IBM, with the introduction of the File Server IOP (FSIOP) board for
the AS/400 signalled its drive for server consolidation by allowing
PC applications to be run under the covers of the AS/400. However,
the current version, the INS (Integrated Netfinity Server), is only
a 333 MHz single processor, which many users may consider
insufficient to run today's processor intensive Windows
applications, such as Terminal Server. What plans does IBM have to
address this situation?
There have been three names for the AS/400 adapter that has on
it an Intel processor and its own memory. Initially this was called
the 'File Serving I/O Processor' (FSIOP), writes Nigel Adams. The
name was clearly intended to convey the message that this device
was to be used for an AS/400 to do File Serving for attached
PCs.
There were a number of enhancements to the processor that was
offered on this adapter. The second name that was given to this
device was the 'Integrated PC Server' (IPCS). The renaming of this
adapter was intended to convey the fact that it now had the
capability to do far more than just file serving.
The most recent name change was to the 'Integrated Netfinity
Server' (INS), and this change was more of a branding change, in as
much as it was intended to show the affinity between IBM's range of
Netfinity PC Servers, and the technology that we have on this
AS/400 adapter. The latest version of the INS does indeed support
an Intel 333 MHz Pentium II processor with a maximum of 1 GB of
memory. It would be fair to say that this is not the most current
of Intel processors, and if one scans the computer press you will
certainly see adverts for PC Servers with processors that have much
faster ratings. However, this is only part of the story.
Firstly, raw processor clock speed is only one element in
overall system response time. We know this very well for the AS/400
itself - in complex business applications there are several
elements which affect response times, and this is not just a factor
of the processor speed. The speed with which data is read from and
written to Dasd, the bus speed, the speed of data transfer across a
Lan and or Wan, the amount of main storage available - all of these
contribute to overall response time. The processor speed is thus
just one element of overall response time, and exactly how
important that element is will depend upon individual
circumstances.
Secondly, it is true to say that it takes IBM time to develop
and test a new version of the INS with a newer, faster Intel
processor on it. Given the fact that the vast majority of AS/400s
are used by customers for core line of business applications, and
that as a consequence the overall reliability of the AS/400 is of
critical importance, it is vital that any new hardware adapter is
fully tested and reliable before IBM makes it generally available
to customers. A delay in coming out with an INS with the latest
Intel processor is, I am sure customers would agree, a price worth
paying for ensuring that each new version has the sort of
reliability that customers expect from their AS/400s.
Thirdly, and this is mentioned by the questioner, all iterations
of this adapter have until now been single processors. Could not
IBM improve performance by having an SMP version of the INS? I
would by no means claim to be an expert on PC Servers, but my
understanding is that SMP processors on these servers do not
produce dramatic results in terms of improved performance, and
certainly not the sort of benefits that we see from SMP versions of
the PowerPC processor on the AS/400. However, there is no doubt
that there are certain applications that would benefit from an SMP
implementation of the INS.
IBM will be addressing the second and third points that I have
mentioned above, and a Statement of Direction was issued early last
year in this area. This announcement was to the effect that IBM
will in the future be coming out with an SMP version of the INS.
This SMP version will take the form of a standard IBM Netfinity PC
Server, and this should allow IBM to address these two points.
Because this will be a standard Netfinity box, this will allow
IBM to get the technology to market quicker than can be the case
with internal AS/400 adapters; and because this is external to the
AS/400, this will allow IBM to offer an SMP version. This SMP
external version of the INS is not yet announced by IBM, and
information about has only been issued in the form of this
Statement of Direction.
However, it should be made clear that this does not mean that
when this external version becomes available that IBM will no
longer offer the internal version in the form of AS/400 adapters.
The internal version will continue to be offered and will be
enhanced. And there are several benefits that will continue to
apply to the internal version - for example the floor space saving
that results from these being standard AS/400 adapters that are
located in the AS/400 towers; and the fact that the maintenance of
these adapters is included in the maintenance costs of the AS/400
itself.
The external version will be an alternative, and will clearly be
of particular interest to customers whose applications will benefit
from SMP processors. However, the external version will offer the
same fundamental benefits as are offered by the INS today - the
easier administration, improved Dasd performance, AS/400 backup and
recovery capabilities, for example - which make the internal INS
such an attractive option today for AS/400 customers to run Windows
NT.
In ending, though I want to emphasise the first point that I
made above; and that is that even when the external SMP version of
the INS becomes available, it will not change the fact that for
very many applications the processor speed remains only a small
part of overall system response times.