Two questions:

1. Terminology:
a. What's the difference between "Running" and "Running High Priority"?
b. What's the difference between "Waiting to run" and "Ready to start?"

2. I'm running 3 different projects and have BOINC Preferences set to allocate 33% time to each project. I often see situations where a project with a later completion date will be running while a project with an earlier completion date sits idle. Is there a way to force BOINC to process projects in order of their due dates?

Thanks,

I'm not familiar with the intricacies of the BOINC scheduling system, but I can give it a shot:
AFAIK, this happens when a task is in danger of not completing before deadline; the client won't switch applications or request new WUs under this condition.

"Waiting to run" means that the task has accumulated run time, but is being preempted by another. "Ready to start" is what a fresh task is labeled as, before any progress has been done.

The behavior you are experiencing is normal; if three projects have an equal share, the client will switch between them to give them all an equal amount of run time. There has been quite a bit of work put into the scheduling system, and it's not perfect, but if you are concerned about not having work get done on time you should be ok. You might lose a couple when first attaching a host and the client figures out how much work it needs to fetch, but after that it shouldn't be fetching any more work than will complete on time. There can be some problems with those projects that don't have a very reliable "To completion" time.

Alice and Dave

I know nothing about how BOINC decides which units to run, but one thing I have noticed is that the larger the cache of work units I keep, the more likely I am to have WUs that won't complete on time. I keep ten days of work on my machines, and almost every day I have to abort units that are not going to finish in time. However, when I keep only a one day cache of WUs, then it is very rare that they don't complete on time. I use BOINC View 1.5 beta 8 to monitor all my work units on my machines, and through it I can manage work units to abort on all my machines at once if necessary. I hope this info helps.

IMHO, there is no real reason to set your clients to cache much unless you are hording WUs to hit some sort of milestone before there are no longer any available. Otherwise, you're just likely to have some problems with things being submitted after the deadline--I have my hosts set to connect & keep additional work for "0 days" and will only get new work shortly before the running WUs are complete.

The only exception that I can think of, is if yours is a rare host where you must manually connect to the Internet (i.e., dialup) to send & receive work.

noderaser

I like to keep a large cache because I don't like to have my computers sitting idle when one of my projects is down. Usually that is rare with the projects I'm currently crunching, but some projects I've run in the past have had long outages. I have my own little challenges running. I am focusing on getting as many of my projects over one million as possible, and I'm just going down my list. Right now I'm concentrating on Leiden Classical. A few weeks ago they were down for several days over the weekend. Since I'm running that project only on my CPUs, they would have been idle for days without the cache. Yes, I know I could run other projects at the same time, but I'm focusing my CPUs on that one project right now. Besides, I like being hands-on with managing my little farm. But, you're right, for most crunchers with only one or two hosts and a permanent internet connection really no cache at all is necessary.

Ah, well I'm pretty sure the scheduling system is designed to make sure that no WUs go past their deadline; if you're having problems with that, you might try decreasing your cache slightly to account for any variations in processing time.

I think its inconsiderate to hord WUs that you know you will not be able to finish and have to abort.
Projects issue WUs to process the data they need for their research. They expect that work
to be performed in a "timely mannor". Needlessly aborting WUs delays their research. It's
best to set your caches to so that any WUs you download, will be processed in the time given.

After 5 years of running BOINC, I've found it best to let it do its thing. Trying to micro manage it
causes problems. I run 30 projects, each set at 100% resources, with a one day cache. If a
project is down, the workload shifts to the other WUs and they get processed faster. I am NEVER
out of work!

Sorceress

Allegations of "hoarding" WUs are not appreciated. I was asking for some guidance as to what may be wrong. The goal is to find it and fix it. I'll try decreasing the cache to 1 day and see what happens.