[Info-vax] The Future of Server Hardware?

[David Froble]

JF Mezei wrote:
> On 12-10-02 18:53, David Froble wrote:
> 
>> For the disks, the MicroVAX 3100 systems had a rather stupid 50 pin SCSI 
>> controller, and the speed was, well, "not speed".  Now, not that I was a 
>> fan of the BI bus, but the 6000 class systems had much better 
>> capabilities at moving data.  CPU speed is not everything in a system.
> 
> But today, does an IBM mainframe use a bus that is faster than
> PCI-Express ?

This I don't know.  My thoughts are that with the prices of memory today 
that massive amounts of data are cached in memory, and memory accesses 
will be much faster than any disk I/O.

I remember the time that some entity was using a VAX with a large (for 
then) amount of memory to load the entire database into memory.  I 
believe it may have been some govermental system for massive inquirirs 
into the database, perhaps driver license, perhaps criminal records, 
something like that.

> In terms of disks, wouldn't mainframes and commodity servers use the
> same type of disk array interfaces ?

Perhaps, but then if you're using storage arrays that allowed many 
connections, why not use cheap systems?  The systems aren't doing much 
while the storage arrays are providing the data.

Perhaps some of these people building their own custom systems can set 
up large memory arrays that can be shared for read by multiple systems.

> Or is this a case of modern mainframes using the same types of
> interfaces, but surporting multiple instances whereas comodity servers
> only have one PCI-Express and one interface card to FC or whatever type
> of interface to a disk array ?

Don't know.  Perhaps one interface card is enough for each system.

It seems to me that much of what Google does is provide an online 
library of information.  Most of the work is finding and retrieving data 
and presenting it to an inquiry.  This to me doesn't seem that it would 
present much of a load to a CPU.  Could be wrong.  But there isn't, as 
much as I can understand the process, any number crunching going on.

Now you can argue that using larger systems to do this work could be 
more economical.  I suggest that is maybe what an entity such as Google 
is doing.  What would you call a system with a large number of "blades" 
plugged into some type of backplane that didn't need individual power 
supplies, that provided access to storage arrays, and access to 
networking, and such?

Perhaps these people are already using their own version of "mainframes"?

I assemble my own x86 systems.  Get a motherboard, CPU, memory, disk, 
optical drive, and power supply.  I've been known to test such without 
an enclosure, just sit them on a table, make the connections, and run 
them.  This can be done for a couple hundred dollars.  I'm sure that if 
I was purchasing them by the thousands that the price would be much less.

AS you've been told more than once, the cost of electricity  isn't the 
only thing in the equation.