7.0 Release Notes

Active Analytics Workbench

• The Active Analytics Workbench (AAW) platform is now available. A brand new API and UI are available to improve existing machine learning workflows. For machine learning, Tensorflow and Blackbox (using Docker containers) models are supported. AAW also supports continuous deployments, on-demand deployments (via an inferencing REST endpoint), and audits of both TensorFlow models trained internally and imported models. AAW has a new ingestion interface that allows ingesting data from variety of sources and methods, including Kinetica, PostgreSQL, and Kafka.

Core

• Kinetica now supports Resource Management. Resource management involves the three following resources:
  • Storage Tiers: Data containment layers within the database (VRAM, RAM, Disk Cache, Cold Storage)
  • Tier Strategies: Data object eviction priorities within each storage tier to help define memory usage and data priorities
  • Resource Groups: Resource fencing -- process scheduling priorities and other limits imposed on specific groups of users
• The ability to alter multiple columns in a single call is possible via the new /alter/table/columns endpoint.

Geospatial / Network Graph

• Kinetica now includes a graph server. The graph server provides a generic and extensible design of networks with the aim of being tailored or used for various real-life applications, including transportation, utility, social, and geospatial.
• Kinetica now offers a Vector Tile Service (VTS) to generate Vector Tiles and support client-side visualization of geospatial data contained within the tiles. Generating Vector Tiles using Kinetica involves passing in the VTS URL to the client-side renderer.
• The /wms endpoint now supports contour plot visualization functionality

KAgent

• The Kinetica Agent (KAgent) UI is now available to automate Kinetica installation and configuration. KAgent can automatically install Nvidia drivers for CUDA installations, the Active Analytics Workbench (AAW) and Kubernetes (required for AAW), configure the cluster for SSL and/or external authentication and high availability. KAgent is also cloud ready and able to provision and/or deploy to instances in Amazon Web Services (AWS), Google Cloud Platform (GCP), or Microsoft Azure. KAgent also allows one to configure the location of the head node, AAW, and the graph server within the cluster. Upgrades will now be managed using KAgent.

SQL / OLAP

• Kinetica is now packaged with a new and improved SQL interpreter.
• A new /execute/sql endpoint is available. You can now send SQL commands to the native API without an ODBC driver.
• The SQL engine uses a query planner that analyzes an entire query for the many different ways it can be solved (e.g., performing a filter before a join, performing a filter after a join, etc.) and selecting the plan that is the most efficient. Once this plan is utilized, it's cached so the plan can be used again without having to re-compute it when the same query is received again.
• Complex SQL queries that involve multiple operations are now analyzed for interdependency. Any operations that have no dependencies on others are executed in parallel.
• Distributed operations, such as UNIONs between sharded and replicated tables or distributed JOINs, are now possible using SQL syntax or /execute/sql. The database will automatically re-shard or replicate tables temporarily as necessary to help process the query. Note that the distributed query can be slower and use more memory than a traditional non-distributed operation. The sql.distributed_joins setting in /opt/gpudb/core/etc/gpudb.conf controls the ability to use distributed operations.
• Correlated sub-queries are now supported.
• New SQL support for:
  • EXPLAIN
  • Partitions
  • Tier Strategy Definitions
  • Logging Levels
• Partitioning is now available. Table data that is sharded or replicated can be partitioned to aid storage tiering and data skipping. The two types of supported partitioning schemes are:
  • Range
  • Interval

UI

• The Kinetica Administration Application (GAdmin) has undergone a visual refresh and now supports functionality for managing resource groups, interfacing with tier strategies and partitions, reviewing graph node and edge counts, and deleting graphs.
• Kinetica Reveal service has undergone a visual refresh as well as received some usability improvements.

