CompuServe recollections - Revision history

Jnc: Add some links

2025-09-20T00:31:57Z

Add some links

Jnc: +Categories

2024-02-20T04:56:12Z

+Categories

Larsbrinkhoff: Internal communications regarding purchase of SC-40's.

2024-01-16T06:29:46Z

Internal communications regarding purchase of SC-40's.

Larsbrinkhoff: Archived here.

2024-01-16T06:27:10Z

Archived here.

New page

=== Some recollections from CompuServe ===

I have no docs but I have examples of a Qbus "Trinode" (3 9-slot DEC-made Qbus backplanes + 3 sets of boot/halt buttons), some of the console wiring boards, and one serial console management interface (Qbus-shaped, sits in the bus for power only, has 8, I think, serial ports to manage 8 "Nodes" from one external interface.

In the old, old days, a Node would have a PDP-11 processor and a wad of DEC serial mux cards and whatever interface to the X.25-like WAN network. By the end of days, the PDP-11 was replaced with that 486 board, and I'm pretty sure they had custom 16-port mux boards, but it was just higher-density versions of the DEC ancestors.

I am pretty sure that sometime around 2000, they migrated a lot of the content over to their 32-bit boxes - "silver bullets" - custom chromed-steel cases that were over 24" long and ~12" wide that would fit two side-by-side in a 24" rack. They ran WinNT and/or some flavor of BSD depending on role.

I was not working there when they decommed that hardware so I didn't get one of the racks (parking lot giveaway from what I heard). I did get 1-2 of the trays from a friend who scooped some stuff up then. I also got a couple of the Silver Bullets. EISA Pentium boxes with high-end SCSI cards. I haven't examined every single disk but a random sampling did not reveal anything particularly interesting in what fell into my hands.

I worked there 1994-95 when the first SC machines were beginning to arrive, then again 2001-2002 when all the DEC 36-bit hosts were completely gone and most or all of the 32-bit hosts were gone. There were still some SC-40s online (which stayed up to past 2005) but as mentioned, they were user administration and accounting (billing for people with ancient PPN accounts who had kept them for vanity reasons). By 2002, all the real activity was on a patchwork of Solaris/AIX/IRIX/HPUX boxes.

I also worked at H&R Block in 2003 and ran their VMS cluster in the same room as some of the remaining SC-40s (since H&R Block used to own CompuServe, they maintained a co-lo agreement even when the building had transferred to WorldCom).

I don't know exactly what was on those SC-40s in the Hilliard Datacenter but I was told it was basically just accounting/billing for some part of the business that WorldCom got in the split. I did hear that they spent $$$ and years trying to replace the billing app but it was written in XF4 and apparently quite difficult to functionally reverse-engineer and rewrite in a modern language so they just kept the SC-40s plugging away year after year.

I nabbed my comms box (with permission) from the loading dock when they were decomming some gear in Hilliard in 2003. No cards. I did see one of those 486 cards on the bench once.

here's a mention: [https://wiki.preterhuman.net/A_Brief_History_of_36-bit_Computing_at_CompuServe A Brief History of 36-bit Computing at CompuServe]

(I worked in "Building 7" at the HQ mentioned that was first occupied in 1973)

(and I later worked with Dr. John Goltz when I was at Software Results - our founder had worked with him when they were both at CompuServe in the 70s)
Looks like the SC-30s showed up a bit earlier than I remember.

the history is, I think, just the progression from the first then second KA10 through several KI10 to present day (1988) and the recent arrival of the SC30s. They went on to run 36-bit hardware close to 20 more years.

FWIW, about 3 weeks ago, a historic marker was placed at the former CompuServe HQ in Upper Arlington.

← Older revision		Revision as of 04:56, 20 February 2024
Line 67:		Line 67:
	questionable whether any more could be produced.		questionable whether any more could be produced.
	</pre>		</pre>
		+
		+	[[Category: Vendors]]
		+	[[Category: History]]

← Older revision		Revision as of 06:29, 16 January 2024
Line 27:		Line 27:

	FWIW, about 3 weeks ago, a historic marker was placed at the former CompuServe HQ in Upper Arlington.		FWIW, about 3 weeks ago, a historic marker was placed at the former CompuServe HQ in Upper Arlington.
		+
		+	=== Internal communications ===
		+
		+	<pre>
		+	SUPPORT SERVICES DIVISION OVERVIEW
		+	November 1995
		+
		+	Alexander B. Trevor
		+
		+
		+	I. HOSTS, SERVERS, SYSTEMS
		+
		+	o 32-Bit Platform - We have 1036 32-bit servers in production (190
		+	assigned to ARTF). We are evaluating several products to help manage
		+	this large number of systems (e.g., SMS, CA-Unicenter).
		+
		+	o 36-bit Platform - Consumption of 36-bit capacity has temporarily
		+	leveled off, but in anticipation of the post-Christmas growth, we have
		+	asked the SC Group to build a few more machines for us -- consuming the
		+	last SC-40 chips. That’s it folks -- we MUST move forums in December.
		+	</pre>
		+
		+	<pre>
		+	SUPPORT SERVICES DIVISION OVERVIEW
		+	December 1995
		+
		+	Alexander B. Trevor
		+
		+
		+	I. HOSTS, SERVERS, SYSTEMS
		+
		+	o 32-Bit Platform - Over Christmas, several unexplained outages occurred
		+	that blocked access to the NISA applications (Forums, mail). Diagnosing
		+	and fixing the cause is the top priority for Systems, as this problem has
		+	stalled the move of 32-bit Forums into production.
		+
		+	o 36-bit Platform - We received 15 more SC-40’s this month. There are
		+	only 10 SC-40 chips remaining that we have not already ordered, and it is
		+	questionable whether any more could be produced.
		+	</pre>

