Cache management for multi-node databases

Patent metadata

Field	Value
Publication number	US-10346304-B2
Application number	US-201715659544-A
Country	US
Kind code	B2
Filing date	Jul 25, 2017
Priority date	Jul 25, 2017
Publication date	Jul 9, 2019
Grant date	Jul 9, 2019

Abstract

Techniques related to cache management for multi-node databases are disclosed. In some embodiments, a system comprises one or more computing devices including a training component, data store, cache, filtering component, and listening component. The training component produces a plurality of models based on user interaction data. The plurality of models are stored in the data store, which responds to requests from the cache when the cache experiences cache misses. The cache stores a first subset of the plurality of models. The filtering component selects a second subset of the plurality of models based on one or more criteria. Furthermore, the filtering component sends the second subset of the plurality of models to a messaging service. The listening component retrieves the second subset of the plurality of models from the messaging service. Furthermore, the listening component causes the second subset of the plurality of models to be stored in the cache.

First claim

What is claimed is:

1. A system comprising:
a training component that produces a plurality of models based on user interaction data;
a cache that stores a first subset of the plurality of models
a filtering component that:
computes a difference value representing one or more value changes between one or more values in an old version of a model in the first subset and one or more values in a new version of the model,
determines whether the difference value exceeds a threshold value,
selects a second subset of the plurality of models that includes the new version of the model if the difference value exceeds the threshold value, and
sends the second subset of the plurality of models to a messaging service;

a listening component that retrieves, from the messaging service, the second subset of the plurality of models and causes the second subset of the plurality of models to be stored in the cache.

2. The system of claim 1 , wherein selection of the second subset is based on one or more criteria that are used to determine which models of the plurality of models are most likely to be requested.

3. The system of claim 1 , wherein the second subset of the plurality of models comprise models to be stored in the cache prior to the cache receiving any requests for the models.

4. The system of claim 1 , wherein the old version and the new version are each represented as a set of coefficients that are used to predict a click-through rate for a content item.

5. The system of claim 4 , wherein the filtering component, in computing the difference value for the model:
determines a set of absolute differences between the set of coefficients for the old version and the set of coefficients for the new version;
adds the set of absolute differences to compute the difference value between the old version and the new version.

6. The system of claim 1 , wherein a number of the second subset of the plurality of models is based on a size of the cache.

7. The system of claim 1 , wherein when the cache is empty, the cache is populated with models selected by the filtering component prior to retrieving, into the cache, any models from a data store.

8. The system of claim 1 , wherein selection of the second subset is based on one or more criteria that comprise a number of times that a particular model has been requested in a particular time period.

9. The system of claim 1 , wherein selection of the second subset is based on one or more criteria that comprise whether a particular model corresponds to a new content item.

10. A method comprising:
producing a plurality of models based on user interaction data;
causing a first subset of the plurality of models to be stored in a cache;
computing a difference value representing one or more value changes between one or more values in an old version of a model in the first subset and one or more values in a new version of the model;
determining whether the difference value exceeds a threshold value;
selecting a second subset of the plurality of models that includes the new version of the model if the difference value exceeds the threshold value;
sending the second subset of the plurality of models to a messaging service;
retrieving, from the messaging service, the second subset of the plurality of models;
causing the second subset of the plurality of models to be stored in the cache;
wherein the method is performed by one or more computing devices.

11. The method of claim 10 , wherein selecting the second subset of the plurality of models further comprises using one or more criteria to determine which models of the plurality of models are most likely to be requested.

12. The method of claim 10 , wherein the second subset of the plurality of models comprise models to be stored in the cache prior to the cache receiving any request for the models.

13. The method of claim 10 , wherein the old version and the new version are each represented as a set of coefficients that are used to predict a click-through rate for a content item.

14. The method of claim 13 , wherein computing the difference value for the model comprises:
determining a set of absolute differences between the set of coefficients for the old version and the set of coefficients for the new version;
adding the set of absolute differences to compute the difference value between the old version and the new version.

15. The method of claim 10 , wherein a number of the second subset of the plurality of models is based on a size of the cache.

16. The method of claim 10 , wherein when the cache is empty, the cache is populated with models selected based on one or more criteria prior to retrieving, into the cache, any models from a data store.

17. The method of claim 10 , wherein selecting the second subset of the plurality of models further comprises using one or more criteria comprising a number of times that a particular model has been requested in a particular time period.

18. The method of claim 10 , wherein selecting the second subset of the plurality of models further comprises using one or more criteria comprising whether a particular model corresponds to a new content item.

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Xu Liqin
• Pu Wen
• Ramanath Rohan
• Liu Kun

Classifications

• G06F11/3438

  monitoring of user actions (tracking the activity of the user H04L67/535) · CPC title

• G06F2201/80

  Database-specific techniques · CPC title

• G06F11/073

  in a memory management context, e.g. virtual memory or cache management (memory management G06F12/00; testing of static memory units G11C29/00) · CPC title

• G06F12/0815 Primary

  Cache consistency protocols · CPC title

• G06F16/24552 Primary

  Database cache management · CPC title