SQL/OLAP

• Added support for partitioning track tables (by TRACKID); automatic list partitioning, hash partitioning, and series partitioning schemes can be used
• Added support for IGNORE NULLS and RESPECT NULLS in the following window function types:
  • FIRST_VALUE
  • LAST_VALUE
  • LEAD
  • LAG
• Added additional framing options for PRODUCT window functions
• Added support for JOIN and subqueries during SQL UPDATE operations
• Added support for SQL procedures to be executed as another user using the EXECUTE AS syntax
• Added four new functions:
  • LOG1P(expr) -- Returns the natural logarithm of one plus expr (a double). Effectively evalutes as log(1+expr)
  • LOG2(expr) -- Returns the binary (base-2) logarithm of the given expression expr (a double)
  • LOG1PF(expr) -- Returns the natural logarithm of one plus expr (a float). Effectively evalutes as log(1+expr)
  • LOG2F(expr) -- Returns the binary (base-2) logarithm of the given expression expr (a float)
• Added support for short-circuiting errors using the conditional function IF and operators AND & OR
• Improved performance of JOIN operations involving inequalities between columns in different tables
• Added support for /insert/records returning any records not inserted due to conflicting primary keys. This is achieved when the user provides either the allow_partial_batch or return_individual_errors flag when update_on_existing_pk is set to false
• Improved performance of /append/records operations; i.e., SQL INSERT with SELECT
• Added support for returning an error when a timestamp is out of range
• Added support for init_with_now within /append/records
• Removed support for using temporary tables within procedures

Geospatial/Network Graph/Visualization

• Improved performance of feature and classbreak rendering in large clusters
• Added support for symbology in classbreak feature rendering
• Added support for track head symbology in non-classbreak feature rendering

AAW

• Added support for Kafka Cloud SASL credentials

APIs/Tools/Connectors

• Python API
  • Added support for custom headers, which enables SQL-like user impersonation at the native API level
  • Improved and simplified the table monitor API
  • Added support for sharding on UUID type columns in multihead egress/ingest
• JavaScript API
  • Added support for custom headers, which enables SQL-like user impersonation at the native API level
• Node.js API
  • Added support for custom headers, which enables SQL-like user impersonation at the native API level

UI

• Reveal
  • Added support for sorting histogram columns by field name

SQL/OLAP

• Added three new functions:
  • IF_ERROR(expr, val) -- Evaluates the given expression expr. If it resolves to infinity or NaN, return val; otherwise return the evaluated expression
  • IF_INF(expr, val) -- Evaluates the given expression expr. If it resolves to infinity, return val; otherwise return the evaluated expression
  • IF_NAN(expr, val) -- Evaluates the given expression expr. If it resolves to NaN, return val; otherwise return the evaluated expression

Geospatial/Network Graph/Visualization

• Added new local turn restriction and penalty identifiers that can be specified at creation time or during solve, query, or match operations and will locally override the *_turn_penalty and intersection_penalty options (if set):
  • WEIGHTS_FROM_EDGE_ID
  • WEIGHTS_TO_EDGE_ID
  • RESTRICTIONS_FROM_EDGE_ID
  • RESTRICTIONS_TO_EDGE_ID
• Added support for using symbol and TRACKs simultaneously in /wms
• Added support for using expressions during classbreak raster style /wms operations
• Added support for imposing weights using WKT POINTs as a factor with the new identifier combination, WEIGHTS_WKTPOINT, WEIGHTS_FACTORSPECIFIED
• Added new options to /solve/graph for SHORTEST_PATH solutions:
  • output_edge_path -- adds a column to the solution table containing the concatenated edge IDs for each source and target pair
  • output_wkt_path -- adds a column to the solution table containing the concatenated edge WKT line segments for each source and target pair
• Improved the /solve/graph endpoint's SHORTEST_PATH solution type accuracy particularly in cases of directed social graphs
• Corrected WKT POINT decimal inaccuracies for purportedly coincident vertices while generating graph nodes
• Removed constraints for matching the number of source_nodes and destination_nodes pairs when using the SHORTEST_PATH solution type in /solve/graph

