Re: Blast from the 1988s (DEC proposal)

I think you're right Mark - the message I've heard from a number of
people is that it's a case of needing to take lots of peoples'
workstations over for a weekend in order to gather enough systems
together to form a cluster that big - most companies only have that
number of PCs available for this kind of job. Few have that many VMS
systems available.

The limit probably got changed with the introduction of LAVc.

Steve

Mark Daniel wrote:
Mark Daniel wrote:
Larry Kilgallen wrote:

In article <87ac0kdrjo.fsf@xxxxxxxxxxxxxxxxxxx>,
prep@xxxxxxxxxxxxxxxxxxx writes:

JF Mezei <jfmezei.spamnot@xxxxxxxxxxxxx> writes:


At the time, the graphs showed from the all mighty microvax II to
the VAX 8978. The document mentions up to 15 8800 nodes in a
cluster. (when did it go from 15 to 96 nodes in a cluster ?)


AIR, the 15* was the limit on the number of nodes on a single system
volume, not the cluster size limit. I *THINK* that back then the size
of a cluster was limited by the size of the statically allocated
System Director Vector for the DLM.



Formerly there was a distinction between the total number of nodes
supported in a cluster and the number of large nodes supported in
a cluster. This was based on what had been tested, not theoretical
limits.


From memory (and it's been a while since I've actually seen a CI -
though I think we've still got the cables between our two, redundant
machine rooms); it was 32 nodes per CI (the star-coupler), half of which
could be VMS systems and the other half storage controllers (e.g. HSCs).
The number of CI's (redundant paths or independent storage) on any one
system depended on the size of the iron. We (HFRD->WASD->SSD->ISRD) had
over the span in excess of a decade a time-variable collection of
systems based on a mixed-interconnect cluster, including a handful of
CI-based systems (including at one stage a VAX9000), along with NI-based
VAXservers, VAXstations, DECstations, AlphaServers, AlphaStations, etc.
Our main cluster exceeded 70 systems at one stage.

What I would have added had I not inadvertantly hit [send]; IIRC the 96
nodes was a supported maximum because that was the largest cluster
Engineering had managed to configure on the test-bench. I'm (idly)
curious about design limitations (ignoring the practical limitations of
memory, bandwidth, etc.) on cluster size. Things like the 65k DECnet
node limit. (Hope the spelling's less objectionable this time, Steve.)

.

Relevant Pages

  • Re: Cannot use GPO Deployment from Cluster Volume
    ... OK Mark, now we're on the right track. ... Windows 2000 MCSE + MCDBA ... Windows NT/2000/2003 Cluster Technologies ... > From there I click "Permissions" and try to add a global security group to ...
    (microsoft.public.windows.server.clustering)
  • Re: Cannot use GPO Deployment from Cluster Volume
    ... Not quite sure what bug you are referring to, but if you have a reference ... Windows NT/2000/2003 Cluster Technologies ... "Mark H" wrote in message ... I have to set and revoke permissions with a script. ...
    (microsoft.public.windows.server.clustering)
  • Re: Blast from the 1988s (DEC proposal)
    ... not the cluster size limit. ... From memory; it was 32 nodes per CI, half of which could be VMS systems and the other half storage controllers. ... over the span in excess of a decade a time-variable collection of systems based on a mixed-interconnect cluster, including a handful of CI-based systems, along with NI-based VAXservers, VAXstations, DECstations, AlphaServers, AlphaStations, etc. ... I'm curious about design limitations on cluster size. ...
    (comp.os.vms)
  • Re: HACMP: clstat
    ... Uzytkownik "nc_mark" napisal w wiadomosci ... > Mark, are you sure snmp working properly? ... clstat it doesn't display the state of the cluster." ...
    (comp.unix.aix)
  • Re: advice on growing from 1 to 2 servers
    ... Mark - thanks for all your comments. ... > If your pocket won't stretch to a cluster and a good tech resource to ... > doesn't run to redundancy on Exchange and you don't get it out of the ...
    (microsoft.public.win2000.advanced_server)