Conference 7.286::digital

	VMS clusters can go more than 96 nodes.

							mike

Don't tell us, tell THEM !

    re .0
    
    Aren't they talking about UNIX clusters here, hence the 4 nodes is
    correct.

    There are a number of sidebars:
    	VMS Remains King of the Clusters
    	Clustering Coming to NT
    
    Although, the comparison in that table is of UNIX Clusters there is
    num erous mentions of OpenVMS clusters.  My point is now that we have
    their attention let's WOW them with what OpenVMS cluster has.  How about
    a FRIENDLY note to the editor indicating some of the features of OpenVMS 
    Clusters like NODE COUNT.
    
    Last article I saw 96 was the magic number.  What is the magic secret
    number these days?
    
    Would you take notice if you saw a table like this?  What would your
    reaction be?  Must be a 'typo' there at 96.  Technologies CAN'T be that
    far off??!!  If true, I'd be interested in finding out WHY this one
    individual's offers were so different.  Might be something I could use.
    
    Vendor No      Product	           No of nodes
    AT&T GIS       LifeKeeper FRS              8
    Data General   AV cluster                  4
    Digital        AdvantageCluster            4
    Digital        OpenVMS Cluster            96
    HP             Enterprise Cluster          4
    IBM            HACMP/6000                  8
    Pyramid        Reliant Cluster Arch        4
    Sequent        Symmetry 5000 SE100         4
    Sun            SPARCcluster PDB            4
    Tandem         HATS                        4
     
    I should mention that there is a additional column with the
    Interconnects.  Basically we're all the same with ethernet, token ring,
    FDDI and a few others.   
    

    
    Did we patent our clustering technology?  Do all these competing forms of
    clustering use completely different technologies?  Where are the
    lawyers when we need them?
    
    len.
    

len:

> Where are the lawyers when we need them?

  Microsoft, same as everybody else!

                                   Atlant

    Anybody else been following Doonesbury the last week or two ?
    
    "How do they get away with it? " (referring to previous comments about
    how Apple-like it is when you start up Windows 95...)
    
    "It says on the box `Supported by 3000 lawyers'".
    
    Or something like that. It loses a lot in the telling, you know.

    Do other people violate our patents?  I'd say yes, but you'd have
    to prove it in court.  Look at the cost (both in time and money) of
    various people fighting software patents these days.  I would think
    that most companies are afraid of uncorking the genie from the
    bottle...  
    
    				-John

    Did anyone take the basenoter's suggestion seriously?
    
    Has anyone written to the magazine to toot Digital's horn?
    
    I think it was an excellent suggestion and deserves to be followed up
    on.
    
    Joan
    

    re .last 
    thanks.  Is anyone from OpenVMS Cluster Marketing going to write the editor?
    
    re .1  
    According to SPD 2978 The maximum number of CPUs supported in a VMScluster 
    is 96.  I assume that CPU = NODE.  Therefore, what are you referring to?  
    Some unpublished number that we cannot tell customers?  
    
    Also, in the future could you refain from  statements like:
    
    	VMS clusters can go more than 96 nodes
    This just leaves the reader's hanging in the air and frustrated.  I'm 
    wondering what the number is.  If you know what the answer is how about 
    pro-actively sharing it and not require the reader's to waste time 
    requesting this additional bit of information, that really should be 
    readily accessible to us.  We're reading here to get educated.  Help us
    to maximize our time.
    
    	Regards,
    
    	 kam

	the maximum *officially supported* nodes in a cluster is as per
	the SPD: 96....the fact that larger clusters have been built is
	something that the cluster folks have historically been reluctant to
	discuss openly...why? my guess is they don't want to get stuck
	supporting the mess if it breaks something....however, a
	"dir/title="large"" in SPEZKO::CLUSTER yields 3176.1 which
	discusses a customer (unnamed) with a 150 node cluster and
	this note was entered in July 1992! so chances are there may be
	more and/or bigger ones out there by now.........

	_kelley

    
    
    I sent the note to the VMScluster product manager - he sent it along to 
    the PR people...  I'm sure something will be said...
    
    
    John

>    According to SPD 2978 The maximum number of CPUs supported in a VMScluster 
>    is 96.  I assume that CPU = NODE.  Therefore, what are you referring to?  
>    Some unpublished number that we cannot tell customers?  

