meteohub.de

My new system with Alix.3D3 has the following values.
recomputation of weather data finished: 184402 records processed in 278 seconds (663 records by second)
Two other tests proved 667 and 932.
These are hardly data on the map.

I find this, nevertheless, very little.
My old system with the NSLU2 was already slow under the average. Now also Alix System. What can I make there?

Greeting Mark


Mein neues System mit Alix.3D3 hat folgende Werte.
recomputation of weather data finished: 184402 records processed in 278 seconds (663 records per second)
Zwei weitere Tests ergaben 667 und 932.
Es sind kaum Daten auf der Karte.
Das finde ich doch sehr wenig.
Mein altes System mit der NSLU2 war schon unter dem Durchschnitt langsam. Nun auch das Alix System. Was kann ich da machen?

Gruß Mark

That looks below expectation. Possible reasons are:
1) slow CF card. What card is used?
2) There were multiple recomputations running in parallel (can happen when setting up weather stations and/or sensors)
3) very untypical setup (many calibrations, extreme number of graphs to be computed, etc)

Thank you for your answer Boris!

1) 4 GB ExtremeMemory 120 X performances CF map of Varia Store, other maps CF are not inserted there

2) these were single calculations

3) few graphics (13), nothing unusual, resolution partly 5 minutes, upload 5 minutes

V 4.4 , no Software-Firewall, Systemlast is down

I am just having Transcend 266x or industrial cards around here, but even 120x should be not that slow. Are there records you have marked as deleted in the raw data? That might keep the count low while data has to be read anyway.

I still have data (10/08-06/09) in the folder "export". They still come from the USB-Stick of the NSLU2 which I have there copied.
RAW data are not available any more.

ALIX.1D with Transcend Extreme Speed 300x SLC 4 GB CF.
Version 4.4: 5200226 records processed in 2959 seconds (1757 records per second).

Two observations:
Going from a from a Sandisk Ultra II CF-card to a Transcend Extreme Speed 300x SLC results in no noticeble performance difference, despite the Transcend being advertised as three times as fast. Most likely the bottleneck is the ALIX.1D and having a faster CF-card than 10-15 MB/second doesn't seem to affect recomputation performance, at least not for me. I'm still not disappointed as my reason for changing was to increase reliability (SLC-memory) rather than speed.

In my installation 4.4 is a bit slower than the previous versions I looked at. Possibly it's because I'm know using the system data options?

It appears to me that the recomputation process has improved quite a bit, these are my numbers over the last couple of months:
wmr928fulleval(04.06.2009 05:27:00): recomputation of weather data finished: 1317545 records processed in 7005 seconds (188 records per second)
wmr928fulleval(22.06.2009 12:13:53): recomputation of weather data finished: 1906184 records processed in 13717 seconds (138 records per second)
wmr928fulleval(25.06.2009 13:00:32): recomputation of weather data finished: 2004352 records processed in 15308 seconds (130 records per second)
wmr928fulleval(30.06.2009 14:17:11): recomputation of weather data finished: 2167163 records processed in 14209 seconds (152 records per second)
wmr928fulleval(20.07.2009 18:10:48): recomputation of weather data finished: 2818637 records processed in 22521 seconds (125 records per second)
wmr928fulleval(24.07.2009 18:30:48): recomputation of weather data finished: 2948314 records processed in 13819 seconds (213 records per second)
wmr928fulleval(02.09.2009 21:01:32): recomputation of weather data finished: 4266103 records processed in 13256 seconds (321 records per second)

This is on an NSLU2 (266) with unchanged hardware, currently running 4.5, so as far as I can tell Boris has done some magic to run things more efficiently!

Thumbs up!

I agree. At the latest two recalculations I've seen well over 2000 records/sec. Previously it was around 1700-1800.

ALIX.1D with Transcend Extreme Speed 300x SLC 4 GB CF.
Version 4.4d(=4.5): 6958889 records processed in 3262 seconds (2133 records per second)

Not observed here, with ALIX.1D

I find that recomputes vary quite a bit. The "after installing 4.5" had a lower records/sec than the previous one, but within (and towards the lower end) of the usual range.

...but my Meteohub does have a lot of sensors to deal with and is running a couple of plug-ins.

David

I too have seen a big improvement in recomputation speed after jumping from 4.4c to 4.5 on an NSLU2 (266).

wmr928fulleval(23.08.2009 11:34:29): recomputation of weather data finished: 5565262 records processed in 28123 seconds (197 records per second)
wmr928fulleval(31.08.2009 16:13:30): recomputation of weather data finished: 5774627 records processed in 24164 seconds (238 records per second)
wmr928fulleval(01.09.2009 13:56:48): recomputation of weather data finished: 5805068 records processed in 18381 seconds (315 records per second)

This speed increase will really help.