Skip to main content

⚠️ Please note that this topic or post has been archived. The information contained here may no longer be accurate or up-to-date. ⚠️

Effect of Metadata on Filter Tool and Browser Speed

Comments

27 comments

  • SFA
    Eric,

    I think you may have found the key.

    Something only vaguely related that I saw in a log file a couple of days ago made me wonder if that (re-reading folder populations and, being similar if one is repopulating, calculating numbers for filters) could be a problem. However I am using Windows and Sessions and not experiencing the dramas reported for Catalogs and, it seems, mainly Macs.

    I wonder if these fairly large catalogs are mainly imported files? This perhaps more likely to have history of IPTC data that varies for most records. In theory that should not happen as far as I understand if soince the IPTC codes are intended to be controlled content and table driven. In practice ... who knows what might be in there?

    Can you imagine an index based on full Geocodes!

    Many many years ago I recall a similar sort of problem (although possibly the inverse of this - lack of information rather than too much) when undertaking a data conversion exercise for a new client.

    Our system worked with more information than their old system required so although most of the data could be assimilated and converted quite well some thing were reliant on unique identifiers and then being "distributed" around some location records for operational purposes.

    The test files for the routine worked well with the sample data set. We estimated an 8 hour process to convert, write tapes and so on. So they copied the files on a Friday night, couriered the tapes to us, we ran the process on Saturday expecting to verify the tapes with the output files on Sunday and send them off to be loaded so the new system was up and running on Monday first thing.

    The conversion ran for 4 days.

    It transpired that the Unique identifiers for the 10% or so of records that did not have accurate locations readily available and that were supposed to be 99% identifiable by unique serial number ... did not have serial numbers.(Or something similar to that - it was a very long time ago now and I forget the complete details.)

    The conversion routine have a great process for getting the required distribution checking for duplicates and so on ... provided the majority of the records had the promised serial numbers. For the few record thought so have missing serial numbers or "duplicates" (like "No Serial Number" where people could not be bothered to find and enter one) the process was designed to cycles through the records and connect them to a location that simply spread them out as equally as possible. It would have worked perfectly within the time planned for a few tens or even a few hundred records. For several thousand it meant that the process was constantly repeating assessing every record - much the same as a key field index finding there are no common keys and every entry has to be listed (and checked for at the next record) in terms of processing effort.

    We worked with the client on their data quality, agreed a revised process which introduce fake but unique keys that could be used in advance rather than after the fact of attempted conversion as they previously insisted, and re-ran the exercise a month or so later well within the cut-over plan timing for the weekend.

    One of the more memorable projects I have been involved with. The conversion exercise was by no means the end of the issues they had not realised they had with the data available from their previous records. We had some interesting weeks getting that particular project up and running fully.

    The problem you described in your post would appear to be the inverse of the one we had - more or less. But the result will, in effect, be very similar.


    Grant
    0
  • Eric Nepean
    Hi Grant

    [quote="SFA" wrote:
    Many many years ago I recall a similar sort of problem (although possibly the inverse of this - lack of information rather than too much) when undertaking a data conversion exercise for a new client.

    Nasty problem you describe. Those kind of things cause gray hair for project managers if the technical staff can't find a quick remedy. And sometimes there just is is no quick remedy.
    As you say, you have to have the relevant test conditions, configurations and a data other wise you will receive surprises.

    [quote="SFA" wrote:
    I wonder if these fairly large catalogs are mainly imported files? This perhaps more likely to have history of IPTC data that varies for most records.

    Not only that, but large catalogs represent many years of work for a person. For example my catalog not only contains my own images, but also from other contributors - we all went on the same wilderness trips, and we all contributed to a common collection of images.

    [quote="SFA" wrote:
    In theory that should not happen as far as I understand if soince the IPTC codes are intended to be controlled content and table driven. In practice ... who knows what might be in there?
    Grant

    A few of the the IPTC fields have a strictly enforced format but many are free form, particularly the "Title" field which can have up to 64 characters - and there are of course no rules for a Title - and it is likely to be different for almost every image.

    [quote="SFA" wrote:
    However I am using Windows and Sessions and not experiencing the dramas reported for Catalogs and, it seems, mainly Macs.

    Altough the Phase One likely try very hard to keep the higher level code the same, the lower level code, drivers and services, will differ. And as a consequence some interfaces will differ.

    Thanks for the comments
    0
  • Frank Lison
    Thanks Eric, very interesting observation. I also see this effect on my system. Unfortunately it comes on top of my other issues with big catalogs, i.e. I can make the system even slower (completely unresponsive) when using filters extensively. So there seems to be several issues: filters and smart albums, all images folder in general, RAM not released when closing catalogs, metadata handling,..

