| Subject | Re: [Firebird-Architect] Feature Request - records before disposal |
|---|---|
| Author | Jim Starkey |
| Post date | 2004-12-16T03:07:51Z |
Nigel Weeks wrote:
Rdb/ELN, and the follow-on private sector Interbase.
VMS clusters had two roots. The first was primordal terror at the
prospect of Tandem eating DEC's Telco lunch. The other was DEC's
institutional inability to make a fast CPU. The first crack was the
PDP-11/74, DEC's first multi-processor based on the 11/70 ttl cpu
running Dave Cutler's RSX-11M+. It didn't have enough address space to
support the cycles available, and got canned, but until creating the
Cerberos T-shirts, by far the most sought after DECstuff.
VAX clusters were based on two pieces of key technology. One was the CI
(computer interconnect, another DEC story), an extremely fast LAN cast
in hardware. The other was Steve Beckhardt's distributed lock manager.
Supporting this was the HSC ("hot shit controller") that provided disk
service over the CI to the cluster. The HSC group in Colorada (aka
ColoSpgs) wanted to morph the HSC into an even higher margin database
machine, and took over sponsorship of Renegade Systems (me, Ann, Don,
and Jeff) after the opening salvo of the Database Wars. JRD's (yes, the
very same acronym) first release was to Jayhawk, a dedicated database
server running on a MicroVax II.
So, yes, Rdb/ELN was the first database designed specifically for
VaxClusters. Rdb/VMS was second, and Interbase was third. But what I
didn't appreciate was the fact that nobody so deluded to blow money on a
VAX would even consider buying third party software. Perhaps my first
major business misjudgement, but far from my last.
The basis for a cluster is shared disk access synchronized by a cluster
wide lock manager capable of delivering blocking asts. Funny -- that's
exactly the same way Vulcan works (if you ask it nicely).
>>Funny, I asked the RMS guys pretty much the same question at DEC inAbsolutely. It was the technology that drove the design of JRD, aka
>>1976. They said, sure, if it doesn't have to be accurate, then "13".
>>
>>
>>
>
>I keep seeing references to RDB here at work - we've got a few DEC Alphas
>running OpenVMS.
>Did RDB(or your JRD) ever support VMS's clustering system?
>That's a weird system I'm still getting my head around...;-)
>
>
>
Rdb/ELN, and the follow-on private sector Interbase.
VMS clusters had two roots. The first was primordal terror at the
prospect of Tandem eating DEC's Telco lunch. The other was DEC's
institutional inability to make a fast CPU. The first crack was the
PDP-11/74, DEC's first multi-processor based on the 11/70 ttl cpu
running Dave Cutler's RSX-11M+. It didn't have enough address space to
support the cycles available, and got canned, but until creating the
Cerberos T-shirts, by far the most sought after DECstuff.
VAX clusters were based on two pieces of key technology. One was the CI
(computer interconnect, another DEC story), an extremely fast LAN cast
in hardware. The other was Steve Beckhardt's distributed lock manager.
Supporting this was the HSC ("hot shit controller") that provided disk
service over the CI to the cluster. The HSC group in Colorada (aka
ColoSpgs) wanted to morph the HSC into an even higher margin database
machine, and took over sponsorship of Renegade Systems (me, Ann, Don,
and Jeff) after the opening salvo of the Database Wars. JRD's (yes, the
very same acronym) first release was to Jayhawk, a dedicated database
server running on a MicroVax II.
So, yes, Rdb/ELN was the first database designed specifically for
VaxClusters. Rdb/VMS was second, and Interbase was third. But what I
didn't appreciate was the fact that nobody so deluded to blow money on a
VAX would even consider buying third party software. Perhaps my first
major business misjudgement, but far from my last.
The basis for a cluster is shared disk access synchronized by a cluster
wide lock manager capable of delivering blocking asts. Funny -- that's
exactly the same way Vulcan works (if you ask it nicely).