| CPC G06F 9/5005 (2013.01) [G06F 9/544 (2013.01); G06N 3/08 (2013.01)] | 20 Claims |
|
1. A system including:
serverless sizing circuitry configured to execute a serverless sizing stack, the serverless sizing circuitry configured to:
at a data-staging layer of the serverless sizing stack, obtain historical data including execution duration data and compute usage data for a serverless function;
at an eligibility layer of the serverless sizing stack:
access the historical data via data-storage operations provided by the data-staging layer;
based on the historical data, determine data completeness for a serverless function represented within the historical data; and
when data completeness for the serverless function exceeds a pre-determined threshold, mark the serverless function as sizing-eligible;
at a metric selection layer of the serverless sizing stack, select an execution duration metric and a compute sizing metric for iterative comparison;
at a compute allocation layer of the serverless sizing stack, determine, via a compute deep-learning analysis of the compute usage data, a base compute size and compute buffer range;
at a duration mapping layer of the serverless sizing stack, conduct an iterative traversal of the compute buffer range by:
for each iteration:
selecting a current compute buffer size for the iteration, the current compute buffer size being within the compute buffer range;
based on a combination of the base compute size and the current compute buffer size and using a duration deep-learning analysis of the execution duration data, determining a current compute duration for the serverless function;
based on the current compute duration, the base compute size, and the current compute buffer size, determining a current computational-performance score for the current compute buffer size;
determining whether the current computational-performance score meets an iteration stop criterion; and
when the current computational-performance score meets the iteration stop criterion:
ending the iterative traversal; and
designating the current compute buffer size as a prescriptive compute buffer size; and
at a requisition layer of the serverless sizing stack, generate a serverless requisition token including:
a request for the serverless function with a requested size corresponding to the prescriptive compute buffer size and base compute buffer size; and
network interface circuitry configured to send the serverless requisition token to a host interface for control of serverless compute resources.
|