Back to results
US 11,381,483 B2Granted
Maintenance recommendation for containerized services
AssigneeVMware, Inc.
InventorsYash Bhatnagar, Naina Verma, Keerthanaa K, Sivaraj M
Life of this patent
Estimating the term…
Abstract
A maintenance recommendation for containerized services can find a time to perform maintenance on a particular service based on resource usage patterns such that the maintenance will have a reduced impact on dependent services. The dependent services can be determined for the particular service based on network interactions between the services.
Claims20
A system, comprising:
- processing resources;
- memory resources;
- and
- a controller coupled to the processing resources and the memory resources, wherein the controller is configured to:
- host a container cluster and provision containers within the container cluster with resources including at least the processing resources and the memory resources;
- collect time series data including:
- resource utilization by each of a plurality of services in the container cluster;
- and
- network interactions between the plurality of services in the container cluster;
- construct a graph with vertices representing the plurality of services, edges representing the network interactions, and edge weights representing average quantities of network interactions;
- and
- identify a subset of the plurality of services that will be affected by unavailability of a particular service based on the graph;
- determine a respective resource usage pattern, if any, for the particular service and each of the subset of services based on the resource utilization data;
- and
- recommend a time to perform maintenance on the particular service based on the respective resource usage patterns.
12 more elements
A non-transitory machine-readable medium having instructions stored thereon, the instructions executable by a machine to:
- host a container cluster and provision containers within the container cluster with resources including at least processing resources and memory resources;
- identify services that are dependent on a particular service in the container cluster, wherein the particular service is to be unavailable for a maintenance duration before a deadline;
- for the particular service and each of the dependent services respectively:
- detect seasonality in resource utilization data on a per-resource basis;
- detect a periodicity for the detected seasonality;
- identify active intervals on a per-resource basis in the resource utilization data based on the detected periodicity;
- and
- identify overlap of the active intervals across resources;
- project the overlapped active intervals to the deadline;
- find a window, equal to the maintenance duration, within which less than a threshold quantity of the dependent services and the particular service are active based on the projection;
- and
- schedule maintenance for the particular service during the window.
9 more elements
A method, comprising:
- hosting, by a controller, a container cluster and provisioning containers within the container cluster with resources including at least processing resources and memory resources;
- by the controller and for a particular service and each of a plurality of services dependent thereon in a container cluster:
- detecting seasonality in resource utilization data on a per-resource basis;
- detecting a periodicity for the detected seasonality;
- calculating a respective moving average value of resource utilization based on the periodicity;
- categorizing the resource utilization data, for each unit of time for which the resource utilization data is recorded, as active or idle based on whether the resource utilization is within a threshold difference from the moving average value;
- identifying a respective maximum contiguous active resource utilization interval within each of a plurality of periods of the resource utilization data based on the categorization as an active interval and a remainder of each period as an idle interval;
- projecting, by the controller, the identified intervals to the maintenance deadline;
- and
- recommending, by the controller, maintenance for the particular service during a time slot, equal to a maintenance duration, within the projected idle intervals.
7 more elements
Dependencies are inferred from the claim text.