[Info-vax] New filesystem mentioned

[johnwallace4 at yahoo.co.uk]

On Thursday, 16 May 2019 11:08:04 UTC+1, IanD  wrote:
> I seem to remember something about Oracle using the VMS clustering goodies too but it was nothing in writing, just spoken about. Perhaps it was something to do with them using a dlm that was similar?
> 
> The Lustre file system makes direct mention of modeling some of it's distributed lock manager off of concepts used by VMS dlm
> 
> Like in lots of things, VMS has sat on it's outstanding technology and charged an arm and a leg for it
>  
> This help foster a catch up market where others had an incentive to create their own. 
> One can argue as much as one likes as to the strength of VMS clustering but other offerings are now possibly just as good or good enough and the VMS premium is considered too expensive now
> 
> What and when was the last technological advance made in VMS clustering? 
> Compare this to innovation made by others now and where clustering and distributed technology innovation comes from now
> 
> Hashgraph and related technologies might hold ways of expanding existing clustering technologies with it's blistering fast transaction rates (100's of thousands of transactions per second). Sure, it's limited by the number of nodes you can gossip with in a reasonable timeframe, but I believe it's around 1000 or so, so maybe it could be adapted to support VMS clusters of that size?
> 
> Perhaps there's opportunity for VMS clusters still if they can truly be a backbone for distributed redundant synched data supporting high speed replication or is that chasing a crowded market?
> 
> But there's so much work to be done, this sort of cutting edge clustering stuff could be 10 years away before being production available and I doubt I could even imagine what other technologies might be upon us by then
> 
> At least VSI has done more for VMS in the short time they tucked VMS under their wing compared to HP who let it rot and worse, let the public think it had no future by being vague about it's future until the 11th hour of it's death announcement (yeah, I'm still a tiny bit bitter how HP treated VMS!!!)

Have you looked at the stuff from Ben Stopford which Kerry has 
referenced a few times? It's technology-independent, but 
covers things like the difference between shared nothing and
shared everything, and highlights once again that what's old is 
new again.

1) If it's shared nothing it is not a meaningful cluster. Even
'sharding' requires shared resources from time to time (or a
system rebuild). Check it out.

2) Ben Stopford was a long time employee of the Royal Bank of 
Scotland, which in 2012 culminated a year of poor IT performance
with a month long service outage following a botched upgrade 
(not VMS, fwiw). See e.g.
https://www.theregister.co.uk/2013/06/21/rbs_chernobyl_one_year_on/

3) RBS's internet banking is still switched off and on again 
overnight every night, and has been since 2011.

Ben writes some interesting stuff, and has clerly worked for a 
business that has some experience with unresolved IT-related
issues.

In some ways the most amazing lesson is that RBS is still in
business, despite their IT failures.

On a related note: Ben seems to have rediscovered 'graph
databases' as though they were a new thing. Various people,
me included, knew them by a different name in the 1980s 
when similar concepts were used for the internal 
representation of programs and data in automation systems 
(high end PLCs etc). 

When automation systems don't run right and they're in 
control of (say) a car assembly line, somebody gets to hear 
about it.