it behaved a little "strange" (long boot time etc.)

What I do not have is an Eprommer that is capable to write the Flash-ROMs of the 28Fxxxx series or I could a) write a spare 2.4x Flash (to keep for the "worst case") and b) clear the 28F010 ROMs I pulled from some old boards. Else I would stuff in a blank ROM in the second socket, have the old 2.xx in the first and run a firmware update 3.x from the DAC960PL on that adapter.

The machines with the older RAID-adapters ("Passplay" and "Cheetah") based on MCA technology are out of the discussion anyway. They are based on the DAC960M technology basically but an older draft of that concept. They use some of the chips of the -M and early -Px adapters (PL / PD) and they are developed by Mylex - but the firmware 3.x is PCI-specific, not MCA. So you can practically forget about using them under Linux since the driver is *particularly* written for the 3.x firmware level.

LVD on Cheetah
>What kind of drives does the RAID take? Is FAST/WIDE DIFFERENTIAL SCSI the right kind? Or are LVD (low voltage differential) different and it needs them instead? I've never dealt with RAID before.
 
 From Peter
 Remember the  "Cheetah"-Adapter's "Real Trade Name" ?  IBM Fast/Wide Streaming Raid Adapter /A.
   It it an ordinary Fast / Wide indended for single-ended SCSI devices. It does however take U/W LVD drives, because these are downward-compatible to single-ended, which the old
 "high-voltage differential" are *not*. 
      If you get - for example - a set of U/W "Low Voltage Differential" (LVD) IBM DDRS 4.5 or
 9.1GB drives then they will nicely run with the Cheetah. I have some of them in "Starship" - my
 Server 520 attached to the Fast/Wide RAID Adapter PCI. No problem. You can even mix them
 with "ordinary" F/W or U/W drives. Same for the Cheetah and even the older Passplay.



Cheetah Specs
SCSI type   SCSI-2 Fast/Wide
SCSI bus path / speed 16 bit / 20 MB/sec
I/O bus path / speed 32 bit / 40 MB/sec streaming 
(80 MB/sec on PC Server 720)
I/O features Streaming data transfer
Address parity and data parity
RAID levels  RAID 0, 1, Hybrid 1, 5
4 ind (A, B, C, D) / 8 logical arrays
Tagged Command Queuing Yes
Processor i960 at 25 MHz
Size Type 3 (full length)
Channels Two (one internal; one internal or external) 
Connectors Three total:
  Two internal - 16 bit wide 
(sidecard not included on Server 500)
One external - 16 bit wide
*Can only use two connectors at once
Devices supported 14 per adapter (7 per max per channel)
Cache std / max 4 MB / 4 MB (with parity) 60 ns
Cache method Soldered on adapter
Cache write policy Write-through or write-back

Controllers Main Page

9595 Main Page

or external)  Connectors Three total:   Two internal - 16 bit wide 
(sidecard not included on Server 500)
One external - 16 bit wide
*Can only use two connectors at once Devices supported 14 per adapter (7 per max per channel) Cache std / max 4 MB / 4 MB (with parity) 60 ns Cache method Soldered on adapter Cache write policy Write-through or write-back

Controllers Main Page

9595 Main Page