Conference 7.286::digital

    
    Yes, this is the way the market is going... first desktop PCs became
    commodities.  Next it will be servers.  How do we differenciate you
    ask?  Software development in enterprise manageability is key.  Then
    another one would be marketing expertise.  Can we get there? Time
    will tell.
    
    gb
     

Re: .1

>>> Software development in enterprise manageability is key.

For the moment, I agree.  However, this differentiation last only a limited 
time!

Intel has already announce that it will be adding server management 
features, such as temperature and fan sensing to it generic motherboards. 
CA or Microsoft will eventually own the enterprise management software 
market.  Once this happens, we need to move on to the next level of 
differeniation.

At one time, our desktop PC were once differentiated from the rest of the 
market.  The PC market eventually caught up and passed us, but we didn't 
move onto the next level.  IMHO, Dell/Micron/Gateway now builds a much 
better desktop PC then the Celebris/Venturis.


							-Paul

    Maybe purchasing this 4/8 processor motherboard make the most
    engineering sense.  I remember reading, EE Times I think, that the 
    Pentium Pro design was based on a joint effort with Sequent and that 
    Intel incorporated its own version of a complex, multiple-transaction 
    SMP bus into the Pentium Pro.
    
    Moreover, that designers have the opton of licensing the bus that was
    implemented on the chip or design a proprietary bus that would link to
    the Pentium Pro.  Supposedily this effort is a bit difficult and
    time-comsuming.  And that many companies are opting not to undertake
    this efforts.
    
    Also, developers that choose to use the P6 bus or even the concepts
    must license a package of Intel intellectural property.  The license
    term requires that you clearly define what it wants to build off the P6
    bus AND that you must share this technology with Intel.
    
    In the end, this provides Intel with the rights to integrate it into
    future chip sets.  Some developers are worried that their work on
    eight-way SMP might be sucked into Intel's Merced-generation of
    processors.  However, this does invest in the P6 core for the future.
    
    I believe that Corollary and NEC or NCR are developing their own
    proprietary ASICS to get around this.  Remember Corollary??? The
    provide the bus for the applicationDEC 433MP.
    
    Maybe Companies are using Intel's motherboard until they can further
    define this behavior.
    
         Regards,
    

    The interesting question here is the "how do you differentiate?" The
    short answer is that you don't...you concede a portion of the market,
    in this case the low end (where that definition is definitely very
    broad). IMHO the differentiators which IBM, HP and WE must find a way
    to market effectively have to do with overall system balance and
    throughput capabilities. 
    
    Does anyone really believe that a PC, no matter how configured, can do 
    the total workload of a mainframe or a good mid-range system? (Cheat Sheet 
    follows - the answer is NO) So the issue is...does the application 
    environment place demands on the system that require that it be robust, 
    balanced and capable of large-scale I/O? If not, go to the PC answer. If 
    YES, then stick with an Alpha or other larger-scale system. In many cases,
    we are doing the customers a disservice by letting PC's eat our lunch in 
    application environments that cannot be properly serviced by that class of 
    machine.
    
    But, that's just my opinion. I could be wrong.
    
    -OFWAMI-  

RE: .4

>    Does anyone really believe that a PC, no matter how configured, can do 
>    the total workload of a mainframe or a good mid-range system? (Cheat Sheet 
>    follows - the answer is NO) 

Pardon me, but I must disagree with that answer.  And, with all respect,
your answer is why Digital (and almost every other server vendor) missed
the PC market, and why Compaq is eating all of our lunches.

Let's define the difference between a PC and a mid-range server (taking the
AlphaServer 4100 as the example):

1) They both use PCI as their external device interface, so they both take
   the same Ethernet/FDDI/graphics/SCSI/etc cards.

2) They use similar speed and types of memory.

3) They have processors which are similar in speed for server operations
   (yes, Alpha is much faster for floating point operations but this is
   not what servers primarily do, and the edge that Alpha has over the 
   Pentium Pro or even P6 in integer operations is depressingly small).

4) The backplane speed if the 4100 is 1.1GB/sec, while the backplane speed
   of the Compaq is closer to 200MB/sec.  

So there are virtually no differences in 1 and 2, a slight performance edge
for Alpha in 3, and a great performance edge for Alpha in 4.  What this
translates to in real life is the ability to run more I/O at full speed on 
the Alpha than on the Compaq.

But from a customers point of view, why not just buy another Compaq?  They
are cheaper than 4100s, and once you have one of them working successfully,
it is really easy to go to upper management and say "That one works, but
demand for services has increased to where we need another one which will
just plug in and work", and have them believe you and fund it.

In my opinion, the low-end *AND MID-RANGE* server market for non-Intel 
products is gone.  We shouldn't waste our time developing more low-end 
products for that space, and we should think hard about the mid-range,
because we will lose every time on price to Compaq and their friends.  This
announcement simply accelerates that process.

The advantage that Alpha has right now is several things:

1) Applications which run on Digital Unix and OpenVMS which don't run on NT.
   We have a few more years before this vanishes, but not too many.

2) C2 and B1 security ratings, which NT does not have (I believe they are
   being evaluated for C2, but the wheels of the NCSC grind slowly).  I 
   could be wrong about this, and expect to hear the announcement of C2
   security for NT any time now.

