RE: Why is the iSeries so slow -- MIDRANGE-L


On Tue, 12 Aug 2003 Ken.Slaugh@xxxxxxxxxx wrote:
>
> At 3:30pm I started the GET command. Now it is 9:40pm and the file now
> contains 1,295,330kb. Facts or not, I'm guessing if I were to move the same
> file to the FTP share of this same PC and use another PC FTP client on this
> same network the throughput would be much better.


You're right, the iSeries is slow as an FTP server.   At dell I can buy a
$400 server-class PC system, I can put FreeBSD on it, and it'll be both
more stable and much much faster for this type of task than the iSeries.

However...   it's not THAT slow.  Your transfer has taken 6 hours to
transfer (approx) 1.2gb?!   Is this a CISC box?  If not, something is
wrong somewhere.   Maybe there's some flaw in the way your network is
configured, or something.   Even at 10Mbit, I can get over 700k on an FTP
transfer from the IFS.


> History seems to prove my point around here and I'm just trying to find
> out why the IFS of the iSeries can't keep up with reality.

See also the thread about why HPT is slow to an HP 4050 laser printer.
There's another place where it's not keeping up with reality.   Of course,
the answer is "use an AFP or IPDS printer."   and that suggestion is quite
correct, it will solve the problem, but should we have to do that?!  The
competition will print quickly to any type of printer.   Seems like
there's room for improvement here, too!


> FTP doesn't seem to be the only slow boat up the iSeries river. Simple drag
> and drop stuff is painfully slow as well and I've got three different
> iSeries boxes in-house and a long list of customer boxes to prove it.

This is absolutely true.


> I suppose if you throw enough money into the iSeries box then it may
> start to compete with the throughput of some FTP freeware product on a
> $800 Dell or Gateway PC. This shouldn't, of course, be the case and I'm
> looking for an answer to justify the differences.

Open-source, not freeware... but otherwise I agree.


> If the platform is more stable but not fast enough to perform the task in
> an adequate time frame, then stability is a moot point.

I think there's room for improvement in both speed and stability.  I've
certainly found my iSeries to be less stable than my FreeBSD-based PCs.

When the iSeries gets low on main storage, the system becomes horribly
unstable.   It just randomly crashes.   FreeBSD doesn't have that problem.
Sure, it's hard to use it (since there's no space) but it doesn't crash!
It just reports errors on the console.   Sometimes it kills processes that
are using too much space in order to make the system usable again so
the operator can fix things, which is annoying, but not as bad as the
random hard-locks I have with the iSeries... it doesn't even tell you
what's wrong.

Another example is hardware problems.  Over the past 10 years I've had 7
disk failures, 2 tape drive failures, 1 power supply failure, 1
shorted-out card cage on my iSeries.   Sure, the PC hardware has problems
too... but I've got a PC running FreeBSD that's had no hardware problems
over 8 years, which is a much better record.   If I do have a problem, I
can replace the whole system with a high-quality $400 dell server.

And, even a 486 running FreeBSD outperforms the iSeries for networking
tasks.  I have a 486SX-25 (really slow!) which blows away my iSeries 270
for DNS, Windows file serving, for FTP, and for HTTP.   That's a 1991
PC outperforming a 2001 iSeries.

But, I'm going to regret posting this, because I just know people on this
list are going to boo at me, and hiss at me, and tell me I'm wrong
(despite that these are my personal experiences).   This really is a bad
place to post things like this, since the bias on this list is very
strongly towards OS/400.

But, on the other hand, there are many good things about OS/400.   The way
CL commands work with their parameters, and prompting, and help is
excellent.  The integrated database is wonderful.   The ease of creating
applications with DDS screens and RPG programs that do all sorts of
complex business calculations is really unmatched.

And, all of that is great.   But, until people start to see where the
iSeries weaknesses are, they aren't going to be improved.   IBM really
needs to think about how to leverage the strengths and minimize the
weaknesses, and we won't do that if people continue to overlook it's
deficiencies.

Sigh...