A multikey index will use the time-series buckets non-positional min and max data to build the index. The same indexes will be created as described in the ascending and descending indexes ticket. This works well for non-positional queries such as db.multi.find({data: 5}) or db.multi.find({data: {$gt: 9}}). For positional queries such as db.multi.find({data: [5]}) or db.multi.find({data: [5, 9]}), the non-positional min and max data can be computed from the query. The indexes can then be used to find the time-series buckets that satisfy the min and max range of the query. Any time-series buckets that match this criterion will need to have their measurements looked over for matches.