US 11,388,210 B1Granted

Streaming analytics using a serverless compute system

AssigneeAmazon Technologies, Inc.

InventorsVinayak Sood, Jia Liu, Mandakini Saroop, Tejas Mahadeo Ghadge, Himanshu Sharma, Nithin Vommi, Tyson Charles Olychuck, Dinesh Saheblal Gupta, Peng Sun

Life of this patent

Estimating the term…

patentgrep

Loading patent…

Abstract

Systems and methods are described implementing streaming analytics on a serverless compute system. A user can specify a data stream against which analytics should be conducted, serverless functions to be used to conduct the analysis, and criteria for the analytics, such as windowing criteria and aggregation criteria. The windowing criteria can specify windows of items within the stream that should be collectively analyzed. The aggregation criteria can specify how to group items within each window during analysis. A poller device can read data items from the stream, window and group the items according to the windowing and aggregation criteria, and invoke serverless functions to conduct streaming analytics on the data items. The poller device can further maintain state between invocations.

Claims20

Dependent claims

A system for implementing streaming analytics using serverless code executions, the system comprising:

a streaming data system comprising a set of computing devices configured to host a data stream comprising messages, individual messages within the data stream being associated with timestamps corresponding to relative positions within the data stream;
a serverless computing system configured to obtain invocations of serverless functions and in response initiate execution of the serverless functions, wherein the serverless functions comprise:
an aggregation function representing code executable to process one or more input messages from the data stream and generate state information representing analysis of the one or more input messages;
and
a destination function representing code executable to process state information from the aggregation function;
and
a poller device configured to:
iteratively retrieve messages from the data stream;
using window criteria, assign each retrieved message to a window from a plurality of windows;
and
for each window of the plurality of windows:
group messages assigned to the window according to aggregation criteria to result in at least a first group of messages and a second group of messages;
invoke a first execution of the aggregation function to process the first group of messages;
obtain first state information from the first execution;
invoke a second execution of the aggregation function to process the second group of messages at least partly by passing the first state information to the second execution, wherein the second execution of the aggregation function results in second state information;
and
invoke an execution of the destination function at least partly by passing to the execution of the destination function the second state information, wherein execution of the destination function provides processing of a result of conducting the streaming analytics against the messages assigned to the window.

14 more elements

A computer-implemented method comprising:

iteratively retrieving messages from a data stream;
using window criteria, assigning each retrieved message to a window from a plurality of windows;
and
for each window of the plurality of windows:
grouping messages assigned to the window according to aggregation criteria to result in at least a first group of messages and a second group of messages;
invoking a first execution of an aggregation function on a serverless computing system, wherein the aggregation function represents code executable to process an input group of messages and provide state information as a result, wherein the first group of messages represents the input group for the first execution, and wherein the result of the first execution is first state information;
obtaining the first state information from the first execution;
invoking a second execution of the aggregation function on the serverless computing system, wherein the second group of messages represents the input group for the second execution, wherein invoking the second execution comprises passing the first state information to the second execution, and wherein the result of the second execution is second state information;
and
invoking an execution of a destination function on the serverless compute system, wherein the destination function represents code executable to process input state information from the aggregation function and provide a result, and wherein the second state information represents the input state information for the execution of the destination function.

7 more elements

Non-transitory computer-readable media comprising instructions that, when executed by a computing system, causes the computing system to:

iteratively retrieve messages from a data stream;
using window criteria, assign each retrieved message to a window from a plurality of windows;
and
for each window of the plurality of windows:
group messages assigned to the window according to aggregation criteria to result in at least a first group of messages and a second group of messages;
invoke a first execution of an aggregation function on a serverless computing system, wherein the aggregation function represents code executable to process an input group of messages and provide state information as a result, wherein the first group of messages represents the input group for the first execution, and wherein the result of the first execution is first state information;
invoke a second execution of the aggregation function on the serverless computing system, wherein the second group of messages represents the input group for the second execution, wherein invoking the second execution comprises passing the first state information to the second execution, and wherein the result of the second execution is second state information;
and
invoke an execution of a destination function on the serverless compute system, wherein the destination function represents code executable to process input state information from the aggregation function and provide a result, and wherein the second stale information represents the input stale information for the execution of the destination function.

6 more elements

Dependencies are inferred from the claim text.