The supported limit is still 96 (that number was based on a bug in
MOUNT/CLUSTER in VMS 5.2 that caused crashes at above 96 nodes, and although
the bug has long since been fixed, the number has never been raised).  As far
as we know, the "real" limit is 224, based on how some data structures are
indexed within a piece of PEDRIVER code.  The rest of the cluster code is set
up assuming a limit of 255 nodes.

I worked with a customer who built a cluster that peaked at 151 nodes.  I knew
of half-a-dozen clusters within Digital that were in the 110-to-125 node range
in the same timeframe.

    Any chance of getting the PR people to inform us what the action will
    be, as a reply here?  If a letter to the editor gets published I might 
    miss it.  If something gets published I'd like to obtain reprints of 
    the article and an update in the letters to the editors to educate our 
    Business Partners.
    
    	Regards,
    
    	 kam

    You can get all of this info & more in the various Cluster evaluation
    reports being published by the major indsutry analysts/consultants.
    This article was about UNIX Clusters, and the Digital person who did
    the interview is Roy Shiderly, who works for me.
    
    If you have any inputs/suggestions as to the accuracy/positioning of
    Digital's UNIX Cluster capability and the competition, please let me
    know.
    
    Regards,
    
    Mark Silverberg
    

I see this as a very positive article for OpenVMS. It's not perfect, but it's a
lot better than some of the gibberish we've seen over the past year or two.
I have no intention of quibbling that "my node count is bigger than your node
count". That would be churlish.

The article is about Unix Clusters. OpenVMS doesn't play in that market. Who
cares how many nodes we support? It's much more important that Digital Unix
clusters can be seen to be better than (or have caught up with) HP/IBM/DG/ATT
Unix clusters.

What comes through clearly is that Digital is THE clustering company.
We get quoted as building Windows-NT clusters, we're clearly in the game
with Unix clusters (with good comments about Memory Channel), and we rule the
roost with OpenVMS clusters. What more could we ask for from the press?

Nick Carr
OpenVMS Clusters Product Manager

    For reference purposes:
    
    In the RS/Magazine 1993 now called the RS/Power PC Magazine there was
    an excellent article:
    
    	The Cadillac of Cluster by Thomas Casey and Richard Cameron
    
    Basically, it's about Clusters based on the RS/6000s,  HACMP/6000 disk
    arrays and advanced networking provide a fast but safe ride for
    mission-critical applications.
    
    ...
    As clustering receives more attention, definitions of a cluster abound. 
    Individual vendors slant their definition to reflect the capabilities
    of their product.  Still, the various definitions share some
    principles.
    
    In general, a cluster is a loosely-coupled collections of independent
    computers (nodes) organized into a network for the purpose of sharing
    resources and communicating with one another.  Each node runs its own
    instance of the operating system.
    ...
    
    The first results of this group was the SP1, the Scalable POWERparallel
    System, offering eight to 16 processors per frame.  The future of
    clustering could encompass four to 16 nodes, each with eight to 16
    processors.  Now we're looking at the processing power of 32-256
    RS/6000 working together, sharing memory, disk resources and networking
    interfaces.
    
    ...
    
    re .-1
    
    That's an interest statement[s]: 
    	The article is about UNIX Clusters.  OpenVMS doesn't play in that 
    	market."
    
    Therefore, if a Digital Reseller reads the article and asks about it I 
    shouldn't have said "running OpenVMS with the POSIX component you could
    compete quite effectly against any one of the UNIX Cluster configurations. 
    You have POSIX (UNIX) and you have the REAL Cluster."
    
    We would rather differentiate UNIX clusters from OpenVMS clusters and go 
    through the pains of now getting them to migrate to Digital UNIX and 
    Clusters rather than the competitors box and clusters.
    
    I guess I'm confused.  I thought Digital went through the efforts of
    getting POSIX branded so that we could say that everything UNIX has
    OpenVMS has and MORE.
    
    Bottomline is that no one in Digital is interested in writing articles.

    -re -.1
    
    >>> bottom line....
    
    Yes, without telling a story, no one will hear, without the story
    you are not in the game. We are clearly not in the game of marketing
    our products. It has been demonstrated time and time again, that we are
    not making marketing noise, therefor we loose.
    
    -Mike Z.