IMAP: large number of folders not working

[s_turner]

Hi,

I'm trying to get the IMAP service working for a Share site with roughly 160,000 documents in 250,000 folders without much success. My folder structure under Document Library has 7 levels of folders starting with 1 folder at the top level, 10 below that, 36 below below each of those with many under those. Documents are stored at the bottom level.

When I mounted the top level folder the IMAP service started but as soon as I tried to access the folders via Outlook the Alfresco server hit 100% CPU and stayed like that. It eventually had to be rebooted after an hour.

I then tried mounting a subfolder containing 10,000 documents in 25,000 folders. Alfresco took 40+ minutes to start, not sure what was happening but the IMAP service seemed to be taking a long time to start. Again, as soon as I tried to access the folders via Outlook the Alfresco server hit 100% CPU and didn't resume normal service until rebooted.

I've tried the IMAP service with a small number of folders and documents, 10 to 100, and it works fine. I see from other forum posts that there has been problems with IMAP performance in the past but that CE 4.0 includes "IMAP enhancements for performance and reliability". Is it still the case that the IMAP service cannot handle this number of folders? Perhaps these changes are not to do with handling large numbers of folders and documents?

[I'm wondering if the problem is with the IMAP protocol or Outlook in that they require the full list of folders/documents up front rather than requesting information from Alfresco as a user navigates the folder hierarchy in Outlook? And that returning details of 25,000+ folders is never going to be viable?]

Oh and this is using Alfresco CE 4.0d on a Windows Server 2008 with 4GB RAM, XEON CPU@2.13GHZ, Java memory -Xms128M -Xmx1152M.

Any comments or help would be appreciated.

Thanks,
Stuart

[mrogers]

Yes the problem is the IMAP protocol itself which is not designed for huge numbers of mailboxes.

To be fair its an email protocol so its not really designed for a content repository.

Are you mounting your entire repository over IMAP?   It will probably be better to mount just parts of your repo.

[s_turner]

Thanks for your quick reply. I suspected that might be the case.

Unfortunately the Outlook integration is needed by back office staff who require access to the full set of "active" folders, (the subfolder with 10,000 documents in 25,000 folders), so mounting smaller subsets of this I don't think is workable.

Thinking about it, since most of what they need to do is upload documents to Alfresco I'm wondering if we could make use of the inbound email server. I assume I could add the Email Alias aspect to a folder when I create it via web services and set the alias property to something unique?

Users could then just email their documents into the repository. Do you know if there would be a performance issue with having so many email aliases defined in the repository, i.e. depending on what level subfolder we add the aspect could result in several thousand aliases? I'm thinking of when an incoming email is processed. Obviously Alfresco needs to find the folder to add it to via these aliases - hopefully it's fairly quick query on the DB to find out.

[mrogers]

Yes you could use email.   Or for back office, how about ftp?

[s_turner]

I'm not sure they're up to using FTP software, I think they're more used to shuffling around pieces of paper.

I guess I should also investigate CIFS as this would provide drag and drop from Windows Explorer.

Thanks for your help.