Escrow transactions  permit non-blocking concurrency for the most common high volume transactions, those that perform updates of hotspot data only by incrementing and decrementing aggregate data items. An example of such a transaction is one that increments dollar sale total of a customer for each purchase that decrements the quantity on hand of a product. A hotspot of a database under a high volume transactional workload consists of the set of data items, each one of which frequently needs to be updated by multiple concurrent transactions in order to maintain needed throughput. While this is impossible using read-write locking, since write operations do not commute, high-volume OLTP applications of this kind can be performed without blocking by a transactional system that performs increments and decrements with the appropriate protocols. Increment-decrement updates have been used for years in the IMS Fast Path product [1, 2], supporting...