Amazon Technologies, Inc.

Brian S O'Neill

Archit Shivaprakash

Omer Ahmed Zaki

Kiran Kumar Muniswamy Reddy

Olga I Kuznetsova

Tobias L Holgers

Vishal Parakh

Anand Sasidharan

Jason Shijia Long

Matthew Lloyd Trahan

Diwakar Chakravarthy

Gary P Raden, Ii

Kevin Christen

A fleet of query accelerator nodes is established for a data store. Each accelerator node caches data items of the data store locally. In response to determining that an eviction criterion has been met, one accelerator node removes a particular data item from its local cache without notifying any other accelerator node. After the particular data item has been removed, a second accelerator node receives a read query for the particular data item and provides a response using a locally-cached replica of the data item.

Independent evictions from datastore accelerator fleet nodes

Generally described, the present disclosure is directed to an eventually consistent replicated data store that uses, for its underlying storage, a computer software library that provides a high-performance embedded database for data. The replicated data store employs a plurality of hosts interconnected to one another, allowing for writes to any host and full awareness of membership across all hosts. With the data replication framework disclosed herein, various modes are allowed to be built up on top of the core system.

Data replication framework

A fleet of query accelerator nodes is established for a data store. A particular node stores a binary encoding of a data item in a local data cache. The encoding identifies a metadata record indicating names of attributes of the data item. In response to receiving a particular query directed to the data item, an executable code component generated to produce a result for the query is cached in a local query code cache. When a second query which meets a matching criterion is received, the executable code component is re-used to obtain a result for the second query.

Binary encoding-based optimizations at datastore accelerators

An accelerator intermediary node (AIN) associated with a data store obtains an indication of a control setting to be applied with respect to a write request directed to a data item, where the control setting specifies a target for one or more of replication count, data durability, a transaction grouping with respect to a write request, or back-end synchronization node. Using the control setting, a write propagation node set is identified for the write request. The write propagation node set includes another accelerator intermediary node and/or a storage node of a data store. Respective operation requests corresponding to the write request are transmitted to one or more members of the write propagation node set. A write coordinator role may be verified prior to attempting a commit together of a plurality of write requests as part of a multi-write transaction.

Using data store accelerator intermediary nodes and write control settings to identify write propagation nodes

A fleet of query accelerator nodes is established for a data store. A particular node determines a write throughput limit for back-end writes from the fleet to the data store. A pool of write-back threads is established at the node. When a write request is received at the node, an entry is added to a selected partition of a write-back journal, and a write completion response is provided to the requester without applying the write to the data store. A write-back thread from the pool eventually schedules the back-end write.

Write-back techniques at datastore accelerators

Disclosed are various embodiments for executing multiple applications in a single virtual machine. The classes of an application executing in the virtual machine are traversed to identify non-sharable static fields. Upon identifying a non-sharable static field, mapping data is created that corresponds to the non-sharable static field. During another traversal of the classes of the application, access to the identified non-sharable static field is translated into an access to the mapping data that is associated with the non-sharable static field.

Sharing applications in a java virtual machine

Systems and methods are directed to an eventually consistent replicated data store that uses, for its underlying storage, a computer software library that provides a high-performance embedded database for data. The replicated data store employs a plurality of hosts interconnected to one another, allowing for writes to any host and full awareness of membership across all hosts. With the data replication framework, various modes are allowed to be built up on top of the core system.

Disclosed are various embodiments for executing multiple Java applications in a single Java virtual machine. Each Java application is stored in a Java distribution and includes a plurality of classes. The executing includes translating an access to a non-shareable static field within one of the classes into an access to mapping data associated with the non-shareable static field. The executing further includes executing an initializer associated with the non-shareable static field when the mapping data is created.

Techniques for generating and vending of unique sequence values are described. Each master of a ring of masters is configured to create and vend the unique sequence values downstream to filling pools for eventual client consumption. When a ring of masters is employed, each master communicates with one another to agree on a current range of values that should next be vended. A master may then vend these values to a layer of one or more filling pools, at least of which includes or has access to persistent and non-persistent storage memories.

Generation and vending of unique sequence values

An eventually consistent replicated data store uses, for its underlying storage, a computer software library that provides a high-performance embedded database for data. The replicated data store employs a plurality of hosts interconnected to one another, allowing for writes to any host and full awareness of membership across all hosts. With such a data replication framework, various modes are allowed to be built up on top of the core system.

Growing community of inventors

Bellevue, WA, United States of America

Brian S O'Neill