International Business Machines Corporation

Salesforce, Inc.

Salesforce Com, Inc.

Venkateswararao Jujjuri

Malahal R Naineni

James John Seeger, Jr

Craig Fulmer Everhart

Ninad S Palsule

Badari Pulavarty

Aneesh K Veetil

Thomas K Clark

Anup Ghatage

Rithin Kumar Shetty

Jason Christopher Young

Nathaniel Wyatt

Chandra Seetharaman

Sukadev Bhattiprolu

Narasimha Nageshrao Sharoff

Charan Reddy Guttapalem

Sripathi Kodi

Mohan Kumar Mohan Raj

Andrey Yegorov

Steven Michael French

Robert Michael Rees

Eric Van Hensbergen

Ankit K Jain

Patrick James Helland

Rajagopal Ananthanarayanan

Anthony Nicholas Liguori

Alexei Kojenov

Haren Myneni

Samuel Wynne Yang

Russell H Weight

Hemanth Siddulugari

Paul A Dorwin

William K Bittner

James D Beecham

James Patrick Smith

Roger C Raphael

Jameison Bear Martin

Sameer Tiwari

Christian Bolik

Ujjwal Lanjewar

Ashish Chaurasia

Rohit Krishna Prasad

Jayant Kumar

Senthil Rajaram

Steven Raspudic

Malahal R Nalnenl

Karan Jitendra Mehta

Sushanth Rai

Ankit Jain

Techniques are disclosed relating to storing location information about storage nodes in cookies. A cloud-based service may send location requests to an orchestration service that instantiated storage nodes included in a storage cluster of the cloud-based service. The cloud-based service may receive location information that identifies in which computer zone that a given storage node is located. The cloud-based service may store the location information in cookies at a metadata store that is shared among the storage nodes. The cloud-based service may receive, from a client node, a search request to identify ones of the storage nodes that store particular data. The cloud-based service may return a set of cookies corresponding to identified storage nodes. The set of cookies may enable the client node to determine whether there is a storage node that stores the particular data and is within the same computer zone as the client node.

Cookie-based network location of storage nodes in cloud

Techniques are disclosed relating to migrating database extents between physical storage nodes. A database system stores current and new assignment metadata mapping virtual groupings of database extents physical storage nodes, the new assignment metadata reflecting a data migration of extents between the plurality of physical storage nodes. During the migration, the database system receives 1) read requests, which it responds to by reading data from a first physical storage node identified using the current assignment metadata and 2) write requests, which it responds to by writing data to a second physical storage node identified using the new assignment metadata. Upon the migration being complete, the system identifies physical storage nodes accessed by subsequent read and write requests using the new assignment metadata.

Database request handling during data migration between physical nodes

Techniques are disclosed relating to upgrade groups. A node of a computer system may access metadata assigned to the node during deployment of the node. The node may be one of a plurality of nodes associated with a service that is implemented by the computer system. The node may perform an operation on the metadata to derive a group identifier for the node and the group identifier may indicate the node's membership in one of a set of groups of nodes managed by the service. The node may then store the group identifier in a location accessible to the service.

Mechanisms for grouping nodes

Techniques are disclosed for managing metadata of a distributed database system in a hybrid manner. A computer system may receive, from a computing device, a request to access a set of data stored in nodes of a distributed storage system that is a caching layer of the system. The system retrieves metadata for a set of data specified in the request, including accessing a reversemap storing a reverse-ordered copy of original metadata stored in a metadata store of the system, where the reversemap is stored on durable storage of the system. Based on retrieving the metadata from the reversemap, the system accesses nodes of the distributed storage system, where the reversemap specifies the nodes of the distributed storage system that store the set of data. The system transmits, to the computing device, information indicating a result of accessing data stored in nodes of the distributed storage system.