AAW

• Added support for deploying models from private model repositories (Docker Hub and Docker Registry)
• Added support for incorporating two types of feature engineering transforms into feature sets to prepare model input data: inline transforms, which are common feature preparation techniques (e.g., normalization, encoding, bucketing), and relational transforms, which leverage materialized views to support constantly changing data. Complete data lineage, including feature engineering, is now available in the audit results
• Added support for restarting models without redeploying
• Improved performance for batch inferencing and audit queries
• Removed support for the TensorFlow model type. Trained TensorFlow models can still be deployed as Blackbox models using the Blackbox SDK

UI

SQL/OLAP

• Added support for the PRODUCT() aggregate and window function. For window functions, the PRODUCT() aggregate is only supported when using default framing.
• Added support for additional optimizations that utilize dictionary encoding

Geospatial/Network Graph/Visualization

• Significantly improved speed and handling of class breaks with many classes
• Improved /match/graph Multiple Supply Demand solver operations utilizing the SAMPLE_PRIORITY match identifier
• Added two new geospatial functions:
  • ST_FORCE2D -- A geospatial function that returns the 2-dimensional version (e.g., X and Y coordinates) of all provided geometries
  • ST_FORCE3D -- A geospatial function that returns the 3-dimensional version (e.g., X, Y, and Z coordinates) of all provided geometries; if the geometry does not have a Z value already, a 0 will be registered as its Z value. An optional second argument can be provided to register a constant Z value for all provided geometries or to register a column's values as Z values for all provided geometries

APIs/Tools/Connectors

• KIO
  • Added support for setting the Kinetica password interactively, i.e. avoiding supplying the password via the command line

SQL/OLAP

• Improved the SHOW SECURITY command to display the following:
  • a roll-up of roles & permissions granted indirectly (via roles) to the requested user or role
  • when the enable_authorization configuration setting is set to false, indicating that no authorization restrictions are in effect
• Added a warning when a user attempts to GRANT permission while the enable_authorization configuration setting is set to false, indicating that the GRANT will have no effect since no authorization restrictions are in effect
• Improved performance for more operations that involve dictionary encoded columns
• Added support for list and hash partitioning TRACK tables

Infrastructure

• Added a new timeout option to every endpoint that can be used to override the global request_timeout configuration setting

Geospatial/Network Graph/Visualization

• Added support for /match/graph endpoint operations to utilize multiple CPU threads using a new max_num_threads option
• Added a new enable_truck_reuse option for the /match/graph Multiple Supply Demand solver that enables the scheduling of multiple rounds for a truck from the same originating depot
• Added a new truck_service_limit option for the /match/graph Multiple Supply Demand solver that enables users to add constraints to the optimization based on the total cost of any truck's delivery routing
• Improved transparency handling in ordered class breaks of points
• Improved rendering of tracks
• Improved handling of shapes wrapping over the dateline

APIs/Tools/Connectors

• Python API
  • Added a new GPUdbTableMonitor class that assists in handling table monitors

SQL/OLAP

• Added support for a new SHOW PROCEDURES * command that shows all SQL procedures
• Added a new RowsPerInsertion parameter to the ODBC and JDBC drivers to control bulk insertion batch sizes
• Added support for materialized views within logical views
• Removed the --password and --truststorepwd options from KiSQL. If the respective environment variables are not set, the user will be prompted for a password instead
• Enhanced performance /append/records operations now support store only data and unbound strings in additional cases
• Improved handling of multi-line SQL procedures in KiSQL
• Improved ODBC and JDBC CSV parsing to better and more consistently handle escape sequences and quoted fields
• Improved handling of inconsistent errors and ignored invalid entries within INSERT statements

Geospatial/Visualization

• Added a new priority-based identifier combination for the /match/graph Multiple Supply Demand solver that enables timely delivery scheduling to customer locations
• Added a new option, unit_unloading_cost, to the /match/graph endpoint that accounts for unloading time in the overall solution cost and optimization
• Added support for WMS rendering wrapping the world seam in all raster and class break raster modes
• Improved line track rendering

