Skip to main content

Product Advantages

Relyt is built entirely on public clouds since day one, inheriting all the attributes of public clouds. This includes limitless storage and compute resources, making it well equipped to handle the exponential growth of data and user scale in the age of big data and generative AI. By taking advantage of economies of scale, Relyt effectively minimizes the cost per unit of resources. Additionally, as an AI-ready data cloud, Relyt offers all capabilities as services, eliminating the need for you to manage and maintain hardware and software. This allows you to concentrate on business innovation and growth. But that's not all. Relyt, committed to continuous technological innovation and exploration, has achieved impressive results in its architecture, engine, security, and ecosystem.


Innovative architecture​

In Relyt, we separate storage from compute, completely decoupling data and metadata from compute. This architecture enables Relyt to pool both storage and compute resources effectively, with compute completely serverless. Relyt not only offers top-notch scalability, cost-performance, and reliability but also excels in resource utilization and cost-effective hybrid workloads. As a result, Relyt achieves a total cost of ownership (TCO) that is 90% lower than its competitors.


Cost-effectiveness​

Pay as you go, without resource wastes​

Relyt adopts an architecture that separates storage from compute, allowing storage and compute to be scaled independently. The underlying storage uses cost-effective cloud data stores, ensuring not only affordability but also scalability. At the upper layer, compute resources are provided in the form of DPS clusters, which can be deployed, resized, suspended/resumed, and released as needed. DPS clusters offer a linear performance increase, ensuring outstanding observability. In actual use, you only need to deploy DPS clusters according to your specific workload requirements, without the need to concern yourself with storage space.


Additionally, you can manually suspend inactive DPS clusters or set up an auto-suspend policy. No charges are applied to suspended DPS clusters. These suspended clusters are automatically resumed within milliseconds once a query is issued to them. This approach not only ensures cost efficiency but also maintains optimal query performance.

Cloud FinOps capabilities​

Relyt is a data-centric cloud service provider. It helps organizations build advanced data infrastructure and is continuously working on minimizing its TCO to make sure that every penny that a user pays counts. Relyt is designed with the cloud FinOps capability since the first day it born, to provide organizations with better visibility of monitoring information about tasks, systems, and costs. This design contributes a lot to optimizing resource utilization and reducing costs. In addition, the two-level cloud account hierarchy used by Relyt makes financial management and control, such as budget management, account splitting, and account consolidation, more convenient. These features offered by Relyt help flatten paths for organizations to data-driven sustainable business development and innovation.


Excellent performance​

Operator-level performance optimization​

Relyt is a cloud-native platform that optimizes the computation density and utilization of cloud resources at the granularity of operators (the smallest units for computation in a compute engine). This operator-level cloud-native approach is pioneered by Relyt, which schedules and optimizes resources at the operator level. Additionally, Relyt continues to carry out engineering optimization in the implementation of operators to improve computing performance, mainly in the following aspects: vectorized query execution, algorithms based on data characteristics, data access based on the operating system memory layout.

Till now, we have finished some performance tests in terms of operator performance statistics by using TPC-H, Relyt operators have obvious advantages over Trino and Spark, and many Relyt operators provide the same or even higher performance as their counterparts provided by ClickHouse. In addition to SQL operators, Relyt continuously works on optimizing operators of other types such as data encryption and decryption and compression and decompression, by taking the computing power characteristics into consideration. The adoption of these techniques allows Relyt to provide advanced and comprehensive data processing functionality at an extremely low unit price. The TCO of Relyt is not even a tenth that of other mainstream data platforms.


Workload isolation and support for parallel running​

DPS clusters are designed to ensure complete workload isolation. This allows for the deployment of multiple DPS clusters, each handling a different workload in a one-to-one mode. This approach not only eliminates resource contention but also guarantees that each workload is assigned adequate, stable computing resources. This enhances data processing efficiency and stability.

Furthermore, Relyt's capability to pool computing resources empowers you to create as many DPS clusters as you need without an upper cap. Regardless of increasing parallelism levels, Relyt ensures responses within milliseconds. This provides a solid base for processing large amounts of data and complex computing tasks.

One data​

Leveraging its SHARED-DATA architecture, Relyt effectively resolves the issue of data consistency in distributed systems. It ensures reliable service operation and consistent system performance, even during data migration and redistribution caused by system expansion. Relyt maintains a single global copy of the data, guaranteeing that all business data originates from a singular, truthful source.

This capability enhances data quality, boosts decision-making efficiency, and promotes data sharing and collaboration. Furthermore, by maintaining just one copy of data, Relyt significantly simplifies data management complexity and reduces management overheads.

99.9% query success rate​


Elastic computing​

DPS clusters can be deployed, configured, scaled, and suspended/resumed on demand, providing Relyt with excellent compute elasticity for handling large-scale data. Each DPS cluster processes its workload independently, ensuring that other data processing tasks remain unaffected even if one cluster fails, thereby guaranteeing high reliability for Relyt.

Additionally, the "DW service units + DPS clusters" architecture isolates data and compute resources. This allows Relyt to dynamically adjust computing resources based on workload requirements, optimizing system performance and improving efficiency. This elastic computing capability ensures efficient operation even under high concurrency, further enhancing system stability and reliability.

