Parting data to improve system performance

A non-transitory computer-readable storage medium having computer-readable instructions stored thereon that are executable by a computing device to perform operations comprising:

• parting a low update frequency data unit from a plurality of data units included in a data item based on a historical access pattern of the plurality of data units;
• determining a regional group for the low update frequency data unit based on the historical access pattern of the plurality of data units, wherein the regional group includes a subset of hosts, wherein each host is coupled to one another by a computer network and within a particular distance of a geographical location, and wherein the regional group is determined for the low update frequency data unit based on routine user access and/or particular user interest associated with the geographical location;
• replicating the low update frequency data unit;
• and
• pushing the replicated low update frequency data unit to the subset of hosts.

A system, comprising:

• a plurality of hardware processor-based hosts located at a plurality of locations, wherein each hardware processor-based host is coupled to one another by a computer network;
• and
• a master device comprising a hardware processor and configured to communicate with the plurality of hardware processor-based hosts, wherein the master device includes a computing device that comprises the hardware processor and a computer-readable storage medium having stored thereon computer-readable instructions that are executable by the computing device to perform operations to:
• calculate update frequencies of data units included in a data item, determine regional groups for the data units based on historical access patterns of the data units, wherein each of the regional groups includes a set of the plurality of hardware processor-based hosts, wherein each hardware processor-based host is located within a particular distance of a geographical location, and wherein the regional groups are determined for a low update frequency data unit based on routine user access and/or particular user interest associated with the geographical location,
• part the data units into low update frequency replication units based on the historical access patterns of the data units, wherein each of the low update frequency replication units includes:
• a subset of the data units, each having an update frequency below a threshold, and
• a location list that indicates a subset of the plurality of hardware processor-based hosts common to the regional groups of the subset of the data units,
• replicate the subset of the data units included in the low update frequency replication units, and
• push the replicated subset of the data units to the subset of the plurality of hardware processor-based hosts.