APIs/Tools/Connectors

• Kafka Connector
  • Added support for Kafka version 2.4; added support for triple properties
  • Added support for multiple topics defined with a topics.regex parameter that will sink to a single table
  • Added support for converting boolean values to integers
  • Added support for . characters in topic names
  • Added support for flattening nested messages

UI

Geospatial

• Added new optimization option to /match/graph endpoint for the MATCH_SUPPLY_DEMAND solver that allows users to control the global max trip cost allowed for any supplier (truck) assigned to a demand (depot)
• Added support for LINESTRING and MULTILINESTRING geometries in the ST_DISSOLVE geospatial function
• Improved accuracy of map matching results

SQL/OLAP

• Improved GROUP BY operations' output for EXPLAIN [VERBOSE|ANALYZE]

Geospatial

• Added two new geospatial functions:
  • ST_POLYGONIZE -- A geospatial function that creates a geometry collection containing polygons formed from given (MULTI)LINESTRING geometries
  • ST_DISSOLVEOVERLAPPING -- A geospatial aggregate function that operates similarly to ST_DISSOLVE but is optimized for overlapping polygons

SQL/OLAP

• Improved the output of the SHOW SECURITY command
• Added support for enabling column-level security functions for decimal and ulong type columns
• Added support for missing leading zeroes for the month, day, and/or hour portions of date and datetime type values
• Added support for T and Z in datetime values to enable interpretation of UTC time expressions, e.g., 2018-01-01T10:11:12Z
• Added support for the following TO_CHAR formats:
  • DAY -- full name of the day, e.g., Wednesday
  • M -- non-leading-zero month
  • C -- non-leading-zero calendar day
  • H24 -- non-leading-zero 24-hour format
  • H12 -- non-leading-zero 12-hour format
  • H -- non-leading-zero hour
• Added support for date, datetime, and time columns to the /aggregate/minmax endpoint
• Improved the output of SQL EXPLAIN plans for GROUP BY operations

Geospatial

• Added support for turn penalties and restrictions to /create/graph, /match/graph, /modify/graph, and /solve/graph
• Added new geospatial aggregate function ST_LINESTRINGFROMORDEREDPOINTS3D. This function creates 3D WKT LINESTRINGs (or POINTs)
• Added support for class break rendering track data based on a given attribute column. One value is allowed per track
• Improved the performance of the ST_DISSOLVE, ST_INTERSECTS, and ST_DIFFERENCE geospatial functions

SQL/OLAP

• Added warnings for when a user attempts to GRANT a permission the user already has or when a user attempts to REVOKE a permission they do not have
• Added support for KiSQL returning a Linux error status code when a query fails
• Table monitors are now persisted through database restart; detailed information about existing table monitors on a particular table can be found in the /show/table response

Infrastructure

• Added support for setting memory limits for rank 0 in the tiered storage configuration section of the database configuration file

Geospatial

• Added new RESTRICTIONS_EDGE_DIRECTION identifier for use in the /create/graph and /modify/graph endpoints
• Improved memory management and empty string handling for geospatial ST functions
• Added anti-aliasing support to smooth and enhance the appearance of lines and polygon boundaries in WMS imagery

APIs/Tools/Connectors

• KIO:
  • Added a --s-drop-malformed-rows flag that skips ingesting rows of a CSV file that do not fit the schema of the table into which they are being ingested. This flag is currently only valid for transfers from CSV files to Kinetica using the Spark engine

Highlights

• Added support for row-level security via the native APIs and SQL. Row-level security is achieved through restricting or granting access to certain rows of a table to a user or role based on expressions
• Added support for column-level security via the native APIs and SQL. Column-level security is achieved through restricting or granting access to certain columns in a table to a user or role
• Added data masking and obfuscation on a per-user or role basis to support row- and column-level security. Data masking will mask part of a value or the entire value; data obfuscation will obfuscate a value via hashing, with the hash being able to be grouped on for aggregation purposes

