In Scrum, work is done in fixed-length time-boxes that begin with Sprint Planning and end with Sprint Review and Sprint Retrospective meetings and is expected to produce working, demonstrable pieces of deliverable product. While Sprints can be any length, they are usually one, two, three, or four weeks in length (up to a maximum of a calendar month). During the Sprint, the ScrumMaster is responsible for ensuring that the goals set during Sprint Planning are not changed during the course of the Sprint (should changes be necessary, a Sprint Termination is performed).
The most effective power behind the concept of the Sprint is created by the shortened timebox. In the typical project, where the timeframes are measured in months and, in some extreme cases, years, a significant degree of variability can creep in. Goals change, priorities change, customer requirements change. And while change is to be expected, trying to maintain a detailed plan, despite the changes, over a lengthy period of time can be extremely difficult. By shortening the timeframes to the length of a Sprint and, essentially, running your project as a series of "tiny projects," Scrum teams are able to "lock-in" priorities and plans and work to complete a small piece of the overall project before substantial change can be realized. By their reduced length, Sprints actually reduce project risk.
Sprints are no more than one calendar month in length and may be any fixed length less than one month. Generally Sprints occur in periods of one to four weeks. Initially, teams favored four-week or one-month Sprints. However, more recent implementations have favored two or three-week Sprints. My personal preference is for two-week Sprints, which provide more opportunities for team learning and improvement. Whatever the length of your Sprints, you should plan to keep them the same length. Scrum teams determine how much they can do during a Sprint is by comparison to previous sprints. When a team successfully completes 15 story points in one sprint and then proceeds to complete 17, 15, 14, and 16 story points during the next 4 sprints (which gives them a velocity of 15 story points), they are fairly likely to complete something close to 15 story points in the next sprint as well (yes, something could happen that causes the team to only get 11 story points done instead, but the odds are likely that the team will be able to identify the causes before a commitment has to be made; e.g., a holiday or scheduled training during the Sprint). This consistency in the amount of work that a team completes during a sprint is, to a large degree, a factor of the sprint length being consistent. As you might guess, a team that can get 15 story points of work done in one 2 week Sprint, will get more done in a 3 week Sprint, and even more done in a 4 week Sprint. So, if we keep using the same Sprint length, it is reasonably likely that the team will complete a similar amount of work each Sprint.
For More Information
For more information about Sprints, see