Mellanox Technologies, Ltd.

Amdocs Software Systems Limited

Guy Shattah

Lior Narkis

Ariel Almog

Ron Yuval Efraim

Michael Binshtock

A method for optimizing hash table lookup speed during hash table resize on a computing device, the method including performing the following on the computing device: providing a first hash table having N slots for entries, designating the first hash table as an active hash table, allocating a second hash table, and performing the following after allocating the second hash table: when a hash table insertion of an entry is requested, performing insertion by inserting the entry to the first hash table and inserting the entry to the second hash table, and when a hash table lookup is requested, looking up the requested entry in the active hash table, one of the performing insertion and the performing deletion including also copying K entries, K being greater than or equal to 1, from the first hash table to the second hash table. Related apparatus and methods are also described.

Equipment and method for hash table resizing

A mechanism is provided for partial offload of connection tracking from a host processor to a network interface device. Software running in the host processor is used for connection establishment. After a connection has been established the software initializes and transfers flow-control to the network interface device. Thereafter, the network interface device continues transferring packets to the destination, while the software in the host processor monitors the flow. If the connection has been terminated or has expired flow control reverts to the software. Modes of operation are provided for network interface devices with and without a specific connection-tracking module.

Stateful connection tracking

Described embodiments include an apparatus that includes circuitry, configured to facilitate writing to a shared memory, and a processor. The processor is configured to compute a local current-version number by incrementing a shared current-version number that is stored in the shared memory. The processor is further configured to, subsequently to computing the local current-version number, using the circuitry, atomically write at least part of the local current-version number to a portion of the shared memory that is referenced by the local current-version number. The processor is further configured to, subsequently to atomically writing the at least part of the local current-version number, store data in the shared memory in association with the at least part of the local current-version number, and subsequently to storing the data, atomically overwrite the shared current-version number with the local current-version number. Other embodiments are also described.

Atomic updates of versioned data structures

A system, method, and computer program product are provided for determining at least potentially duplicate records. In use, when a new record is received, a first entry of a first predetermined field of the newly received record is identified, and a data structure is selected, based on the first entry. Further, a second entry of a second predetermined field of the newly received record is identified, and the second entry is compared to entries already stored in the data structure, where the entries stored in the data structure are associated with a plurality of other received records. Still yet, it is determined whether the newly received record is at least potentially a duplicate of at least one of the other received records, based on the comparison.

System, method, and computer program for determining most of the non duplicate records in high performance environments in an economical and fault-tolerant manner

Growing community of inventors

Tel Aviv, Israel

Guy Shattah