SQL/OLAP

• Added support for loading from local CSV files into a new or existing table
• Added support for CHAR, LOWER, and UPPER functions in JDBC escape sequences

Infrastructure

• Added a new /insert/records/fromfiles endpoint that enables fast ingest of local CSV files into new or existing tables

Geospatial / Network Graph

• Added a new /modify/graph endpoint to enable non-streaming graph updates and deletes
• Added support for many-to-many queries via /query/graph. These queries are also performant at scale
• Added support for label index recycling when node or edge label restrictions are enabled

APIs/Tools/Connectors

• APIs:
  • Added support for synchronicity override headers in the Python and C++ APIs
• KIO:
  • Added a version (-v / --version) flag to KIO
• Kinetica Spark Connector:
  • Added new egress processor option egress.batch_size to control the batch size
  • Added support for the egress processor to use lazy iterators for streaming data into partitions

AAW

• Added support to the Blackbox SDK for multi-row inferencing for performance gains
• Added support for container event lifecycle tracking at the container replica level
• Added support for capturing and tracking container inference and round-trip metrics and throughput
• Improved authentication support on Blackbox models
• Added expanded ML frameworks and libraries by default for the AAW RAPIDS Lab

DevOps

• Improved performance for cluster verification
• Improved performance for cluster activities on very large clusters

UI

Security

• Sanitized the input to the /gadmin/gpudbManager/getLogs endpoint of GAdmin to address a command-injection vulnerability. (as of 7.0.11.5)

SQL/OLAP

• Added support for using wildcard characters (*) in place of schema and table names for GRANT and REVOKE statements
• Added support for exposing detailed join filter plan information when using EXPLAIN ANALYZE VERBOSE with operations that do not explicitly join tables but are processed using joins to achieve the desired result
• Added support for using escape characters in LIKE statements
• Added support for the following bitwise operators:
  • & (AND)
  • | (OR)
  • ^ (exclusive OR)
  • << (left shift)
  • >> (right shift)
  • ~ (NOT)
• Added new ipv4_part function that enables extraction of any of the four parts of an IPV4 address

Geospatial / Network Graph

• Added support for batch solving several node and destination pairs using the SHORTEST_PATH solver in /solve/graph or using the MATCH_BATCH_SOLVES solver (with the appropriate Match Identifier Combinations) in /match/graph
• Added a new modify option to /create/graph that enables an existing graph to be updated with new options or new nodes, edges, weights, and/or restrictions from new/updated data in its source tables
• Increased the maximum number of cells that can be produced by the ST_*GRID functions to 100 million (from 20 million)

APIs/Tools/Connectors

• Added new assume_no_nulls to the RKinetica connector that optimizes JSON deserialization when the expected result is not expected to contain NULL values
• Added new offset and limit options to the Spark connector egress processor to enable batch/paginated reading

UI

General

• Improved performance for repeated group-by queries
• Added three new options to /insert/records:
  • return_individual_errors -- if set to true, any error that occurred when inserting an individual record will be returned
  • allow_partial_batch -- if set to true, all records that did not error during insertion will be inserted and all records that did error will be rejected and reported
  • dry_run -- if set to true, no records will be inserted and any errors will be returned
• Added support for flushing column data from memory to persist using the new evict_columns option in the /alter/system/properties endpoint

AAW

• Improved NVIDIA RAPIDS support thanks to Jupyter Notebook integration, multi-tenancy, and fast workspace persistence
• Added support for selecting either a CPU- or GPU-bound container when deploying a model in the Active Analytics Workbench (AAW)

SQL/OLAP

• Added support for SQL procedures. There is currently no support for variables or input/output parameters, but a given procedure may return a single result set or output-producing statement.

