Oracle International Corporation

Vijay Sridharan

Kothanda Umamageswaran

Jia Shi

Alexander Tsukerman

Zuoyu Tao

Juan R Loaiza

Sumanta Kumar Chatterjee

Krishnan Meiyyappan

Richard L Frank

Nilesh Choudhury

Bang Nguyen

Wei Zhang

Yunrui Li

Alex Tsukerman

Graham Ivey

Adrian Tsz Him Ng

Semen Ustimenko

Zheren R Zhang

Ke Hu

Xin Zeng

Huy Nguyen

Kai Zhang

Masakazu Bando

Salini Selvaraj Kowsalya

Mehrnoosh Raoufi

Somayeh Sardashti

Thakur Vishwesh Singh

Saeed Abedigozalabad

Yiliang Jin

Zheren Zhang

The illustrative embodiments provide a consensus protocol with a priority-based leadership election approach. Each server or node in the cluster has an associated priority based on the capabilities and capacities of the servers. The leadership election is divided into two phases: a term increase phase and an election voting phase. A candidate sends a message to peers in the cluster to determine whether all peers agree to allow the candidate to increase the term number. If there is no opposition, then the candidate increases the term number and then proceeds to the election voting phase. Each peer determines whether the candidate requesting the term increase has a log that is up to date. If the log of the candidate is equally as advanced as the peer's log, then the peer approves or opposes the term increase based on a priority comparison.

Leadership election in consensus protocol with heterogeneous nodes

Described is an improved approach to implement fast detection of node death. Instead of just relying on multiple heart beats to fail in order to determine whether a node is dead, the present approach performs an on demand validation using RDMA to determine whether the node is reachable, where the approach of using RDMA is significantly faster than the heartbeat approach.

Fast node death detection

A shared storage architecture persistently stores database files in non-volatile random access memories (NVRAMs) of computing nodes of a multi-node DBMS. The computing nodes of the multi-node DBMS not only collectively store database data on NVRAMs of the computing nodes, but also host database server instances that process queries in parallel, host database sessions and database processes, and together manage access to a database stored on the NVRAMs of the computing nodes. To perform a data block read operation from persistent storage, a data block may be transferred directly over a network between NVRAM of a computing node that persistently stores the data block to a database buffer in non-volatile RAM of another computing node that requests the data block. The transfer is accomplished using remote direct memory access ('RDMA).

Database with NVDIMM as persistent storage

A method and apparatus for intelligent network resource manager for distributed computing systems is provided. A first priority is assigned to a first virtual channel set that includes at least two virtual channels of a plurality of virtual channels associated with a physical communication channel. A second priority is assigned to a second virtual channel set that includes at least one virtual channel of the plurality of virtual channels. The first virtual channel set has more virtual channels than the second virtual channel set. Outbound messages of the first priority are directed to virtual channels of the first virtual channel set. Outbound messages of the second priority are directed to virtual channels of the second virtual channel set. The virtual channels are processed in a round-robin order, where processing includes sending the outbound messages over the physical communication channel.

Intelligent network resource manager

A shared storage architecture persistently stores database files in non-volatile random access memories (NVRAMs) of computing nodes of a multi-node DBMS. The computing nodes of the multi-node DBMS not only collectively store database data on NVRAMs of the computing nodes, but also host database server instances that process queries in parallel, host database sessions and database processes, and together manage access to a database stored on the NVRAMs of the computing nodes. To perform a data block read operation from persistent storage, a data block may be transferred directly over a network between NVRAM of a computing node that persistently stores the data block to a database buffer in non-volatile RAM of another computing node that requests the data block. The transfer is accomplished using remote direct memory access ('RDMA). In addition to techniques for performing a data block read operation to NVRAM, computing nodes perform a data block write operation to data blocks stored in NVRAM of the NVRAM shared storage architecture. The data block write operation is referred to herein as a one-sided write because only one database process needs to participate in the writing of a data block to NVRAM in order to successfully commit the write.

Remote one-sided persistent writes

According to embodiments, a derived cache that is derived from a first instance of particular data is used to speed up queries and other operations over a second instance of the particular data. Traditionally, a DBMS generates and uses derived cache data only for the database data from which the derived data was derived. However, according to embodiments, derived cache data associated with a first instance of database data is relocated to the location of a second, newly created, instance of the database data. Since the derived cache data is derived from an identical copy of the database data, the cache data derived for the first instance can successfully be used to speed up applications running over the second instance of the database data.

Relocating derived cache during data rebalance to maintain application performance

In a write by-peer-reference, a storage device client writes a data block to a target storage device in the storage system by sending a write request to the target storage device, the write request specifying information used to obtain the data block from a source storage device in the storage system. The target storage device sends a read request to the source storage device for the data block. The source storage device sends the data block to the target storage device, which then writes the data block to the target storage device. The data block is thus written to the target storage device without the storage device client transmitting the data block itself to the target storage device.

Data block movement offload to storage systems

Systems, methods, and other embodiments associated with detecting a node death in a clustered distributed system are described. In one embodiment, a method includes transmitting a ping message to a peer node in the network. If a reply to the ping message is not received from the peer node, a query is sent to table of port identifiers that lists ports in the cluster. In one embodiment, the query includes a port identifier associated with the peer node. The peer node is declared as inactive/dead when the query fails to locate a match in the table for the port identifier. When the query locates a match in the table for the port identifier, another ping message is periodically transmitted to the peer node.

Node death detection by querying

Growing community of inventors

Santa Clara, CA, United States of America

Vijay Sridharan