Conference netcad::hub_mgnt

    
    Don't confuse what you're told you'll get on newly announced products
    with the discussion of what you _may_ get some time in the future.
    
    With all announced Ethernet-capable modules you can have at most 6
    Ethernets on the backplane of a HUB900 (with a little care you can
    count the thinwire as a 7th).
    
    This is not a software restriction - it is a limitation of the hardware
    interfaces for backplane Ethernets on the modules.  No software change
    can change it.
    
    _Theoretically_ future Ethernet modules _could_ be created that can do
    more, because the etch is there on the backplane (currently not used by
    Ethernet - we don't need to 'use less etch for more bandwidth', we have
    much more etch than we use) and the hub manager can handle the
    calculations to allocate it (already).  
    
    The problem is mostly that the cost of those extra drivers and
    receivers pushes up the per-port cost of the Ethernet modules. 
    

    What you can and can't do: Think of the DEFMM (12 port fiber) as
    having two levels of switching. Level one is grouping of ports together
    and level two moving groups of ports onto the HUB backplane. The DEFMM
    has 12 ports. The 12 ports are made  up of 6 pairs of ports. Each of
    these 6 pairs may be grouped together into shared lans at level one.
    For example, you could take ports 1 and 2 (pair 1) and group them
    with port 7 and 8 (pair 4) and call this group A. You can then LAN hop
    group A onto any the 6 supported Ethernets on the 900 backplane or
    just leave group A off the backplane as a standalone ethernet segment.
    
    The DECbridge 900MX will allow you to bridge between backplane LANs
    on the DEChub 900.
    
    HUBwatch V3.0 will support all this capability.
    
    Hope this helps.
    Bruce

    Based on the replies in .1 and .2, I did't make myself clear. I
    understand that the DECbridge 900MX only supports 6 Ethernets. However,
    if I have 2 DECbridge 900MX modules, I have 12 total Ethernets
    available to me. If I make them standalone, I can link the 2 of them
    together via the front panel FDDI A/B connectors and connect all twelve
    Ethernet together. So, now lets move the 900MX modules into the DEChub
    900 backplane. I then want to put each of the 6 Ethernets for EACH
    module onto its own TIC configured as an Ethernet channel for a total
    of 12 Ethernets across the backplane. Can I create 12 Ethernets? Why
    you ask?
    
    	I then insert 2 DECrepeater 900FL modules and using HUBwatch match
    each 900FL port-pair to an each Ethernet I've created on the backplane.
    At then end of each 900FL port, I put a DEFLM to a system to connect a
    total of 12 systems. Then, I create an FDDI channel across the
    backplane and connect the two set of 6 Ethernets together via FDDI.
    
    Is this possible under HUBwatch V3.0? If not, when?
    Is this clearer as to what I want to do?

>    Based on the replies in .1 and .2, I did't make myself clear. I
>    understand that the DECbridge 900MX only supports 6 Ethernets. However,
>    if I have 2 DECbridge 900MX modules, I have 12 total Ethernets
>    available to me. If I make them standalone, I can link the 2 of them
>    together via the front panel FDDI A/B connectors and connect all twelve
>    Ethernet together. 

Correct.

>    So, now lets move the 900MX modules into the DEChub
>    900 backplane.  I then want to put each of the 6 Ethernets for EACH
>    module onto its own TIC configured as an Ethernet channel for a total
>    of 12 Ethernets across the backplane. Can I create 12 Ethernets?

NO.

When you move the first 6 ethernets to the backplane (connecting each
port pair on the DEFMM to a bridge port) you use 6 flexible channels
as Ethernets.

When you try to move the 6 ethernets from the other DEFMM to the
backplane _YOU_FAIL_; the 6 channels the second DEFMM can reach are
the SAME 6 CHANNELS that the first one could reach and used up.  The
backplane ethernets are busses.

YOU CAN ONLY HAVE 6 BACKPLANE ETHERNETS (+ 1 thinwire).

>    Is this possible under HUBwatch V3.0? If not, when?

The problem is not HUBwatch - it is the module hardware.

>    Is this clearer as to what I want to do?

It is completely clear what you want to do, unfortunately you can't do
it.

    I'm still not getting the answer I think I want. I don't want to be a
    pain, but your answer contradicts the theory of the DEChub 900 which
    was presented at Network Academy.
    
    Is the DEChub 900 limited to creating only 6 Ethernets? I know each of
    the modules can only talk to 6 Ethernets due to MODULE hardware design.
    Are you saying that the DEChub 900 can only create 6 Ethernets? That
    was not what was said at Network Academy. The presenter indicated that
    the amount of etch available on the DEChub 900 will allow for creating
    many kinds of bandwidth.
    
    So, I'll try again. Imagine a 2 slot DEChub 900. In slot 1, I put a
    DECbridge 900MX and create 6 backplane Ethernets using the flexible
    channels. I install a DR900FL and tie each one of the port-pairs to an
    Ethernet. This is my FIRST 6 Ethernets. How many flexible channels have
    I consumed in doing this? How may flexible channels are left over after
    the 6 Ethernets are created? 

Let me try it:

>    Is the DEChub 900 limited to creating only 6 Ethernets? I know each of
>    the modules can only talk to 6 Ethernets due to MODULE hardware design.
>    Are you saying that the DEChub 900 can only create 6 Ethernets? That
>    was not what was said at Network Academy. The presenter indicated that
>    the amount of etch available on the DEChub 900 will allow for creating
>    many kinds of bandwidth.