Geospatial / Network Graph

• Improved error message logging for the graph server
• Aligned the /solve/graph endpoint parameters with the other graph endpoints for consistency
• Improved the performance and reliability of the match_supply_demand solver in the /match/graph endpoint

APIs/Tools/Connectors

• Added HA failover support for servers returning a 503 error (offline) to the Java, C++, and Python APIs
• Improved performance for batch reading from PostgreSQL for ingestion into Kinetica via KIO

UI

General

• Operations that consist of dividing by 0 will now throw an error
• Enabled support for using the following endpoints with join views and materialized views:
  • /filter/byarea
  • /filter/byarea/geometry
  • /filter/bybox
  • /filter/bybox/geometry
  • /filter/bygeometry
  • /filter/bylist
  • /filter/byradius
  • /filter/byradius/geometry
  • /filter/byrange
  • /filter/byseries
  • /filter/bytable
  • /filter/byvalue
• Added support for dates containing the year 9999, sometimes used to mark a date as unknown or in the future; however, year values between 2900 and 9999 are not supported

SQL/OLAP

• Improved performance for count distinct operations involving a dictionary encoded column
• Improved CSV support for the ODBC/JDBC connector through the following features:
  • Wildcards are now supported in filenames
  • Tab delimiters are the default for TSV files; pipe delimiters are the default for PSV files
  • Comments are now supported
  • Custom null string is now supported
  • Custom quote character is now supported
  • Custom escape character is now supported
  • Option to clear table prior to ingest is now available
  • Option to skip the first n rows of the file is now available
  • Option to limit the ingest to n new rows is now available
  • Option to specify error handing mode is now available:
    • Permissive -- Kinetica will attempt to ingest all rows regardless of error
    • Skip -- Kinetica will attempt to ingest all rows that did not error during validation
    • Abort -- Kinetica will stop ingest upon encountering an error

Infrastructure

• Added support for user impersonation features when external authentication and/or HTTPD is enabled

Geospatial / Network Graph

• Improved performance of double precision WKT rendering

APIs/Tools/Connectors

• Added support for users to select a graceful failure mode when ingesting CSV files via KIO:
  • permissive -- KIO will attempt to ingest all rows regardless of error
  • dropmalformed -- KIO will drop malformed records
• Added support for unsigned long values to the C# API

UI

General

• Enabled a new system user admin role that allows management of user passwords, assignment of roles and groups, and access control of other users
• Enabled the unsigned long data type

AAW

• Added support for importing data via Bring Your Own Container (BYOC) ingests

SQL/OLAP

• Added various new security functions for identifying current user and associated roles
• Added support for executing queries as a provided user
• Implemented support for unsigned long values
• Added support for performing SQL UPDATE and DELETE operations using distributed queries (IN sub-queries)
• Wider support for is true comparisons
• Added new hints to force executing a particular SQL statement synchronously or asynchronously on other clusters in a High Availability ring:
  • KI_HINT_REPL_SYNC - when a server receives a SQL statement with this hint included, the HA process will execute the SQL statement across the cluster synchronously
  • KI_HINT_REPL_ASYNC - when a server receives a SQL statement with this hint included, the HA process will execute the SQL statement across the cluster asynchronously
• Added support for displaying the definition of materialized views using SHOW <materialized_view_name>. Only applies to materialized views created using SQL
• Various performance improvements to HAVING clauses and /aggregate/unique-related operations
• Enabled inclusion of post-mortem information to SQL EXPLAIN plans should a query fail
• Added support for multi-column LIST partition key
• Added the USER ADMIN permission in support of the new system user admin role
• Added support for replicating the right table when joining two randomly distributed tables, i.e. neither sharded nor replicated

Infrastructure

• Various performance improvements to cluster start-up
• Implemented support for the system user admin role
• Added support for tagging a UDF during UDF execution; this tag can be referenced when displaying the status of a UDF or killing a UDF

Geospatial / Network Graph

• Added support for cancelling graph server jobs
• Implemented a generic multiple supply-demand optimization solver to /match/graph
• Various performance improvements to class break operations in WMS

