Computer data distribution architecture connecting an update propagation graph through multiple remote query processors

A method comprising:

• assigning a first sub-graph of a query graph to a first query processor;
• assigning a second sub-graph of the query graph to a second query processor;
• creating, at the second query processor, an object to represent a replica of a result of the first sub-graph from the first query processor;
• sending a subscription request from the second query processor to the first query processor to receive consistent updates to the result of the first sub-graph;
• and
• responsive to receiving a notification at the second query processor, updating the replica of the result and propagating the changes through the second sub-graph at the second query processor.

A method comprising:

• assigning a first sub-graph of a query graph to a first query processor;
• assigning a second sub-graph of the query graph to a second query processor;
• creating, at the second query processor, an object to represent a replica of a result of the first sub-graph from the first query processor;
• sending a subscription request from the second query processor to the first query processor to receive consistent updates to the result of the first sub-graph;
• receiving a notification at the second query processor corresponding to a change in the result of the first sub-graph from the first query processor;
• responsive to receiving the notification at the second query processor, acquiring an update lock, updating the replica of the result, propagating the changes through the second sub-graph at the second query processor, and releasing the update lock.

A nontransitory computer readable medium having stored thereon software instructions that, when executed by a processor, cause the processor to perform operations comprising:

• obtaining current logical clock time and current logical clock state;
• determining if the current logical clock state is idle;
• when the current logical clock state is idle:
• reading current data as a result snapshot, wherein the current data is data that is current for the current logical clock time;
• when the current logical clock state is not idle:
• reading previous data as the result snapshot, wherein the previous data is data that existed prior to the current logical clock time;
• querying a new current logical clock time;
• and
• determining if the new current logical clock time and the current logical clock time are identical;
• when the new current logical clock time is identical to the current logical clock time:
• sending the result snapshot as the initial snapshot of the result;
• when the new current logical clock time is not identical to the current logical clock time:
• acquiring an update lock;
• reading new current data;
• releasing the update lock;
• and
• sending the new current data as the initial snapshot of the result.