An in-memory column-oriented DBMS for transactional-analytical data processing in real time
Scroll
What is transactional-analytical data processing?
Traditionally, transactional and analytical systems are separated from each other. As a rule, transactional information is copied into a data mart or a storage, and then analytical queries are performed on it. This is necessary so that analytical queries don’t slow down operating systems.
In-memory technologies allow real-time transaction processing (OLTP) and analytical query processing (OLAP) within one system.
Such processing is called HTAP (Hybrid transactional/analytical processing).
The Tarantool Column Store column DBMS implements HTAP processing. This makes it possible to perform analytic queries instantly on the latest data, as well as reduce the cost of transferring and copying information.
What problems does Tarantool Column Store solve?
Real-time analytics
Real-time report generation
Acceleration of analytical queries and calculations
Consolidation of analytical data
Storing and managing data and metadata (Feature Store) for machine learning tasks
Tarantool Column Store use cases
Speeding up and increasing the accuracy of anti-fraud systems
Problem: Financial organizations require quick access to large volumes of constantly updated data to instantly detect suspicious transactions.
Solution: Tarantool Column Store (TCS) allows you to detect suspicious transactions with almost zero latency. TCS is built into an anti-fraud system as a high-speed analytical storage, makes it possible to store over 5 TB of data with 1000+ attributes (columns) and carry out analytical calculations in real-time.
Improving the performance of a loan issuance system
Problem: The bank needs to quickly generate a number of loan offers for a potential customer based on the data supplied through a loan application form. These offers must be as attractive to the customer as possible and, at the same time, be low-risk for the bank.
Solution: Tarantool Column Store, in response to an application, carries out real-time analytical processing of a matrix that contains more than 100,000 options for loan offers and additional services. As a result, the loan issuance system instantly assesses the risk of non-repayment of borrowed funds for each loan offer and generates a set of the most suitable and low-risk options.
Generation of financial reports in real time
Problem: Large businesses need to quickly make decisions on financial and management matters. To support this, there are various reports for business management at different levels. Reports should be generated quickly and contain up-to-date and reliable data.
Solution: Tarantool Column Store consolidates data on financial and economic activities (assets and liabilities, purchasing activities, product sales, cash flows, etc.) in a single repository and runs analytical queries in real time for quick reporting. Denormalized data is stored in tables with 400+ columns and a capacity of more than 6 TB.
Scroll
Advantages of Tarantool Column Store
High performance due to multithreading
Real-time data analysis using in-memory technologies
Possibility of massively parallel processing of large volumes of data
Horizontal scaling through sharding
Development by a Russian vendor: support and adaptation of the product to your needs
Flexible configuration of data schemas (tables, columns, and indexes)
Secondary indexes with depth adjustment
CRUD operations
SQL support, including aggregate and statistical functions
Single/mass data recording
Data persistence
Auto-eviction of data when RAM is full
Data replication
Sharding
Multi-threaded reading
Data compression
API for reading/writing data from business applications in Java, Python, or Go (for now, HTTP API is supported)
Administration and Security Options
ACID support
Role-based model (RBAC) for assigning access rights to objects
Security audit log
Operational Capabilities
Support of Russian operating systems: Astra Linux, REDOS, and Viola
Export monitoring metrics to Prometheus and Grafana
Automation of installation and cluster launch
Cluster management
Deployment with Kubernetes
Tarantool Column Store architecture
Tarantool Column Store stores and processes data in the form of columns, integrating with business applications via JDBC/ODBC, HTTP, SQL, and Apache Arrow Flight. Scaling tools available: clustering, replication, and sharding.
Still have questions?
Tell us all about your project tasks, and we will create a Tarantool-based solution tailored to your requirements