    Best
    Frank
    0
  • Eric Nepean
    [quote="FL" wrote:
    Thanks Eric, very interesting observation. I also see this effect on my system. Unfortunately it comes on top of my other issues with big catalogs, i.e. I can make the system even slower (completely unresponsive) when using filters extensively. So there seems to be several issues: filters and smart albums, all images folder in general, RAM not released when closing catalogs, metadata handling,..

    Best
    Frank

    Hi Frank
    How many images do you have in your big catalog?
    0
  • Frank Lison
    My original catalog (Aperture) has 250.000+ images and I have split it into several catalogs with about 25.000 each as V8 was not able to handle this. These worked fine in V8 and V9 and show issues with V10. I also see similar issues but not as slow with smaller catalogs (4000 images each). While the bigger ones are referenced with the images being on a NAS I have generated two copies from one of the smaller ones, a referenced and a managed one, so that I can move them around and test them on different disk and computers. The behaviour is always the same and disk speed does not affect the performance fundamentally.

    Best
    Frank
    0
  • Eric Nepean
    Hi Frank

    As they say in my team, when you are troubleshooting "you have to peel the onion one layer at time - What you find in the next layer can be a surprise."

    Your experience with Drive speed matches mine. There's very little disk i/O, the process is not I/O bound, it's CPU bound. Excessive resource usage by the filter could be responsible for all the symptoms you mention.

    I think there is no benefit to using a filter to configure a smart folder, you can achieve the same image selection using additional search terms. However, a filter carries a lot of computational overhead, especially with current implementation.

    In your shoes, I would reconfigure smart folders and presets to not use the Filter, and remove the Filter tool from the Library tab.
    0
  • Frank Lison
    Eric,

    thanks for the hint. I have removed all smart folders and the filter tool but the difference in performance is marginal, i.e. I still spend a lot of time looking at the nice spinning wheel whereas V9 is just behaving fine in the same constellation. Let's hope that the people at PhaseOne do their homework and will soon release a solution.

    Frank
    0
  • Eric Nepean
    This may be a good time to open a new support case .

    I am working with a support guy now who is head and shoulders above the ones previously helping me.

    The following information I sent is much more extensive and will be much more useful to the investigators than what I have sent before, I recomend the same:
    - The file from from COP Filepackager
    - Screen shots of various Tooltabs of the inspector. e.g. Library, Metadata
    - Quicktime movie of the screen showing the problem, with:
    --- Activity window showing
    --- Installed ClockMe to show an on screen HUD with a clock, and also show user key strokes
    --- Installed Lagente PinPoint configured to clearly identify mouse position (cross hairs) with click indicator on.
    - Zipped coctalog.dB file (not the entire catalog). (I got it from the backup folder, zipped it is quite small)
    - Sample of Capture One process while it is non-responsive (zipping may help here too)
    - Screen shot of the activity monitor while COP is non-responsive, show memory usage and CPU usage
    - Screen shot of the Filter configured to show each of the Metadata fields with data (one by one for the big ones) Then Filter tool shows the the list of all Metadata values for that field. Sent the start and the end for a long list
    --- I removed the filter tool from the Library tab, and put it on a custom tab. More room for each tool, more control of which one is active.
    0
  • Eric Nepean
    I made a little Applescript to monitor the status of the the Capture One 10 process, and display on screen when it is logged and how long for. The end of the hang and the duration are also logged in the Applescript log.

    Since this utility checks the process status as indicated in the OS with a Unix command, it should not interfere with COP performance - I have checked that and it seems OK.

    Use the Script Editor to open a new empty Applescript Document, and copy and paste all of this code into the window. Push compile. It should compile without errors.

    Then start Capture One Pro 10, then start the script. It will run continuously, reporting and measuring hangs. If Capture One 10 is not running, the script will crash. I could fix that, but I must go to bed.
    It can be easily adapted to other versions of COP.


