Verizon Patent and Licensing Inc.

Oath Inc.

Verizon Media Inc.

Yahoo Ad Tech LLC

Idit Keidar

Edward Bortnikov

Eshcar Hillel

Yonatan Gottesman

Anastasia Braginsky

Ohad Shacham

Eran Gilad

Guy Golan Gueta

Dmitry Basin

Lee Rhodes

Gali Sheffi

Maya Arbel

Sameer Paranjpye

Alexander Spiegelman

David Carmel

Hagar Meir

The disclosed embodiments describe techniques for isolating and managing models via versioning. In one embodiment, a method is disclosed comprising reading a configuration document associated with a first model; incrementing an internal write version of the configuration document; storing the internal write version in the configuration document; generating documents belonging to a second model such that respective document identifiers of the documents include a next external version a first field and the incremented internal write version in a second field; and uploading the documents to a serving system, causing the serving system to replace the first model with the second model.

Method and system for managing embeddings and serving recommendations in real-time

The present teaching relates to a method, system and programming for recovering a data storage. In response to a failure of the data storage, a last checkpointed version is retrieved from a checkpoint file created previously for the failure of the data storage. A recovery table is updated to indicate the last checkpointed version and an incarnation number indicating a number of previous recovery of the data storage. The data storage is recovered based on the recovery table. A version number of the last checkpointed version is reset to zero upon a successful recovery of the data storage. The incarnation number is incremented.

Method and system for searching a key-value storage

Methods, systems, and programming for retrieving content items for a search are described herein. In a non-limiting embodiment, a query including a plurality of terms may be received. For each of the plurality of terms, a posting list of one or more content items may be obtained. The posting list may include a ranked list of term scores corresponding to the one or more content items, each of the term scores being indicative of a level of relevance of a corresponding content item to a term associated with the posting list. A list of relevant content items for the query may be determined based on the term scores in each posting list for the one or more content items identified with respect to each term. At least one of the relevant content items may be provided as a response to the query.

System and method for top-k searching using parallel processing

The present teaching relates to a method, system and programming for operating a data storage. The data storage comprises of different portions including: a first portion having a plurality of metadata objects stored therein, each of the metadata objects being associated with a filter and corresponding to a range of keys, wherein at least one of the metadata objects is associated with a data structure, and a second portion having a plurality of files stored therein, each of the plurality of files being associated with one of the plurality of metadata objects; The data storage synchronizes a scan request with respect to one or more write requests based on a parameter associated with the scan request and each of the one or more write requests.

Method and system for synchronizing requests related to key-value storage having different portions

The present teaching relates to a method and system generating a data sketch. A plurality of local data sketches are generated, wherein for each local data sketch, it is determined whether a propagation condition is satisfied. In response to the condition being satisfied, the local data sketch is processed when a synchronization parameter satisfies a criterion. The synchronization parameter is set to a predetermined value to enable propagation of the local data sketch to the data sketch. Upon the local data sketch being propagated to the data sketch, the above steps are repeated.

Method and system for concurrent generation of data sketches

The present teaching relates to a system and method for providing a response to a query related to a data sketch. A local data sketch is obtained in response to a synchronization parameter being set to a first value. The data sketch is updated based on the obtained local data sketch. The synchronization parameter is set to a second value that is different than the first value and thereafter propagated. A response to a query is generated based on the updated data sketch.

Multi-thread systems and methods are described for concurrently handling requests to commit data updates to a database by a plurality of data transactions. The database preferably supports multi-versioning and the data transactions are preferably isolated by snapshot isolation. In one embodiment, concurrent and lock-free handling of requests to commit data updates includes performing two types of concurrent data conflict detection. A transaction proceeds to commit only if it passes both types of conflict detection. The first type of conflict detection is based on a hash map between data keys and their commit timestamps whereas the second type of conflict detection is based on a log that keeps track of the status of transactions whose requests to commit are actively being processed. In another embodiment, concurrent conflict detection for data items in concurrent transactions is broken down into buckets and locks are used for accessing each bucket. These systems and methods maintain transactional integrity to database while improving throughput by maximizing concurrency of data commits in a multi-thread environment.

Scalable conflict detection in transaction management

The present teaching relates to a method, system and programming for searching a data storage. A key is extracted from a request and a metadata object associated with the key is identified. Further, a determination is made as to whether the metadata object is associated with a data structure stored in a first portion of the data storage. In response to a successful determination, the data structure is searched to retrieve a value associated with the key from the first portion. In response to an unsuccessful determination, a cache is searched to retrieve the value associated with the key, and in response to the key being absent in the cache, a file associated with the metadata object is searched to retrieve the value associated with the key, wherein the file is stored in a second portion of the data storage.

In one embodiment, a set of lock and unlock instructions in a read phase of a computer-readable program is replaced with a first set of tracking instructions, wherein the first set of tracking instructions track a set of locked objects identifying objects that would have been locked by executing the set of lock and unlock instructions. A second set of tracking instructions is inserted into the read phase of the computer-readable program, wherein the second set of tracking instructions track a set of read objects indicating versions of objects that are read. Validation instructions are inserted into the computer-readable program, wherein the validation instructions validate that the versions of objects in the set of read objects have not changed since they were last read and lock the set of locked objects that would have been locked upon completing execution of the set of lock and unlock instructions. Update instructions are added to an update phase of the computer-readable program, where the update instructions increment a current version of an object each time a value of the object is updated or a lock of the object is released.

Automatic lock removal method for scalable synchronization in dynamic data structures

Briefly, embodiments disclosed herein relate to managing transactions in a distributed content transaction system.

Management of transactions in a distributed transaction system

Methods, systems, and programming for accessing data are described herein. In a non-limiting embodiment, a request associated with a key may be received. A memory chunk in a first memory site associated with the key may be located. An entry in the memory chunk corresponding to the key may be determined. A first pointer to the key corresponding to a first location of the key within a buffer at a second memory site may be obtained from the entry. The key stored at the first location may be retrieved from the buffer at the second memory site.

System and method for managing memory for large keys and values

The present teaching relates to committing transactions over a database in a semi-distributed manner. In one example, a transaction is executed to access a plurality of keys stored in the database. A request is transmitted to a transaction management engine to commit the transaction. The transaction management engine performs a conflict check between the transaction and one or more transactions running concurrently with the transaction, and generates an instruction based on the conflict check. The instruction is received from the transaction management engine. Based on the instruction, one or more of the plurality of keys are updated in the database or data added by the transaction to the database are erased.

Growing community of inventors

Haifa, Israel

Idit Keidar