Mauro,
Just a final comment or two before I bore the pants off all the non techies !
In a previous life I was a system architect in a company developing high
performance computer products (non-Intel). The last product I worked on
used multiple EIDE drives in an array to achieve high disk bandwidths.
Having researched the subject quite thoroughly I am convinced that for
many high-performance disk applications (not all), multiple EIDE drives
and multiple controllers can offer comparable performance to SCSI at lower cost.
I believe there are only two things that SCSI can do which EIDE cannot -
1) Support longer cable runs (the bus drivers/receivers are better)
2) Better support multiple devices on a single bus (the new U-2 has
better B/W, and the "transaction" architecture allows periods of
dead time to be used)
However, I do not believe that either of these aspects is sufficient
to protect SCSI in the longer term - i.e. the cable length advantage isn't
that big (and Fibre does it better), and the relatively high price of
SCSI means that you can almost put in twice the EIDE infrastructure at
the SCSI price.
I believe that in the longer term, Fibre will become the connection
medium for inter-box, and EIDE (or derivative) will become the
connection medium for intra-box. In the short term, SCSI will not go
away, but unit volumes are unlikely to increase significantly, and
unit costs are likely to fall. This means there will be a continuing
revenue decline over the next few years, which, unless ADPT gets
dramatically more efficient, will mean continuing earnings decline (for
their SCSI business).
Mark |