    -- AppleScript to detect and time Capture One Pro 10 hangs for the purpose of screen capture and feedback to Phase One.
    -- Eric Valk January 2017 Version 0.2
    -- Start the script running after starting Capture One Pro 10.
    -- Hangs will be displayed as notifications (make sure you have notifications enabled) and the total duration is displayed as an Alert, and also logged in the Applescript log.
    -- Times are only roughly accurate, as the Applescript delay timer is accurate to about 1/60 of a second

    set FinalAlertTime to 2.5 -- time before the alert times out and disappears
    set theApp to "Capture One 10"
    set wait_time to 1.0 / 2.0 -- the time in seconds between checks
    set wait_div to 1 / wait_time -- the divisor needed to scale the output
    set dispTime to 2 as real -- the time between notifications
    set dispCount to (dispTime / wait_time) as integer -- the interval count between notifications
    if false then -- set to true to get some deugging info
    log wait_time
    log wait_div
    log dispTime
    log (dispTime / wait_time)
    log dispCount
    end if

    set nextDispCtr to 0 --The count for the next display notification
    set hang_Ctr to 0 as real -- counts the number of intervals COP is hung
    set isHung to false -- state this time
    set wasHung to false -- state last time

    if true then -- set to false to skip the actual measurements
    tell application "System Events"
    set PID to unix id of process theApp as Unicode text

    repeat while true

    set state to word 7 of paragraph 2 of (do shell script "ps -j " & quoted form of PID & " ")

    if state = "R" then
    set isHung to true
    set hang_Ctr to hang_Ctr + 1

    if hang_Ctr ≥ nextDispCtr then
    set nextDispCtr to hang_Ctr + dispCount
    set disphungtime to ((((100 * (hang_Ctr / wait_div)) as integer) as real) / 100.0) as text -- truncate to two digits
    display notification "COP is hung for " & disphungtime & " seconds"
    end if

    if not wasHung then set nextDispCtr to nextDispCtr - 1 -- try to align display on integer seconds
    else
    if wasHung then
    set disphungtime to ((((100 * (hang_Ctr / wait_div)) as integer) as real) / 100.0) as text -- truncate to two digits
    display alert "COP was hung for " & disphungtime & " seconds" giving up after FinalAlertTime
    log "COP was hung for " & (hang_Ctr / wait_div) & " seconds" & (get current date)
    end if
    set isHung to false
    set hang_Ctr to 0
    set nextDispCtr to 0 -- so that the first display occurs as soon as the hang is detected
    end if
    delay wait_time
    set wasHung to isHung
    end repeat
    end tell
    end if
    0
  • paintbox
    Eric,
    That workaround helped out quite a bit.
    I don't use filters a lot either, mostly because it sucks the life out of C1. So I've learned to use other methods.
    One thing I do use filters for is trying new techniques on older files. If I wanna try a newer noise reduction system, I can quickly find the higher ISO files and get to work. Same goes with C1's new Diffraction Correction. I can pick out the narrower f-stops to see how the files are affected.
    I really hope they get this issue resolved and I appreciate you taking the initiative on getting it resolved.
    Thanks, Jimmy
    0
  • Frank Lison
    Thanks Eric. Will test this tomorrow on a freshly converted catalog as catalogs seem to get slower and slower the longer I use them.

    Best
    Frank
    0
  • Frank Lison
    Eric,
    this is what I get when opening one of my normal (25.000 images) catalog:
    *COP was hung for 12,5 secondsThursday, 12 January 2017 at 22:02:48*)
    (*COP was hung for 2,5 secondsThursday, 12 January 2017 at 22:02:53*)
    (*COP was hung for 3,5 secondsThursday, 12 January 2017 at 22:02:59*)
    (*COP was hung for 13,5 secondsThursday, 12 January 2017 at 22:03:16*)
    (*COP was hung for 1,5 secondsThursday, 12 January 2017 at 22:03:20*)
    (*COP was hung for 5,5 secondsThursday, 12 January 2017 at 22:03:29*)
    (*COP was hung for 84,0 secondsThursday, 12 January 2017 at 22:05:01*)
    (*COP was hung for 5,0 secondsThursday, 12 January 2017 at 22:05:08*)
    Now I can start browsing through the catalog and get these times (not editing, just browsing).
    (*COP was hung for 0,5 secondsThursday, 12 January 2017 at 22:05:12*)
    (*COP was hung for 0,5 secondsThursday, 12 January 2017 at 22:05:17*)
    (*COP was hung for 0,5 secondsThursday, 12 January 2017 at 22:05:22*)
    (*COP was hung for 0,5 secondsThursday, 12 January 2017 at 22:05:25*)
    (*COP was hung for 0,5 secondsThursday, 12 January 2017 at 22:05:28*)
    (*COP was hung for 1,5 secondsThursday, 12 January 2017 at 22:05:32*)
    (*COP was hung for 3,0 secondsThursday, 12 January 2017 at 22:05:40*)

    Not sure how to interpret these values / how they compare to yours.
    Frank
    0
  • Eric Nepean
    Hi Frank

    How to interpret..

    The log entries which are like this: (*COP was hung for 0,5 seconds Thursday, 12 January 2017 at 22:05:28*)
    means there was a very short hang you might not even notice. Half a second is the shortest period the script can determine, so it might have even been shorter than that. There are probably other short hangs the script missed. But identifying short hangs is not the purpose of the tool.

