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

Re: THREDDS slow with lots of files aggregated?

Hi Bob:

1. What is the aggregation type? union, new or existing?

2. Are you sure you are asking for data for "only one file"? If you are asking 
for data across the aggregation dimension, then the TDS has to open all the files to 
satisfy it. If so, it looks like a few msec per file, which seems in the right ballpark.

3. How are you doing the data access? An OpenDAP request? From what client?

BTW, you might want to use our new thredds "esupport" mail address: 
address@hidden for these kinds of questions

Bob Simons wrote:
We have a THREDDS server running on Linux here at ERD serving a lot of data sets (http://oceanwatch.pfeg.noaa.gov:8081/thredds/catalog.html). The data sets are aggregates, created by putting lots of individual files in each data set's directory.

The problem is: I have noticed that the time to get data from 1 data file in a dataset is roughly proportional to the number of files in the directory. And access to data in directories with lots of files is very slow.

Here are the results from a test in the order that the subtests were done (in one run of the test program). The GAssta hday subtest was added to test the theory that the number of files in the directory was correlated with the thredds opendap response time.

AG ssta 3day: 190 files, 719 ms
CM usfc hday: 2138 files, 13359 ms
GA ssta hday: 1018 files, 7063 ms
MB chla 1day: 185 files, 625 ms
QN curl 8day: 537 files, 3141 ms

That looks like a great correlation to me.

Any idea why THREDDS is so slow at opening one file? Linux is slow with so many files, but not close to this slow. Is there anything we can do to Linux or THREDDS to improve this?

Thank you.


Bob Simons
Satellite Data Product Manager
Environmental Research Division
NOAA Southwest Fisheries Science Center
1352 Lighthouse Ave
Pacific Grove, CA 93950-2079
<>< <>< <>< <>< <>< <>< <>< <>< <><


To unsubscribe thredds, visit:

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.