First, consider the DEChub 900 backplane separate from the modules. The
*backplane* could support more than six channels, but the current
Ethernet *modules* cannot *use* this capability due to hardware
restrictions (saving a bunch of expensive chips). The Ethernet *modules*
can only use the first six backplane channels, whatever you do. The rest
of the backplane is available for other things, e.g. modules using FDDI
channels. 

>    So, I'll try again. Imagine a 2 slot DEChub 900. In slot 1, I put a
>    DECbridge 900MX and create 6 backplane Ethernets using the flexible
>    channels. I install a DR900FL and tie each one of the port-pairs to an
>    Ethernet. This is my FIRST 6 Ethernets. How many flexible channels have
>    I consumed in doing this? 

Six.

>    How many flexible channels are left over after
>    the 6 Ethernets are created? 

None that the Ethernet modules could use. Think of 'Channel' as an
abstract term. It exists only through a module that uses it. The mapping
to the 'physics', i.e. wires, is done at a lower level and not visible
to the user. 

Marketing likes to talk about the theoretical capabilities of the 
backplane to emphasize the potential for future growth. This does *not* 
mean that every module is able to use it *now*.

*Robert

    I think I now understand. The way that the modules connect on the
    backplane only allow access to a certain number of the flexible
    channels (from top to bottom, probably a simplistic explanation). So
    the previous noters were trying to point out that you need more chips
    in each module to access more flexible channels. 
    
    Given this, why can't the problem be solved by simply having variants
    which can use another group of flexible channels or make this
    selectable via HUBwatch? I'm not an electrical engineer, so I probably
    don't know how hard/easy this is to do. Or is this so hard-wired into
    the modules that this isn't possible? How does this compare with our
    competition? Can they create more that 6 Ethernets across their
    backplanes?
    
    I mean if NEW modules were switch selectable, we wouldn't have to add
    more chips to access more flexible channels. We'd just tell the module
    to attach to a certain group of flexible channels.

    There are special rules for how the modules are allowed to "touch" the
    flexible channels.  The rules will not allow any switching such as you
    envision.  If the module is _ever_ going to touch a flexible channel,
    the rules force us to buy and hard-wire in the expensive interface 
    chips which increase the per-port cost.
    
    It would be possible to make variants of the modules which use another
    group of flexible channels.  Note that the variant modules would not be
    able to connect over the flexible channels to the non-variant modules.
    All ethernet modules which now connect to the flexible channels connect
    to the same six channels, and can thus communicate with each other.

    We're moving the right direction now that I understand the interface.
    If we could make a DECbridge 900MX variant which could use 6 different
    flexible channels and the COMMON FDDI flexible channels, it would be a
    start.
    
    I don't know about anyone else, but with 8 module slots available,
    variants which use different sets of flexible channels could be a real
    differentiator for us. If these modules had some sort of physical
    switch or an internal connector which could switch the flexible channel
    chips to a different set of pins (which is how I understand the modules
    connect to the flexible channel etches on the backplane), could we
    reduce our engineering costs on creating variants and still allow us to
    use more flexible channels?
    
    If you go back to my original note, I'd like the ability to put 4
    DECbridge 900MX modules + 4 DECrepeater 900FL modules into a DEChub 900
    and have *24* seperate Ethernet channels linked externally thru a
    GIGAswitch via the front-panel FDDI connectors.
    
    Does anyone else envision this kind of need?

    
    *** Futures Speculations Warning ***
    
    We are aware of the kind of requirements that you're trying to meet,
    and we have looked into it.  There are a variety of approaches that
    have been looked at.  
    
    The cost sensitivity of the Ethernet repeater market is something I'm
    sure you appreciate.  The ability of our modules to operate either
    standalone or in a hub means that we have to be able to build them
    cheaply enough to sell standalone even though they incur all the costs
    of operation in a hub (all those backplane driver chips are unused
    overhead in a docking station).
    
    One approach (just to show that we have thought about this) would be to
    add an additional mux such that a module could drive any of (for
    example) 12 backplane channels, but no more than 6 at any one time;
    this would interoperate smoothly with existing modules because the hub
    manager can ensure that the more restricted modules get the channels
    that are common (it can already do this).  This would get you up to a
    theoretical max of 12 for the backplane.   
    
    Using the current Ethernet backplane scheme, the maximum if all signals
    were in use would be 15 (+thinwire), 4 of which could reach only half
    of the backplane (some signals are split between slots 4 and 5).  Other
    schemes might be possible, but the cost of building new modules to use
    them _and_ be interoperable with the now-existing ones might be pretty
    high.
     

    >>  ... If these modules had some sort of physical switch or an 
    >>  internal connector which could switch the flexible channel
    >>  chips to a different set of pins ...
    
    This is exactly what I was trying to say we cannot do.  The electrical
    interface rules for the flexible channels require us to hard-wire the
    interface chips to the connector pins.

    
    I would like to re-visit this note. Please review reply .10 which has
    the necessary speculation discussion. I have been told by a Network
    Sales "Warrior" that we are working on the ability for DECbridge 900 EF
    modules to be able to have more flexibility and thus allow us to create
    multiple set of Ethernets on the DEChub 900. Since my original
    discussion was in March and there was some speculation on this issue,
    has any engineering work been done to make this possible in the short
    term?