APIs/Tools/Connectors

• Added the system_user_admin permission in support of the new system user admin role
• Implemented support for unsigned long values in multi-head I/O operations to the native APIs
• KIO enhancements:
  • Added custom parameter support for ingesting CSV data into Kinetica, e.g., null, quote, and comment characters
  • Added support for transferring data from Kinetica to PostgreSQL via the Spark engine
  • Added support for schemas in PostgreSQL URIs

UI

AAW

• Added support for persisting training metrics during Model Training for model performance comparisons

SQL / OLAP

• Added support for grouping by WKT columns

• Added support for character delimiters and specifying columns in a file or destination table when using insert from CSV functionality, e.g.,

  INSERT INTO employee (first_name, last_name, date_of_birth)
  SELECT fname, lname, dob
  FROM FILE."/home/gpudb/data/emp_2017.csv"
  DELIMITER=',';
  

• Added support for converting various charN types to other types, e.g., char64 to date

Infrastructure

• User name and source IP address are now included with job information

Geospatial / Network Graph

• Added support for preventing conflicting graph component configurations
• Added new All Paths solver for querying all paths between two nodes
• Added two new query identifiers to retrieve edges or nodes associated with the given labels:
  • QUERY_NODE_LABEL
  • QUERY_EDGE_LABEL

APIs / Tools / Connectors

• Added all /get/records options to the RecordRetriever class for the Java, Python, and C++ APIs
• Added date, datetime, and time parsing convenience support to the Java API

UI

AAW

• Added support for multi-record (row) outputs to Blackbox functionality
• Enabled cluster multi-tenancy via custom Kubernetes namespaces

SQL / OLAP

• Added support for ASOF joins
• Added security features to logical views
• Added table monitor triggers for DELETE and UPDATE events
• Added support for temporary table names via the # prefix, e.g., #table_name
• Added support for renaming joins and views
• Various performance improvements for using Tableau with Kinetica
• Various performance improvements to predicate equi-joins (joins on the basis of equality)
• Added support for converting all numeric, date, time, datetime, and timestamp type values to charN

Geospatial / Network Graph

• Added new geospatial functions:
  • ST_LINELOCATIONPOINT
  • ST_HEXGRID
  • ST_POINTGRID
  • ST_SQUAREGRID
  • ST_TRIANGLEGRID
• Added a WMS handler for the /visualize/isochrone endpoint
• Enabled dictionary-encoded multi-property label support for improving performance when using multiple LABEL identifiers
• Implemented an origin-destination (OD) pair matching solver to /match/graph
• Various performance improvements to graph solvers and WKT class break for WMS

APIs / Tools / Connectors

• Added High Availability (HA) support to multi-head input/output for the C++ API
• Added support for the init_with_now column property to the Python API

UI

AAW

• Improved installation diagnostics
• Fixed timestamps on audit logs to reflect UTC time only

SQL / OLAP

• Added ROOT and KINETICA to the reserved words list to align with SQL standard referencing of the database catalog and root schema
• Various optimizations and improvements to window functions and distributed operations
• Expanded support for the IPV4 type
• Added support for propagating dictionary encoding to result tables
• Added support to KiSQL, JDBC, and ODBC for designating the primary processing node in a cluster
• Added support for string concatenation of numeric columns using the || operator

Geospatial / Network Graph

• Various improvements and simplifications to /query/graph
• Added a new QUERY_TARGET_NODE_LABEL query identifier for simplified querying of graph nodes

APIs / Tools / Connectors

• Various performance improvements to the Kafka connector
• Added support to JavaScript and NodeJS APIs for converting quoted Infinity, -Infinity, and NaN values to null for double and float type columns
• Added support for importing geospatial data from SQL Server via KIO

UI

AAW

• Added support for NVIDIA RAPIDS to all three deployment modes
• Improved configurations for all container assets (for inside-firewall use)
• Bundled dependencies in the RPM