Hybrid database system for strongly consistent and highly scalable metadata storage

Techniques are disclosed relating to managing distributed storage of data for various entities according to classifications for these entities. A database node of a distributed storage system may receive, from a first entity of a plurality of entities, a request to store a set of data. The database node may further obtain metadata associated with the first entity, wherein the metadata specifies one of a plurality of classifications for the entities. The database node may provide the set of data to one or more of a plurality of caches for storage. The caches may be located in two or more availability zones and are configured to store the set of data based on the classification for the first entity identified in the metadata associated with the first entity. The database node may also store the set of data in a shared object storage coupled to the database node.

Cloud storage class-based variable cache availability

Techniques are disclosed relating to managing distributed storage of data across availability zones and the replication of data in case of storage server failures. A distributed storage system may include storage servers distributed across availability zones with an auditor instantiated in at least one storage server. The auditor manages the replication of data in the event of one or more storage servers failure in an availability zone. In the event of the failure, the auditor may determine the extent of the failure and whether the failure involves a small number or a large number of storage servers. In the event a large number of storage servers being affected, the auditor may delay replication of data to see if the failure is temporary and avoid unwanted data transfer of large amounts of data across availability zones.

Grey failure handling in distributed storage systems

Techniques are disclosed relating to a database system synchronizing a current state of the database system among a plurality of nodes configured to handle requests for data of the database system stored in a distributed storage with one of the plurality of nodes being currently active and the other nodes of the plurality of nodes being currently standby nodes. In various embodiments, the active node receives a request to perform a transaction that includes committing data to the distributed storage. In response to receiving the request, in some embodiments, the active node commits the data to the distributed storage to update the current state of the database system and causes storing of metadata that provides an indication of the commitment in a transaction log stored in the distributed storage.

High availability database through distributed store

Techniques are disclosed relating to fencing out a first one of a plurality of nodes configured to handle requests for data stored in a distributed storage. A database system, in various embodiments, stores a value indicating that the first node is permitted to update a catalog stored at a metadata server. In response to a determination to prevent the first node from updating the catalog and writing to the distributed storage, in various embodiments, the database system updates the value to indicate that the first node is not permitted to update the catalog and instructs the distributed storage to prevent write operations to a particular portion allocated to the first node for writing data.

Fencing out nodes in a distributed clustered system

In some embodiments, a server node may transmit metrics to a client piggy-backed on (appended to) responses to regular read/write requests. Other embodiments may be disclosed and/or claimed.

Dissemination of node metrics in server clusters

Among heterogeneous nodes in a cluster, one node's capabilities can vary widely from that of others in the same cluster, for example, in a cluster of storage servers. As each of the clients/writers to the cluster may contact different server nodes in the cluster over time, with each contact the client/writer may receive server metric information, including both the contacted node's current metrics as well as metrics for any other nodes of which the contacted node is aware. The client/writer may provide updated information to the contacted node regarding the metrics for other nodes of which the client/writer is aware or has contacted. The exchange of metric information may be piggybacked upon read/write requests to each node to conserve bandwidth.

Opportunistic gossip-type dissemination of node metrics in server clusters

Distributed Systems are becoming very large and are being served by 10s of 1000s of individual machines. Among heterogeneous nodes, one node's capabilities can vary widely from that of others in the same cluster, for example, in a cluster of storage servers. Each of the clients/writers may retrieve, periodically, from all the servers, important metrics such as free disk space capacity, average disk IO utilization over the last N minutes, etc. We call these client to nodes (servers) messages as 'periodic metrics query.' Using received metrics, clients dynamically assign weights to each of the servers as an indication of their current capabilities. Server metrics and or weights are distributed across the clients, and each client may make independent placement decisions based on their view of the cluster. Because clients are using substantially the same information for making decisions on placement, the result is improved balance across the servers.

Decentralized, resource aware load distribution in a distributed system

Log entries are provided with unique entry identifiers, which may be sequenced in an incremental or decremental order, to create or to update a distributed log which may be replicated and distributed to multiple servers in a network. The entry identifiers may be appended to their respective log entries. Files, records or data which are identified by their respective log entries may be transmitted regardless of the sequence of the entry identifiers.

Coherent distributed logging

Growing community of inventors

Beaverton, OR, United States of America

Venkateswararao Jujjuri