@@ Line 1: / Line 1: @@
 === Some recollections from CompuServe ===
-I have no docs but I have examples of a Qbus "Trinode" (3 9-slot DEC-made Qbus backplanes + 3 sets of boot/halt buttons), some of the console wiring boards, and one serial console management interface (Qbus-shaped, sits in the bus for power only, has 8, I think, serial ports to manage 8 "Nodes" from one external interface.
+I have no docs but I have examples of a [[Qbus]] "Trinode" (3 9-slot [[DEC]]-made Qbus backplanes + 3 sets of boot/halt buttons), some of the console wiring boards, and one serial console management interface (Qbus-shaped, sits in the bus for power only, has 8, I think, serial ports to manage 8 "Nodes" from one external interface.
-In the old, old days, a Node would have a PDP-11 processor and a wad of DEC serial mux cards and whatever interface to the X.25-like WAN network.   By the end of days, the PDP-11 was replaced with that 486 board, and I'm pretty sure they had custom 16-port mux boards, but it was just higher-density versions of the DEC ancestors.
+In the old, old days, a Node would have a [[PDP-11]] processor and a wad of DEC serial mux cards and whatever interface to the X.25-like WAN network.   By the end of days, the PDP-11 was replaced with that 486 board, and I'm pretty sure they had custom 16-port mux boards, but it was just higher-density versions of the DEC ancestors.
 I am pretty sure that sometime around 2000, they migrated a lot of the content over to their 32-bit boxes - "silver bullets" - custom chromed-steel cases that were over 24" long and ~12" wide that would fit two side-by-side in a 24" rack.   They ran WinNT and/or some flavor of BSD depending on role.
-I was not working there when they decommed that hardware so I didn't get one of the racks (parking lot giveaway from what I heard).  I did get 1-2 of the trays from a friend who scooped some stuff up then.  I also got a couple of the Silver Bullets.  EISA Pentium boxes with high-end SCSI cards.  I haven't examined every single disk but a random sampling did not reveal anything particularly interesting in what fell into my hands.
+I was not working there when they decommed that hardware so I didn't get one of the racks (parking lot giveaway from what I heard).  I did get 1-2 of the trays from a friend who scooped some stuff up then.  I also got a couple of the Silver Bullets.  [[EISA]] [[Pentium]] boxes with high-end [[SCSI]] cards.  I haven't examined every single disk but a random sampling did not reveal anything particularly interesting in what fell into my hands.
-I worked there 1994-95 when the first SC machines were beginning to arrive, then again 2001-2002 when all the DEC 36-bit hosts were completely gone and most or all of the 32-bit hosts were gone.  There were still some SC-40s online (which stayed up to past 2005) but as mentioned, they were user administration and accounting (billing for people with ancient PPN accounts who had kept them for vanity reasons).  By 2002, all the real activity was on a patchwork of Solaris/AIX/IRIX/HPUX boxes.
+I worked there 1994-95 when the first [[SC]] machines were beginning to arrive, then again 2001-2002 when all the DEC 36-bit hosts were completely gone and most or all of the 32-bit hosts were gone.  There were still some SC-40s online (which stayed up to past 2005) but as mentioned, they were user administration and accounting (billing for people with ancient PPN accounts who had kept them for vanity reasons).  By 2002, all the real activity was on a patchwork of Solaris/AIX/IRIX/HPUX boxes.
-I also worked at H&R Block in 2003 and ran their VMS cluster in the same room as some of the remaining SC-40s (since H&R Block used to own CompuServe, they maintained a co-lo agreement even when the building had transferred to WorldCom).
+I also worked at H&R Block in 2003 and ran their [[VMS]] cluster in the same room as some of the remaining SC-40s (since H&R Block used to own [[CompuServe]], they maintained a co-lo agreement even when the building had transferred to WorldCom).
 I don't know exactly what was on those SC-40s in the Hilliard Datacenter but I was told it was basically just accounting/billing for some part of the business that WorldCom got in the split.  I did hear that they spent $$$ and years trying to replace the billing app but it was written in XF4 and apparently quite difficult to functionally reverse-engineer and rewrite in a modern language so they just kept the SC-40s plugging away year after year.
@@ Line 24: / Line 24: @@
 Looks like the SC-30s showed up a bit earlier than I remember.
-the history is, I think, just the progression from the first then second KA10 through several KI10 to present day (1988) and the recent arrival of the SC30s.  They went on to run 36-bit hardware close to 20 more years.
+the history is, I think, just the progression from the first then second [[KA10]] through several [[KI10]]s to present day (1988) and the recent arrival of the SC30s.  They went on to run 36-bit hardware close to 20 more years.
 FWIW, about 3 weeks ago, a historic marker was placed at the former CompuServe HQ in Upper Arlington.