Core

• Various performance improvements and bug fixes

SQL / OLAP

• Added support for RANGE and NULLS FIRST/LAST in window functions
• Added support for using NOT IN within an UPDATE/DELETE query
• Added support for managing resource groups
• Added support for CSV & Multi-head ingestion via ODBC
• Tuning improvements, including:
  • Faster STXY_* functions, geo-joins, & non-geospatial equi-joins
  • Avoiding creating projections when paging through result sets, to reduce memory

Geospatial / Network Graph

• Added support for Adaptive LOD Vector Tiles (Client side VTS)
• Added options to image/chart requests for applying log scaling for min/max request parameters
• Added ST_ConcaveHull geo function
• Added solution type support to geo functions:
  • ST_DFullyWithin
  • ST_LongestLine
  • ST_MaxDistance
  • ST_Segmentize (Euclidean)
• Graphs
  • Improved interface for applying restrictions when querying
  • Added support for constant expressions in all endpoints

APIs / Tools / Connectors

• Added HA support to multi-head ingest & egress via Java API
• Added primary cluster identification for the C++ API

DevOps

• Added support for Docker EGL rendering
• Included S3 Cold storage in backups

UI

AAW

• Added support for NVIDIA RAPIDS Model Training (Regression) and Deployment/Inference
• Added facility: BlackBox Container Wizard
• Added auto-launcher for State Manager
• Upgraded all endpoints to be capable of asynchronous operation

Core

• Added support for assigning resource groups to roles; users will inherit resource groups from roles if they have none set explicitly

SQL / OLAP

• Added support for standard SQL views
• Added support for SQL UPDATE and DELETE statements that reference other tables via JOIN
• Improved efficiency of geospatial filters
• Added JDBC support for multi-head ingest, including from CSV files
• Various optimizations and bug fixes

Geospatial / Network Graph

• Added Map Matching feature, which can map a given set of points to an underlying graph
• Added geodesic distance option to ST_buffer geospatial function

APIs / Tools / Connectors

• R ODBC connector support for row.names and bind & query parameters
• Added primary cluster identification for Java & Python APIs
• KIO
  • Added support for connecting to public S3 buckets
  • Improved transfer time between Kinetica instances via Spark
  • Added support for multi-head ingestion into Kinetica
  • Added support for ingesting into a table within a collection
  • Added ability to tune the JVM parameters for Spark memory limits
  • Improved logging

UI

AAW

• Introduced Batch Model deployments
• Added Credentials concept to handle reuseable connection profiles
• Added support on Containerized Ingests to utilize saved Credentials concept
• Introduced second process (state manager) to manage long-running processes
• Added more filtering options for audit searches
• Added Config-Driven K8s Cluster Selection
• Added capture & searchability of stack traces on individual inferences
• Faster inventory/entity dashboard operations

Core

• Added ability to bypass Disk Cache tier for persistent objects when Disk Cache & Persist tiers are similarly performant (tuning option)

KAgent

• Improved handling of upgrading httpd 2.2 configurations (when upgrading from GPUdb 6.2)
• Initial support for KAgent creation of Nvidia-enabled Kubernetes clusters

SQL / OLAP

• Improved Partitioning
  • Added List partitions
  • Added support for Date columns
  • Improved partition syntax
• Added ability to apply the current date/time as a default for date/time columns
• Added support for Alteryx
• Improved support for SQL with HA
• Various optimizations and bug fixes

Geospatial / Network Graph

• Added ability to query graphs by edge/node ID constant value
• Added Probability Rank Solver (Hidden Markov Chain)
• Added new edge/weight configurations (identifier combos)
• Added ST_partition geospatial function

APIs / Tools / Connectors

• New R ODBC connector
• Added high availability support to the basic client class for C++, Java, & Python and to the multi-head operations for Java & Python
• KIO
  • Added support for Oracle
  • Added support for Teradata
  • Added support for SybaseIQ
  • Added support for SQLServer

UI