[email protected] (R.S.) writes:
> So, 12 years old FICON is simply obsolete and we should expect
> something new? And don't tell me about FICON enhancements, ESCON was
> enhanced as well!
> So? Shall we expect wi-fi based channels?

FICON is features on top of FCS (fiber channel standard). FCS started in
the late 80s at LLNL ... it had non-blocking switch with serial copper
interface. and LLNL effort to standardize non-blocking switch with gbit
serial fiber optic (concurrent gbit transfers in both direction).

ESCON had been fiber interface that had been knocking around POK since
the 70s. About the time ESCON appeared, one of the RS6000 engineers
worked on taking ESCON specs, making it full-duplex (concurrent transfer
in both directions), media speed 10% faster than ESCON ... and
significantly less expensive (as well as reliable) optical drivers. This
was the SLA (serial-link adapter) feature on the RS6000. He then wanted
to start work on enhancing SLA (effectively enhanced ESCON) from
220mbits to 800mbits.

We had been working with LLNL and convinced the engineer to join the FCS
committee rather than doing a proprietary 800mbit SLA (SLA started out
being enhanced ESCON).  Then in the early 90s, some POK channel
engineers started participating in FCS committee meetings ... wanting to
layer some unnatural IBM channel features on-top of the base FCS
standard (which eventually appears as FICON). There are a lot of FCS
mailing list from the period over the conflict layering FICON features
on top of FCS.

A major part of FCS standard was optimal cost/effective driver speed
available at the time ... for commodity priced standard (aka 1gibt/sec,
full-duplex, simultaneous transfer in both directions) ... and
full-duplex operation used for helping mask end-to-end latency (aka ibm
channel protocol operation with half-duplex paradigm was becoming
more&more latency limited ... dead-time on the channel waiting for
end-to-end turn arorund). Once FCS addressed the end-to-end latency
issue ... then changes in media speed was what-ever commodity parts
supported at any point in time.

In that time-frame, IBM had developed its own serial-copper for disk
operation (internal name harrier out of hursley) and announced as
9333. It ran 80mbits/sec ... full-duplex (aka concurrent 80mbit/sec in
both direction) using "packetized" SCSI commands ... aka rather than
half-duplex SCSI bus signaling ... the SCSI commands were encapsulated
in message packets and transmitted asynchronously (using effectively
identical SCSI disks ... harrier had significant higher aggregate
thruput than equivalent scsi disks in a half-duplex scsi-bus
configuration).

We tried to get harrier enhanced to interoperate with FCS ... multiple
fraction of FCS media ... still using serial copper ... but running into
FCS non-blocking switch and speed-matching handled.  Instead, it morphed
into "SSA" (running at 160mbits/sec in each direction) ... and
non-interoperable with FCS.

old post mentioning SSA & FCS 
http://www.garlic.com/~lynn/95.html#13

also references early Jan1992 meeting in Ellison's conference room doing
loosely-coupled cluster scaleup ... with proposal to have 128-way
configurations shipping to customers by ye1992.

128-way ye1992 would both support numerical intensive environments
... but also use FCS and non-blocking loosely-coupled access to massive
disk farm in large-scale DBMS operation (mainframe DB2 people started
whining that if I was allowed to continue, it would be light-years ahead
of where they were).

old email about cluster scaleup ... including several references
to LLNL
http://www.garlic.com/~lynn/lhwemail.html#medusa

-- 
virtualization experience starting Jan1968, online at home since Mar1970

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to