Optimizing scheduling of developers: "who works on what, and when"?

Question

I'm trying to plan the next 9-12 months of development using a Gantt chart. I've input all the tasks, their durations, and their dependencies, and all is well. However, as I assign developers to tasks, I've noticed that some developers have big gaps where they aren't assigned to any tasks, while other developers are overloaded with too many tasks at times. Since each task has a varying start date and duration, it is not easy to come up with the optimal scheduling of developers.

In my example, I have:

• 10 developers (let's assume equal skill sets)
• ~40 well-defined tasks, each with a different start date and duration. Some tasks must wait for other tasks to be completed before they can begin. (These are not simple tasks that last only a few hours; but more long-term tasks that last anywhere from a few weeks to a few months.)

Is there a known algorithm/solution/tool for coming up with an assignment of developers to tasks, so as to simultaneously:

• Minimize the amount of time each developer is not assigned any tasks
• Create the shortest possible scheduling, so all the tasks are done as fast as possible?

Note that the algorithm/solution/tool is allowed to assign multiple developers to a single task (so that that task would take half as long) and it may also have some overlap of tasks, i.e., a developer may be assigned to more than one task at a time, as long as his total effort sums to 100% at any given time.

Answer 1

When people sit down to schedule, they schedule tasks. In fact, you are scheduling resources-human and otherwise. So if you ended up with a severely un-leveled schedule, do it again with the principle of scheduling resources.

Some tools level for you, but the results are mostly unusable.

Another piece of advice: don't look to level with perfect results. Getting close is good enough since the second you load actual, your leveling goes out the window.

Answer 2

As others have pointed out it sounds like you are looking for information on resource levelling. If you did want to look into using a formula/algorithm for calculating optimum resource allocation and delivery time then you might want to start with this brief article on available algorithms for the task. Though in my opinion the cost of developing a system of your own would be far greater than the cost of simply purchasing software that does this job for you.

What's really important to remember though (and I think is one of the main reasons auto levelling doesn't work) is that you need to take into account all of the interdependencies between tasks way ahead of their actual delivery date. The risk is that your early projections will give an overly-optimistic view of what's possible over a 9-12 month period. As your 'cone of uncertainty' reduces over the lifetime of the project you may realise that the interdependencies are more complex than originally anticipated or that (surprise, surprise) the scope has changed to the point where your interdependencies are no longer relevant.

Answer 3

Check it out MS project has auto levelling function that will help you to achieve your objective more or less.

Answer 4

Don't even try to level or balance resources out so far in advance. There is too much uncertainty that could totally change the picture.

People get ill, resign and working situations change and business constraints impact. Only look at assignments in the very short term and the further you go out the more rough the allocation is. i.e. Work for the next day can be allocated but work for 6 months out should be in the form of "We need 10 developers for another 6 months".

Trying to plan so far in advance with so much uncertainty, will just mean you are going to have to do a lot of constant rework to keep the plan up to date. I am sure you have better things to do that constantly rework the plan.