公开(公告)号：US10523712B1

公开(公告)日：2019-12-31

申请号：US15604484

申请日：2017-05-24

Applicant: Amazon Technologies, Inc.

Inventor： Wei Huang

IPC: H04L29/06 ,  G06N7/08 ,  G06F17/17 ,  G06Q30/02

Abstract: A stochastic estimation approach can be used to provide percentile determinations for data, such as may be received on one or more data streams. A stochastic approach can analyze each received request on a data stream and update the percentile values accordingly, providing near real time adjustment of the percentile values. One or more scaling factors or adjustment boundaries can be applied such that the estimations do not fluctuate excessively in response to individual requests. The near real time updates enable actions to be taken on the data stream, such as to generate alarms or initiate request throttling. The stochastic approach is very light weight, requiring minimal resources and having minimal latency.

公开(公告)号：US10372572B1

公开(公告)日：2019-08-06

申请号：US15365852

申请日：2016-11-30

Applicant: Amazon Technologies, Inc.

Inventor： Nitin Kesarwani ,  Sharon Xuelun Yang ,  Wei Huang

IPC: G06F11/26 ,  G06N7/00

Abstract: A prediction model testing system includes a test environment that is used to test a prediction model under test (PMUT). A metrics collector in a production environment collects and stores production metrics data generated from computing resources in a production environment. A production predictor in the production environment generates production predictions for the metrics, using a production prediction model. A test manager may make the production metrics data available in a test environment. Test predictions are generated in the test environment from the metrics data using the PMUT. The test manager may then calculate respective prediction errors of the production prediction model and the PMUT, and generate a report indicating the differences between the two sets of prediction errors. The report may be used by the test management system to determine whether a test of the PMUT was successful.