The Yellowbrick Data Warehouse. This appliance consists of a self-sufficient hardware platform with Yellowbrick software pre-installed and pre-configured. The appliance is custom-built to support various analytic applications. Everything you need to create and use a Yellowbrick database is included in the appliance.
bulk loader (ybload)
The client program used for running bulk loads on Yellowbrick tables.
capacity expansion
The process of adding blades to an appliance.
The appliance hardware that contains the shelf of compute blades, InfiniBand switches, and power supplies. The appliance comes in single-, dual-, and triple-chassis configurations.
See HA cluster, worker cluster.
Cluster Manager
Software that manages the compute blades.
column-level encryption
Automatic encryption of sensitive data loaded into specific columns, as defined in the CREATE TABLE statement.
column store
Persistent columnar storage of table rows in shards on the worker node file systems.
compute blade
A blade with dedicated storage, CPUs, and memory, deployed as a worker node. Parallel worker processes on the blades execute the query workload for the appliance.
compute node
See worker node.
control node
See manager node.
cross-database queries
Queries that reference tables and views in different databases.
cumulative backup
A backup that records all the changes to a database since the last full or cumulative backup.
distribution key
A column in a table that is declared in the CREATE TABLE statement as the key for distributing data evenly among the analytic blades.
external table
A table that is stored outside the database in an NFS-mounted file system.
fabric blade
Hardware assembly that provides high-speed interconnect access among the compute blades and the manager node.
flex pool
A WLM resource pool with flexible concurrency settings.
Background operation that periodically moves table rows from the row store to the column store.
full backup
A backup of the entire contents of a database.
HA cluster
Cluster of two manager nodes that provide high availability (HA) for the appliance.
A starting point for a new backup may be thought of as a horizon. Full and cumulative backups reset the horizon, but incremental backups do not; therefore, a new full or cumulative backup picks up all of the changes that were already captured by an unbroken series of incremental backups.
incremental backup
A backup that records all the changes to a database since the last backup of any kind (full, cumulative, or incremental).
InfiniBand switches
Network fabric or interconnect for the cluster.
leader node
See manager node.
Lime compiler
Compiler on the control node that generates C++ code for incoming SQL commands and queries. This code is shipped to the worker nodes for processing.
manager node
A server that communicates with the outside world and assigns work to the processes on the worker nodes. The appliance comes with two manager nodes. The second node is a failover node for the first, should the first node fail or require maintenance. Only one manager node is in use at any given time.
object storage
Remote source files supported by ybload, such as Azure and S3.
A Yellowbrick database is resilient to the loss of individual compute blades. All of the data is stored in a way that allows it to be reconstructed from the failed blade(s), using file-level RAID6. Parity-protected files provide the required degree of redundancy.
PostgreSQL front-end database, running on the manager node. Connects via internal JDBC.
primary manager node
The manager node that is currently active, serving user requests.
replicated table
A table that is copied rather than distributed across all of the analytic blades so that all worker processes have immediate local access to all of its data.
The process of maintaining a replica of a physical Yellowbrick database on a secondary Yellowbrick appliance.
row store
Temporary storage for results of INSERT INTO...VALUES statements and ybsql \copy operations. Rows are periodically flushed to the column store.
A marker within a transaction. Transactions may be rolled back to savepoints.
A unit of data storage on a worker node.
shard store
See column store.
The analytic engine for the appliance is a shelf of compute blades that are provisioned as worker nodes.
A logical reference to the state of a database at a given point in time. Backup snapshots result from both backups taken with the ybbackup tool and database replication cycles.
sort key
A column in a table that is declared in the CREATE TABLE statement as the key for sorting data in the storage system on the analytic blades.
Secure Sockets Layer, a communications protocol for authenticated and encrypted connections over a network. See TLS.
stored procedure
Executable PL/pgSQL code that combines application logic and SQL constructs to perform actions in the database.
System Management Console (SMC)
UI for monitoring activity on the appliance, analyzing query plans and statistics to troubleshoot performance issues, running system management operations, running queries, and so on.
system table
A table in the system catalog that contains persistently stored data (data from log files). System tables describe database objects, storage data, and other metadata about the appliance.
system view
A view built over one or more system tables or virtual tables. System views are available from the sys schema.
Transport Layer Security, a communications protocol for authenticated and encrypted connections over a network. The TLS and SSL terms tend to be used interchangeably. SSL/TLS is also used.
virtual table
A table in the system catalog that is dynamically generated based on current activity in the appliance. These tables capture statistics for queries, CPU and memory utilization, and file system information.
worker cluster
Interconnected shelf of worker nodes (compute blades) that process Yellowbrick queries and loads in parallel. A standard configuration has a cluster of up to 15 blades.
worker node
A compute blade that is actively in use for Yellowbrick queries and load operations.
worker processes
Processes launched on the worker nodes (blades). In general, Yellowbrick queries use massively parallel processing, with all of the nodes sharing the work. One process on a given worker node is elected to run the final phase of a given query.
ybbackup, ybrestore, ybbackupctl
Yellowbrick backup and restore client tools.
Yellowbrick CLI for system management.
Yellowbrick client tool that generates DDL for users and roles.
Yellowbrick client tool for generating DDL.
Yellowbrick bulk load client tool.
Client tool that works with Apache Spark to load data from source files that ybload cannot read directly.
Yellowbrick SQL client tool.
Yellowbrick client tools package.
Yellowbrick client tool for unloading data.
Yellowbrick appliance
See appliance
Yellowbrick database
A physical database created with the Yellowbrick appliance.