How MSR Maps Works

Database Partitioning

The Microsoft Research Maps data is a set of three servers known as a Bunch. Each server is referred to as a "Storage Brick" or "Brick". The Microsoft Research Maps site has two bunches containing identical data. The web applications can retrieve imagery from bricks in either bunch.

Imagery is distributed between bricks in a bunch based on the location of the image. From experience we know that users have a tendency to view images where they live. During peak times, MSR Maps is busiest in the mornings in the time zones where our users live and MSR Maps has imagery. Thus, we attempt to evenly distribute data for a single time zone across all three bricks in a bunch.

The diagram below shows how the data themes currently available on Microsoft Research Maps are distributed between each brick. Both bunches are identical, i.e. Brick 1 in Bunch 1 is identical to Brick 1 in Bunch 2.

All Microsoft Research Maps themes are in the UTM map projection. UTM divides the globe into sixty, six-degree wide zones. A time zone is roughly 15 degrees wide. Therefore about 2 UTM zones cover each time zone.

The above diagram shows the current database sizes for each theme and zone.