Emulator stopping during mnesia writes
Ulf Wiger
etxuwig@REDACTED
Thu May 11 12:56:30 CEST 2000
On Wed, 10 May 2000, Klacke wrote:
>We at bluetail have had some serious troble with disk io too. In
>particular loads of disk operations that modify a directory entry
>such as open/creat, unlink, rename
>
>What makes the situation a lot worse is that while disk io is
>pending, the emulator can't run no reductions at all and the overall
>performance drops to an almost grinding halt.
>
>It also turns out that different file systems are better or worse.
>In particular FreeBSD with Soft updates performed well whereas
>UFS/Solaris and FFS/BSD performed bad (They are almost identical)
At AXD 301, we've also had some problems with disk I/O.
One of the things we uncovered was that Solaris likes to handle
/tmp as a virtual partition (unless you specify that you want it
to be a true disk partition). This can cause conflicts between
swapping and disk I/O to /tmp, which can lead to some magnificent
delays.
Another problem we've had is when testing an IDE drive, as a
replacement for our current SCSI drives. Access times were
significantly slower, and (I believe in combination with a few
driver or HW bugs) could lead to whopping delays in disk access
-- which of course caused the whole VM to hang.
We've found ways around these problems, but would very much like
to see a modification of the current I/O implementation.
BTW, in my own pet project - a web server for source code browsing
in ClearCase, I've had great difficulty with the ClearCase client
sometimes causing delays of > 1 minute*. It's not a great feature
of a web server to get stuck for that long, servicing a single
request. My solution was to create a number of slave nodes and
spreading out the file requests to those nodes. This seems to
work reasonably well (although I had to increase (!) net_ticktime
from 1 minute to 5 in order to keep the distribution going.)
* For those of you unfamiliar with ClearCase, the ClearCase client
hooks into the file I/O system, so if the ClearCase server is bogged
down, your file ops will hang.
/Uffe
--
Ulf Wiger, Chief Designer AXD 301 <ulf.wiger@REDACTED>
Ericsson Telecom AB tfn: +46 8 719 81 95
Varuvägen 9, Älvsjö mob: +46 70 519 81 95
S-126 25 Stockholm, Sweden fax: +46 8 719 43 44
More information about the erlang-questions
mailing list