Performance issues

I’ve become really quite keen on DT, to my surprise - it’s the first tool of its kind with which I think I could have a long-term relationship!

I’m plagued, however, by the spinning beachball, and by performance in general. I have a fast Powerbook with 1G RAM, it’s not heavily loaded with other things, I optimize my database regularly, but I still get frustrated by the slowness of DT. It seems to go away regularly for extended periods of what i assume are garbage-collection.

I had it indexing about 5000 documents on my hard disk, which was really useful but made it crawl, so I’ve deleted those and have accepted that it will probably be more of a notebook for my clippings than a replacement for Spotlight. It does still have a couple of thousand past emails in it, but I’m hoping it will be able to cope with those.

I was wondering whether this issue was likely to be much improved in DT Pro?

With a big database (I’ve certainly got one :slight_smile: ) DEVONthink is making heavy use of RAM and Virtual Memory. There’s a lot of processing going on when you do some types of searches, and certainly when you use the Classify and See Also features. VM scratch files build up, and data gets paged back and forth to RAM.

Your practice of using the Backup & Optimise tool regularly is a good one, and helps keep the database running well – as well as assuring a recent backup of your work.

Another trick that I use, when DT gets sluggish, is to close the database, then relaunch it.

Another trick – an important one – is to keep System and disk operations clean and healthy. Disk Utility can be used to repair Permissions, and this should be done before and after major software installations such as OS updates and security updates. There’s a ‘donation ware’ utility called AppleJack that allows disk repair, cache cleaning and so forth to be done from Safe mode – much faster and more convenient than starting up from a CD. There are several utilities, some freeware, that allow cron scripts to be run without going to the Terminal. DiskWarrior does a great job of catching and repairing disk errors and optimizing the directory. ‘Housekeeping’ helps ensure that your computer is running at its maximum performance, and protects your data.

I’m using a beta of DT Pro. Although DT PE and DT Pro were using the same basic ‘engines’ when DT PE 1.9.2 was released, the developers have continued to speed up DT Pro operations such as phrase searches. Phrase searches were recently tweaked to run an order of magnitude faster. Improvements like that will find their way into future versions of DT PE, as well. For example, you may be noticing a lag in response the first few times you do a search, or use the Classify button. DT Pro beta is now dramatically faster.

My own computer is a more than 4-year old TiBook 500 MHz, with 1 GB RAM. I continue to be awed by how fast DEVONthink performs on my database, which is much larger than the one you had tried.

Bottom line: DT (including PE) can run fast, even with databases much larger than mine. A bit of routine housekeeping can ensure continued quick operations.

Future news: DEVONtech will introduce a revamped database structure in upcoming versions. Transition to the new structure will be smooth. Memory usage will be reduced. Many DT operations will receive a 5 to 10-fold speed boost (on top of the boosts I noted above in DT Pro beta). Spotlight will be able to see your files under OS X 4. But DT will do so much more than Spotlight that you will want to put more and more things into DEVONthink.

Bill, do you think it might be possible for you to write an article at the DevonWiki about DT-Database housekeeping and other hints to improve performance?

Thanks. I’ll put that on my ‘priority’ list.

Another one planned for ASAP is a section on Tables & Records in DEVONthink Pro.

Any other suggestions for “soon” sections of the Wiki?

How about collecting together your forum posts on PDF to text options?

Another would be the collected wisdom of printing rich text format documents, with its margin and pagination issues, with workarounds.

Still another would be use of Wikilinks, both automatic and manual

Obviously, not all requests directed at you specifically (although the first request has usually been answered by you.) We could just create the sections, and hope that someone came along and populated them.