    I see that you had one really long hang of almost a minute and a half, a couple in the 10-15 second range, and the rest are under 10 seconds. Would you say that matches your user experience?

    If I have the Metadata I started with, and if filters tool is open, I get many many hangs in the 20-30 second range. Literally every time I mouse click or click a key on the filters tool, or on the search window, I get a 25 second hang. Typing a 10 letter search term in the search window takes 5 minutes. I have one version of a catalog where it's even worse (I did something with the Metadata that COP really doesn't like), and each mouse click or key click carries an 84 second penalty.
    0
  • SFA
    Eric,

    I find it interesting that 84 seconds is a regular hangup number for you and also appears in Frank's output.

    If the cause was variable one might expect a more random distribution of numbers.

    Is this the sort of process that, once something has kicked it off, will stall things for a fixed period of time whether it needs to or not?

    Grant
    0
  • Frank Lison
    The long hang up is happening every time I start this catalog. It is not always this particular number (this might be due to the tool as I then see several larger hangs up that would add to something in the range of 80 to 90s) but always happening in the moment when images appear in the browser tap. What is interesting is that
    - CPU load is <15% for the complete time
    - Memory usage is reasonable and below 25% of the available
    - GPU usage is basically none
    - GPU memory is high (1.75 out of 2) but this is usual for C1P
    - Disk and Network usage is basically none

    Though C1P completely stalls for about 1 1/2 minutes the computer is completely fine to use for everything else.

    Note: there was a severe bug in V9-V9.1 with GPU memory not being released properly on NVIDIA cards and the system consistently crashing at some point in time. To me it looks like there is some "limit" in GPU memory usage "implemented" that prevents these crashes. However as GPU memory is not released it may well be that everything gets incredible slow and this is limited all processes. But I can't get any better performance when disabling GPU which was possible with the old bug.

    Best
    Frank
    0
  • Eric Nepean
    Hi Frank

    Here is my hang data . I have updated the script a little to discard hangs under 2 seconds, and to improve the accuracy, the original was reading about 10% low.

    (*COP was hung for 4 seconds Friday, January 13, 2017 at 1:32:10 AM*)
    (*COP was hung for 54 seconds Friday, January 13, 2017 at 1:33:11 AM*) - selected a filter
    (*COP was hung for 106 seconds Friday, January 13, 2017 at 1:35:06 AM*) - selected another filte
    (*COP was hung for 27 seconds Friday, January 13, 2017 at 1:35:50 AM*) - opened the advanced search window
    (*COP was hung for 29 seconds Friday, January 13, 2017 at 1:36:49 AM*) - setting up a search
    (*COP was hung for 29 seconds Friday, January 13, 2017 at 1:37:54 AM*) - setting up a search
    (*COP was hung for 28 seconds Friday, January 13, 2017 at 1:38:39 AM*) - setting up a search
    (*COP was hung for 27 seconds Friday, January 13, 2017 at 1:41:01 AM*) - setting up a search
    (*COP was hung for 28 seconds Friday, January 13, 2017 at 1:42:54 AM*) - setting up a search

    Here is catalog opening. About 60 seconds altogether
    *COP was hung for 27 seconds Friday, January 13, 2017 at 2:08:20 AM*)
    (*COP was hung for 28 seconds Friday, January 13, 2017 at 2:08:51 AM*)

    If I get rid of that one metadata field which is the original file name, then the delays associated with the filter tool go away.

    Most of the activity indications I get are the same as yours, except that the hangs associated with the filter tool uses up about 1 core (out of 😎 -100%

    Loading the catalog uses 2 cores, or 200%. But I'm less worried about catalog loading than the key click delays while using Capture One.
    0
  • Frank Lison
    Eric,

    thanks, very consistent with my observation. I completely removed filter, smart albums,... some time ago as V8 and V9 had big bugs when using Smart Albums. So I do not see your delays but it is still slower than V9 when browsing and editing. Instead of filters and albums I have split into catalogs accordingly and therefore need to be able to switch fast in between these, i.e. loading and startup time and memory releae is the bigger issue for me.

    What bothers me most is that all has been documented and communicated to C1P. They have investigated my full log files and acknowledged the issue, proposed several changes that did not really help and then become quiet. Having bugs in a new release is one thing, not communicating properly makes it much worse.

    Frank
    0
  • Robin Procter
    So, if I understand the advice offered so far (ignoring the highly technical gobblygook which goes above my head but is meaningful to others who are more expert than myself in this respect), the following setting should reduce the browser speed problem....

