US 12,216,653 B1Granted

Thrashing detection and avoidance for a tiered data processing service

AssigneeAmazon Technologies, Inc.

InventorsInduja Sreekanthan, Sriram Subramanian, Athanasios Papathanasiou, Vijayan Prabhakaran

Life of this patent

Estimating the term…

Loading patent…

Abstract

Methods for improving performance of tiered storage of a data processing service by detecting and reducing thrashing of data blocks between warm and cold storage tiers are disclosed. In order to understand the frequency of hits by incoming queries to data blocks that are not currently stored in the warm storage tier, the elapsed time between query hits to the respective data blocks may be tracked using timers. Times below a given amount of time may be used to indicate thrashing. For example, recently evicted data blocks that are subsequently hit by a query within a short amount of time since eviction may indicate thrashing. In scenarios in which thrashing may be occurring, a threshold corresponding to the number of times a given data block in the cold storage tier receives a query hit before being added to the warm storage tier may be turned on.

Claims19

Dependent claims

A system, comprising:

one or more computing devices configured to implement a data processing service, wherein the data processing service uses a plurality of storage tiers to store data blocks, the plurality of storage tiers comprising at least:
a warm storage tier implemented using storage included in nodes of a node cluster of the data processing service, wherein the node cluster has a first storage capacity to store a first quantity of data blocks;
and
a cold storage tier implemented using storage of a data storage service;
and
wherein the data processing service is further configured to:
maintain, in one or more buffers of the warm storage tier, metadata comprising block identifiers for an additional number of data blocks that are stored in the cold storage tier that are in addition to the first quantity of data blocks stored in the warm storage tier, wherein the additional number of data blocks that are stored in the cold storage tier, and for which metadata comprising block identifiers are maintained in the one or more buffers, are data blocks which have previously been evicted from the warm storage tier with a number of query hits below a threshold number;
responsive to a query targeting a data block stored in the cold storage tier with a respective block identifier included in the one or more buffers of the warm storage tier, increment a counter corresponding to the block identifier for the data block targeted by the query;
and
promote the data block, targeted by the query, from the cold storage tier to the warm storage tier, in response to the counter corresponding to the block identifier exceeding a threshold count, wherein the threshold count is two or more.

7 more elements

A method comprising:

storing data blocks, for use by a data processing service, in a plurality of storage tiers, wherein the plurality of storage tiers comprise at least:
a warm storage tier implemented using storage included in nodes of a node cluster of the data processing service, wherein the node cluster has a storage capacity to store a first quantity of data blocks;
and
a cold storage tier implemented using storage of a data storage service;
maintaining, in one or more buffers of the warm storage tier, metadata comprising block identifiers for an additional number of data blocks that are stored in the cold storage tier;
tracking, via the metadata maintained in the one or more buffers of the warm storage tier, respective elapsed times since the data blocks having block identifiers included in the one or more buffers of the warm storage tier have been accessed via a query;
adjusting a threshold used by the node cluster having the first storage capacity based on the respective elapsed times tracked via the metadata, wherein a given elapsed time less than a time threshold indicates thrashing of one or more data blocks of the first quantity of data blocks between the warm storage tier and the cold storage tier, and wherein adjusting the threshold comprises increasing a query hit count required for a data block to be promoted from the cold storage tier to the warm storage tier when thrashing is indicated;
and
promoting one or more data blocks to the warm storage tier based on the adjusted threshold.

6 more elements

One or more non-transitory, computer-readable, storage media storing program instructions that, when executed on or across one or more processors, cause the one or more processors to:

maintain, in one or more buffers of a warm storage tier, metadata comprising block identifiers for an additional number of data blocks that are stored in the cold storage tier that are in addition to a first quantity of data blocks stored in the warm storage tier, wherein the additional number of data blocks that are stored in the cold storage tier, and for which metadata comprising block identifiers are maintained in the one or more buffers, are data blocks which have previously been evicted from the warm storage tier with a number of query hits below a threshold number;
responsive to a query targeting a data block stored in a cold storage tier with a respective block identifier included in the one or more buffers of the warm storage tier, increment a counter corresponding to the block identifier for the data block targeted by the query;
and
promote the data block targeted by the query from the cold storage tier to the warm storage tier, in response to the counter corresponding to the block identifier exceeding a threshold count, wherein the threshold count is two or more.

1 more element

Dependencies are inferred from the claim text.