3) The reputation for reliability that OpenVMS and VMSclusters has.  Many
   people around the industry are not yet ready to bet their business on
   NT, people such as the airline's SABRE system, or the Hong Kong Jockey
   Club, or nuclear power plants, or hospitals with systems that maintain
   patient life support, etc.  I expect this advantage to last less than
   10 years, as more and more people adopt NT.

4) The physical number of slots for I/O in the systems (I believe the largest
   Compaq system has 16 slots, while the 8400 has more than 100).  This leads
   to excluding NT for **BIG** storage arrays (>1TB).  This is true for both
   Digital UNIX and OpenVMS.

5) And the best one: 64 bit addressing.  We can put **LOTS** of memory in 
   Alpha systems (14GB today, 28GB real-soon-now, and it won't stop there).
   This leads to dramatic improvements in getting certain classes of jobs
   done, such as the Oracle demo of late last year.  Again, this is true for
   both Digital UNIX and OpenVMS.

All of our advantages are on the high end!  When we are competing for the
20-50GB storage system with 256MB of memory for the several hundred office
users, Alpha just edges out Compaq in performance, loses on price, and loses
big-time in market acceptance and support.

-- Ken Moreau

P.S. I consider Digital UNIX TruClusters and NT clusters as equivalent in
     terms of capabilities, reliability, failover, growth capacity, etc.  
     Both of them will catch up to VMSclusters in 2-5 years, but right now
     neither is there yet.  This is not slamming those products, it just
     reflects their current stage of development.  Therefore, I don't think
     that TruClusters has a significant advantage over Compaq NT clusters,
     which means that Alpha and Intel are equivalent in this space.

    Re: power station control (in general, nuclear excepted): this is one
    of the areas I support. There is a _stampede_ to NT. I happen to think
    it's unwise from a manageability and serviceability and support point
    of view if customers have got accustomed to OpenVMS or Digital UNIX.
    
    But our application partners have made the decision, and it is that
    they will have an NT offering. It therefore follows, given what has
    been said before, that they will offer NT on Intel. It therefore
    follows that, because of sales metrics, the business will go to someone
    other than Digital, because no-one outside the PCBU gets credit for a
    PC sale.
    
    Bob, in public, talks about a "dual Intel/Alpha strategy". That may be
    the engineering strategy but it isn't the reality on the streets that
    our customers see, because it isn't what the salesforce have as
    metrics.
    

re.6:

>therefore follows that, because of sales metrics, the business will go to
>someone other than Digital, because no-one outside the PCBU gets credit for a
>PC sale.

That's what happened last year. This year direct sales get rewarded for Alpha &
Intel server sales. It's the desktop piece that's still missing.

You are right about visability of our dual platform strategy though, we need to
see some corporate product advertising that shows the whole server range in one
ad.

Dave.

  

>    Therefore, I don't think
>    that TruClusters has a significant advantage over Compaq NT clusters,
>    which means that Alpha and Intel are equivalent in this space.

     For the record, Compaq doesn't have an NT cluster offering. And
     won't until Wolfpack ships next year. They do have something
     called Standy Server which certainly does NOT measure up to
     TruClusters.
     

    Re: .8
    
    Besides Standby Recover Server they also have On-Line Recovery
    Server & team with Tandem for ServerNET (though, they don't have
    anything to offer yet as Tandem's stuff is so new - though,
    Compaq implies in their presentations that they're teaming with
    Tandem because of their proven, almost de fact industry standard
    approach ;-)
    
    ...petri

Re .5: Windows NT was evaluated for C2 a year or two ago. If you have a Jensen 
or one of a couple of Compaq models which escape me right now, you can have C2.

PJDM

     I think that the creation of a commodity based motherboard has
    broader implications.  Apparently 32 bit processors have hit
    a speed wall at 200 MHz.  It is inevitable that multi processor
    systems will replace the current desktop machines.  I have already seen
    examples of this in the motherboard market.
    
    One should not only consider these as server motherboards, a standard
    4 processor motherboard should be able to penetrate the desktop
    market.  This will take pressure off Intel to produce a faster or
    better microprocessor.  Specifically it will allow them more time
    to remain at 32 bits.
    
    

	There is no such thing as a "speed wall" in the world of
	computers. Every time someone says that, they are ALWAYS
	proven wrong.

	As for a 200Mhz "speed wall" with Intel processors, that is
	totally bogus. Intel is NOT resting on its laurels.

							mike

    >Apparently 32 bit processors have hit the speed wall at 200Mhz.
    
    You've been brainwashed.

200Mhz speed of Intel systems is not a wall, but it is a definate plateau, 
which is expected to last until January '97.  This is due to several 
reasons:

	- Intel needs time to move to the next level of CMOS.  Current
	  yields of P5-200 parts in currenet technology are very low.
	- Software developers need time to incorporate MMX instructions.
	- Feedback from vendors asking Intel to slow down new CPU          
          introductions since rapid changes in CPU speed make customers
          delay their purchases decisions.


						-Paul

    P5 is essentially obsolete; only used these days for home market,
    notebooks, and Alpha performance comparisons.

Yesterday, we heard the CEO of Tandem speak at CA-World.  One of his
main points was that the industry would standardize on motherboards
from Intel with 4 processors.  He predicted great cost reductions 
(commoditization) for this configuration.

BC

    re .15
    Oooooooohhhh!
    That was especially nasty!
    8^)
    r