Re: [ARMini-support] Slow Transfer of Large Files on ARMx6 with LanMan98 / Sunfish

  • From: Michael Ben-Gershon <mybg@xxxxxxxxxxxxxxxx>
  • To: armini-support@xxxxxxxxxxxxx
  • Date: Wed, 11 Jan 2017 19:38:09 +0200

On 11/01/17 18:42, A Rawnsley wrote:

We're doing some significant testing about this at three different
premises at the moment, so I think any suggestion that we're not taking
this seriously is rather harsh.


I don't have all the results in yet, but we're already seeing significant
variation.  And the key here is, it is variation based on *source*.

(snip)
I felt these results were very interesting, as it suggests that the
sophistication of the source unit is very important.  The Intel NAS is
advertised as a much faster/better unit, and that's what we see in
practise.

It is worth noting, however, that neither unit produced "poor" results,
and nothing that fits with the chronically slow results reported by
Michael.



Now, word just in from Richard @ Orpheus who was helping out....

(snip)
As with any test I appreciate that without the same equipment and
replacing each bit one by one it is very hard to tell where a problem
may lie. So I hope this data will be viewed in the correct way.

(end quote)


What I think we're seeing here is that different source devices behave
differently, but past experience has also shown that for a given piece of
hardware, cables and switches also make a difference too.  Today's info
shows a variance between different sources wrt how much hourglassing
occurs, but nothing that matches the results showed by Michael's setup.


(snip)

The fact that Iyonix behaves differently to ARMx6 does underline how different devices may behave.

Running LanMan98:

My problem is that the linux server (running samba) reads and writes fine to and from PCs and Iyonix.
The PCs read and write fine to and from each other, linux and the Iyonix.

The ARMx6 is the one which is different here: It writes fine to all the machines, but its reads are VERY slow to both linux and PCs. I can't get ShareFS working reliably between the Iyonix and the ARMx6 in order to test that option.

Sometimes the ARMx6 will read very fast just after it is switched on, for a number of reads. Unpredictable though.

But, you also need to remember that LanMan works fine, whereas LanMan98 does not. Why? Why is LanMan fast when LanMan98 is not? This is the weird thing. And this is why I thought that getting the different developers together might help.

Something is being done differently in LanMan98 versus LanMan which is triggering this speed problem.

Of course, LanMan has progressed enough to mean that it could be used instead of LanMan98, except for its problem with large files (which looks from other sources as it concerns the last 64k and). It has another problem with directory names containing both a 'dot' and a 'space' on the server (the dot is converted to a forward slash on Risc OS).

Michael


--
Michael Ben-Gershon
mybg@xxxxxxxxxxxxxxxx

---
To alter your preferences or leave the group, visit //www.freelists.org/list/armini-support
List-related queries to info@xxxxxxxxxxxx

Other related posts: