Has there been any thought to having multiple hosts under 1 account pool work from multiple projects over a LAN?

Would there be any efficiencies gained by having multiple hosts work from the same pool of work?

You can't even have multiple cores running the same workunit, so you definitely can't do it with multiple separated computers. If you mean a single work pool but each host computing different workunits (ie. have one host fetch work and give it to the rest), then it's VERY hard to do AND you won't gain anything.

You should just attach each computer to the projects (on same account) and let them each have its own work queue.

No, meant a single pool... not working the same work unit...

That won't give you any benefit. Why do you want all to use a single pool? You should just let each computer keep its own queue.

Also, not even BOINC developers know how to get such a thing working, see Superhost Design and Topology (devs have said it won't be implemented anytime soon; that page is just an idea).

The idea bsed on the premise that the server sends work based on the user account.

User account is defined by the hosts attached. (e.g. multiple platforms / processor capabilities) etc.

Single program manager controls the multiple hosts.

Work assigned by efficiency when possible but ultimately by deadline.

Slower machines chew on the work thats due at a later date
Faster machines crunch any deadline misses

Linux machines keep the linux apps clear, etc

Work is downloaded based on the efficiencies of the 'MiniGrid'

If a machine is removed from the 'MiniGrid' work still get's done...

I think that it would be best used say in an environment where people work on the pc computers for say an 8 hour shift... thus taking it out of the pool.

In the current environment each computer has its own silo of work. Hypothetically speaking, my computer pool has ~100 pcs and some machines will always be on hold for 'BOINC' crunching, based on preferences, when an employee is working approx ~8 hrs. Chances are high that WU's on multiple individual work queues will expire when other pc's are crunching away on non-deadline intensive work.

Not necessarily saying that it would change the face of BOINC as we know it, but perhaps some efficiencies could be gained.

I've seen a thread or two that stated that certain apps from various projects have more success on one platform vs another... Apps could cary an efficiency rating within the minigrid, based on performance with different machine types.

Any computers that have higher than average pc user time could be fed wu's with shorter processing time.

This was pointed out to me on another message board if anyone is interested:

http://boinc.berkeley.edu/trac/wiki/SuperHost

I think your idea could be the same as Find-A-Drug used to be.

It was downloaded onto one computer and if I remember right, you copied a shortcut to each of your other computers and clicked the shortcut on each computer. the work was downloaded to the first computer and then individual work units were assigned to all the computers on the network.