[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20050901: clock skew on striker2? (cont.)

>From: David Knight <address@hidden>
>Organization: SUNYA
>Keywords: 200508312300.j7VN0ojo025668

Hi David and Kevin,

re: PC clocks suck

>Even on a linux box running ntp?

The clock hardware sucks.  One _must_ run nptd to keep the clock
set correctly.  My comment was meant to compare/contrast the clocks
in "real" Unix machines (e.g., Sun SPARC, Compaq OSF/1, etc.) with
PCs.

>This is an issue for us since we might want to replace
>gusher with a linux box - I'll probably follow up with
>you later on this - think spring 2006 Unidata equipment
>proposal...

We have moved to PC hardware for the great majority of our machines.
Our current darlings are dual Opteron SunFire V20Zs.  They serve
as the data backends of our IDD relay cluster, idd.unidata.ucar.edu,
and are stellar performers... when they have enough memory.

re: we didn't look hard at the length of the latency period
>yep, that is what I thought.

I apologize, but it still does not change my opinion that the problem
is closer to SUNYA than to sites receiving the NLDN data.

re: we don't think the problem is some sort of internet thing
>Me either, but it could be, see below.

>If memory serves, the ingest process *is* running outside
>the LDM process group. A seperate process runs that collects
>the strike data, puts in 6 minute bins, then inserts it into
>the product queue where the LDM takes over.

That is how I remember the setup (from years ago).

>But this procedure has not changed since just before y2k.

Yup.

re: comparison of ping latencies
>OK I'll set it up.

I am eager to see the results!

re: we don't think the latency weirdness has anything to do with running
an older version of the LDM

>Good ;) I really don't want to change the LDM at the beginning
>of the semester, and during a hurricane!

I totally agree.  I stopped updating McIDAS installations on Community
Data servers during the hurricane.

re: By the way, this pattern is relatively new.

>This could be an important clue. According to our crackerjack
>UAlbany networking group:
>"
>> Nysernet will be removing some equipment in the NYC POP on Wednesday
>> night from 9-9:30.  During
>> the outage, I2 traffic should be diverted to the commodity internet
>> link.  Once the equipment is removed,
>> configuration changes will be made in existing equipment and we will be
>> reconnected.  If the outage lasts
>> longer than a half hour, they will contact me at home and I will send
>> you email.
>"

Interesting.

>Did the strangeness start at the same time?

I don't know.  We just happened to see the strange pattern today while
we were reviewing data ingestion at the University of Northern Florida.

>Maybe their new configuration has a QOS scheme that pumps data
>through if it has been delayed by 30 seconds.

We had discussed this sort of thing being a possibility.  We looked for
downstream non-NLDN feeds from gusher, but we didn't find any that
were also reporting real time stats.  How about I setup a feed
from gusher for some other datastream(s) so we have something to compare?

>I wonder if only internet2 sites show this periodicity.

Good question.  I will have to look to find a machine feeding
NLDN data  that is not on I2.

>Yep. Got to love a interesting puzzle, as long as it doesn't
>interfere too much with your real work.

Right.  It is always nice to have a non-critical puzzler to keep the
mind sharp, however.

Cheers,

Tom
--
NOTE: All email exchanges with Unidata User Support are recorded in the
Unidata inquiry tracking system and then made publicly available
through the web.  If you do not want to have your interactions made
available in this way, you must let us know in each email you send to us.