Adaptive Query Scaling​

Relyt has introduced a feature called Adaptive Query Scaling (AQS) for DPS clusters. This feature identifies and routes large queries to the AQS shared storage pool during runtime, ensuring the performance and stability of DPS clusters enabled with AQS while executing large queries successfully.

AQS requires you to set a limit on the compute resources it can allocate for processing large queries from specific DPS clusters. This helps you manage their budgets in response to unexpected business requirements. Consequently, there's no need to adjust the specifications of the DPS clusters to handle workload peaks effectively, ensuring a query success rate of not lower than 99.9%.

Advanced engine​


Vectorized engine​

Vector DPS, developed by Relyt, is a powerful engine for embedding vectors, tailored for the management, search, and analysis of unstructured data. It aims to help users leverage the capabilities of generative AI, Retrieval Augmented Generation (RAG), AI agents, and beyond. With Vector DPS, users can instantly build AI applications for hybrid search and analysis for structured data and unstructured data such as video, audio, images, and text. Key features offered by Vector DPS include:

  • High throughput and concurrency: Vector DPS utilizes HNSW + online PQ indexing algorithms to offer high throughput (millions of TPS writes, real-time indexing), high concurrency (tens of thousands of QPS, real-time queries), and real-time incremental training.

  • Real-time high-dimensional vectors: With support for vector data up to 5000 dimensions, Vector DPS enables real-time operations such as adding, deleting, modifying, and querying vector data.

  • Hybrid search: Vector DPS allows for flexible combination of search conditions to meet specific needs, ensuring over 99% accuracy.

  • Cost-effectiveness and simplicity: Vector DPS not only offers parameter auto-tuning for optimized performance but also ensures cost efficiency with a pay-per-use billing model.

CBO​

Relyt uses a cost-based optimizer (CBO) to select the most efficient and cost-effective query plans. The CBO estimates the cost of each potential query plan based on the characteristics of the data and the distribution of computing resources. It then chooses the option with the lowest cost for the execution sequence and procedure, ensuring minimal time and space complexity. This makes the CBO an excellent choice for handling complex queries and large-scale data scenarios, outperforming rule-based optimizers.

High-throughput storage​

To meet high-throughput storage demands, Relyt has made bold and effective design choices. For example, its sort key-based pruning feature significantly reduces I/O operations by up to 99%. Additionally, Relyt uses advanced hybrid row-column storage technology, optimizing both single-column and multi-column queries to minimize unnecessary data I/O. This technology supports various compression algorithms and encoding methods, adaptively selecting the most suitable ones based on data characteristics, ensuring high I/O efficiency.

Robust security​


Relyt has adopted a range of advanced security technologies and measures to keep your accounts, data, and network connections secure. With multiple security compliance certifications under its belt, Relyt is committed to providing you with the highest standards of protection.

Highly reliable object storage:

Relyt persistently stores data in cloud object storage, ensuring 12 nines of data security.

Bucket-level isolation:

Each user's data is securely stored in separate buckets, ensuring strong isolation and protection from unauthorized access. This approach also simplifies backup and recovery, guaranteeing data availability and security even during disruptions.

Table-level encryption:

Used to encrypt data in transit and at rest.

Role-Based Access Control (RBAC):

Role-based access control (RBAC), used to restrict access based on the permissions assigned to specific roles. Relyt leverages the account/subaccount two-level hierarchy offered by RBAC to enhance account security and ease of use. In addition, Relyt develops a mechanism called "administration and use separation". This feature does not allow administrators to create accounts for other users. Instead, administrators invite users to create their own accounts. In this way, the password of an account is known only by the user who creates the account and thus advancing your privacy security.


Access token management:

Used to advance authentication isolation and management for different applications on the same DW service unit.


PostgreSQL-compatible and easy-to-use​

Compatible with PostgreSQL protocol and ecosystem​

Relyt is compatible with the PostgreSQL protocol and ecosystem. The PostgreSQL ecosystem is nurtured by a powerful community and provides comprehensive SQL functionality and rich extensions. Compared with MySQL, PostgreSQL is better suited for data analytics and complex queries. What's more, workloads that are compatible with PostgreSQL are not vendor lock-in, and thus possessing a wide range of community freedom. Relyt inherits all these advantages of PostgreSQL and continues developing its own secure, open, and collaboration-oriented data platform and data cloud that has regions distributed worldwide, while contributing to the PostgreSQL community.

Support for ACID transactions​

As a hybrid workload processing platform, Relyt supports ACID transactions, ensuring data consistency, durability, and isolation between transactions. This significantly improves Relyt's reliability, stability, and efficiency.

Out-of-the-box global data cloud infrastructure​

Relyt provides organizations with out-of-the-box global data cloud infrastructure of which the data warehouse service units are distributed across regions deployed in multiple clouds. These data warehouse service units can be interconnected on demand. Such a highly integrated architecture helps Relyt to leverage its seamless data exchange and migration features, allowing organizations to access required data anytime, anywhere. In addition, Relyt supports single sign-on (SSO). This enables organizations to integrate their systems with existing lightweight directory access protocol (LDAP) or third-party authentication services. Based on the data.cloud domain, an organization can customize its exclusive data cloud domain name (organization_name.data.cloud) and portal. All these operations are completed with a few clicks, and then the organization can start using its own global data cloud infrastructure.