    Top Menu > View > Remove Tool From Library Tab... > Filters

    Would a CO Quit and then re-Launch be necessary?

    I shall try it out shortly but trust that this action will not remove any previously set Star Ratings or Color Tabs from any image files. Such cleansing would be absolutely disastrous!
    0
  • SFA
    Robin,

    I think, from what I have read, this problem only appears in catalogues and then is mainly notable when quite a lot of files are catalogued in that one catalogue.

    If you are using sessions I doubt that you would have a problem.


    Grant
    0
  • Eric Nepean
    [quote="RedRobin" wrote:
    So, if I understand the advice offered so far (ignoring the highly technical gobblygook which goes above my head but is meaningful to others who are more expert than myself in this respect), the following setting should reduce the browser speed problem....

    Top Menu > View > Remove Tool From Library Tab... > Filters

    Would a CO Quit and then re-Launch be necessary?

    I shall try it out shortly but trust that this action will not remove any previously set Star Ratings or Color Tabs from any image files. Such cleansing would be absolutely disastrous!

    As SFA says, this is a strong affect on catalogs with more than about 8000 images. I think there is no effect on sessions or catalogs with less than 2000 images.

    After removing Filters my experience is that the effect is immediate, no reboot required.
    0
  • Robin Procter
    [quote="Eric Nepean" wrote:

    As SFA says, this is a strong affect on catalogs with more than about 8000 images. I think there is no effect on sessions or catalogs with less than 2000 images.

    After removing Filters my experience is that the effect is immediate, no reboot required.


    ....Also now my experience since removing 'Filters' from view in the Browser. Thanks 😊

    I only have Sessions, no Catalogs, and some of those Session folders have far more than 2,000 images as I usually dedicate each Session folder to a subject (Birds for example) for the year.
    0
  • Jerry C
    I removed the filter tool, and Capture One version 10.0.1 became unresponsive for 30 min. Activity Monitor says it was reading and writing data. I have learned to go do something else when this happens since C1 appears to be doing something and force quitting is sometimes corrupts the catalog.
    0
  • Jerry C
    I removed the filter tool, and Capture One version 10.0.1 became unresponsive for 30 min. Activity Monitor says it was reading and writing data. I have learned to go do something else when this happens since C1 appears to be doing something and force quitting is sometimes corrupts the catalog.
    0
  • Jerry C
    Excellent suggestions.

    COP version 10.0.1 behaves quite fast with my 52000 image catalog after moving the Filter tool from the Library tab to a custom tab. Unfortunately, removing the Filter tool from the Library tab took 30 minutes of watching the spinning beach ball, so you might need patience and a look at the Activity Monitor to see that activity is actually happening. Thereafter, advanced searches were near instantaneous as each search criterion was entered. I don't know why CO has to scan the catalog with each little addition to the search criteria (unlike with a Smart Album), but since it was instantaneous, why not.

    Also, before doing this, I could not edit the key word library without hanging CO infinitely. Now it works normally.

    Jerry C
    0
  • Jerry C
    I spoke to soon.

    Although deleting the filter from the Library tab does as I described, when I reopened my Catalog, today, I had a huge problem. Every time I tried to click on an image in the All Images Collection, Capture One became unresponsive. I actually waited 2 hrs while the OS10 Activity Monitor showed Capture One doing a lot of reading and writing (GB worth). Force quitting erupts the database beyond repair. I have lots of backups , so it is not catastrophic.

    So, if your version 10 is not crashing, you should see the speedups I described.

    I am back to version 9.3, which is very stable. I did find that moving the filters tool out of the Library tab does speed up the search and the data entry in version 9.3.
    0
  • Eric Nepean
    If you're engaged in removing or moving the filters tool because it may be slowing down filter/browser performance in a larger catalog, it's faster to do it if you first select a small user collection rather than the "All Images" collection.

    The slow down due to the filters tool occurs when it is working on a large number, e.g. >>5000, images. But the Filters tool works on the all images in the selected collection(group, album, project). So selecting a group showing no images, or a project or album with a few hundred images temporarily removes the delay.
    0
  • Jerry C
    Neil is correct, which is how I did it. My problem with Version 10 relates to slow performance navigating to All Images after relocating the Filter tool. Selecting an image in All Images results in waiting 15 min on my Mac Pro for C1 to respond. This time gets shorted as I navigate back and forth to All images a couple of time, but this is more annoying than starting a Model T on cold day. On my Macbook Pro, I gave up and force quit after 3 hrs. and was rewarded by an unfixable corrupted database.

    Jerry C
    0

Please sign in to leave a comment.