Conference csc32::consolemanager

I know of one setup that's of a reasonable scale... 


One customer has approx 300 ALpha servers (OSF/1 tho) as the managed systems, 
just to increase the load there's 20+ (running Pathworks and LAN managing local 
PCs themselves).  These are managed by a central 64Mb 2100Axp, and a backup/test
2100 which both run PCM, PSW and Netview.

Display is to VXTs served from an Infoserver.

They're also doing this across tcp and LAT with DECbrouters adn even conversion to
X25 by a Motorolla 6520. Oh and I think a bunch of DS90tls come into it... 

And with that setup although they've had a few problems I've not yet heard 
performance/ capabilities of the 2100 questioned. 

I think it's a great setup - every now and again I sit there and think 
it's amazing the capabilities we have nowadays really.   Daunting when 
you first try to get your head around the situation if you know 
you're going to end up supporting it, must've been interesting to install 
(one day I'll find out if there is a written summary of all kit involved 
and an accurate config map somewhere and get a copy since I'm sure it'd make
life easier knowing what you were trying to fix :-)     but it is amazing really.

    Sort of related...
    
    I know nothing about PCM so please excuse a naieve PCM question.
    
    A partner would like to know what config of system we would recommend
    to manage 48 console lines. An Alpha and a VAX solution would be good.
    If I should ask some pertinent questions then I will. Or point me to a
    sizing guide, I couldn't find one.
    
    Thanks
    
    Andrew

    There is no sizing guide. Do you have an ide of how much the customer
    wants to spend? A VAXSTation 4000-60 or 90 would work fine here. Be
    sure it has plenty of memory - at least 64 megs.  You would get better
    performance from an Alpha - maybe a 1000 or 400 series to keep the cost
    down. Again, don't skimp on the memory.
    
    Regs,
      Dan

I need to size a system that will manage around 256 UNIX console from 
HP, Sequent, Sun, IBM and Digital.

This system must be able to grow to support a possible maximum 1400 
consoles, and must also be resilient.

I'm going to propose two dual 4100 systems in an ASE environment, but 
what I need is memory, disk and CPU sizing.

Anyone got any pointers?

Thanks,

Greg

    Just to make sure that you are aware of this, PCM currently supports
    a maximum of 200 consoles per PCM engine.
    
    There simply are no real-world sizing guidelines as there are just to
    many variables. For example, the more events thats are scanned for, the
    more CPU cycles required. If you have a lot of busy consoles then you
    need diskspace and I/O bandwidth and you want to use multiple disks
    for the active logging directories. If you have a need to keep a lot of
    console data online, then you need a larger archive disk. The more
    concurrent PCM users, the more memory you want to have. 
    
      Having said that, the 4100 systems your proposing should have the
    horsepower to handle 200 systems. Unix systems simply don't generate
    that much console output by default. Obviously, your customer may have
    applications that log a lot of application specific messages to the
    console. You could ask the customer to provide you with some data
    that give a reasonable idea of how much console traffic there is. They
    could make daily checks of the size of the logd daemons logfile to
    get some idea. Once the customer provides you with this, you can 
    start thinking about how many logging disks that you want. Since PCM is
    write intensive you want to select disks and adapters that can handle 
    this kind of environment. If you use something like RZ29's you should
    be able to have 40 systems assigned to each disk. If this winds up
    being too much, the customer can always add more disks later as it's 
    very easy to change the logging disk for any particular system.
    
      The archive disk should be separtae from any active logging disk and
    it should be large enough to contain the desired amount of archived
    console data plus 25 to 50% more. This latte is cushioning to insure
    that if the customer has a flurry of console traffic and they have to 
    archive one or more of the logging disks that they have the free space
    available. If it isn't obvious by now, the archive area is limited to
    a single disk, raid set or bound volume set. Thus all systems are
    archived to the same "archive area".
    
    As far as memory, I wouldn't want less than 256 megabytes in a system
    thats going to manage that many consoles.
    
    Good luck!
    
    Dan

When you say 'supports only 200 per engine', can you be more 
specific?

Is it 200 concurrent connections at any one time, ie it can monitor 
more, but only handle 200 actual console connections?

What would you suggest for the maximum number I could support? Just 
up the number of machines? ie 200 users per machine?

Thanks,

Greg

>When you say 'supports only 200 per engine', can you be more 
>specific?
    
    I stated in -2:
    
    
    >Just to make sure that you are aware of this, PCM currently
    >supports a maximum of 200 consoles per PCM engine.
    
    If the above doesn't make sense then I'll put it another way:
    
    Any one PCM engine may manage up to 200 consoles. This has nothing to
    do with the number of "users" that may use the product. PCM is licensed
    either according to the number systems that will be managed or as
    unlimited use which means that you can manage up to the limit of
    200 consoles.
    

>Is it 200 concurrent connections at any one time, ie it can monitor 
>more, but only handle 200 actual console connections?
    
    Theoretically you can configure *more* than 200 systems, however, it
    is not supported to do so. In other words, the PCM configuration editor
    would allow you to configure more than 200 systems but if more than 200
    are enabled, you are unsupported. It simply not in your best interest
    to even do so. I can see it now: A customer says that they only want to
    manage 200 systems but want to keep 400 in the database so that they
    can enable/disable systems at will and then they complain when they
    realize it's not as simple as it sounds. The only way to enable/disable
    a system is to use the PCM config editor to change the state of the
    enable/disable button. Once done you must save the new configuration 
    and tell the software to reconfigure itself. The latter is quite
    intense when your talking about 200 systems. 

>What would you suggest for the maximum number I could support? Just 
>up the number of machines? ie 200 users per machine?
    
    The 4100 should be able to support 200 console connections quite
    easily. Here in CXO we have a system manager managing 200 consoles of
    various devices and he's using a VAX 6520 for the PCM engine. It runs 
    just fine.
    
    Regs,
       Dan

Ok Dan,

So I can handle 200 consoles per PCM engine, can I run 2 PCM engines 
on a single box?

If not I guess a cluster of 2 4100 systems could manage 400 consoles.

Greg

>Ok Dan,
>
>So I can handle 200 consoles per PCM engine, can I run 2 PCM engines 
>on a single box?
    
    Nope. A PCM engine is a single system with PCM running on it. Yes, that
    system may be a multi-processor system but it's still a "single
    system".

>If not I guess a cluster of 2 4100 systems could manage 400 consoles.
    
    Right you are!

Regs,
    Dan