Hazelcast Sets Industry Standard for Consistency and Performance for Data-Intensive, Mission-Critical AI Workloads

The latest release of Hazelcast Platform features an Advanced CP Subsystem and increased performance at a low total cost of ownership

Palo Alto, Calif., April 17, 2024 – Hazelcast, Inc., the company that enables instant action on all data, announced the latest release of its unified real-time data platform, a core component of the AI application and data infrastructure for Global 2000 enterprises. The new features in Hazelcast Platform 5.4 ensure data consistency, resilience, and high performance. 

Nearly one-third of executives identified data-related challenges among their top three concerns¹ impacting enterprise AI initiatives. These data challenges, often referred to as the “Day 2 problem,” particularly impact performance, resilience, and scale in production systems. These challenges emerge after the initial focus on model design and expose the underlying AI application and data infrastructure limitations.

Hazelcast Platform 5.4 directly addresses these challenges with several new features, including:

• Advanced CP Subsystem for strong consistency that retains a performance advantage over other comparable systems;
• Thread-per-core (TPC) architecture that extends Hazelcast Platform’s industry-leading performance;
• Access to larger data volumes with Tiered Storage.

“Industry leaders, especially within financial services, are positioning AI at the center of their application modernization strategies, yet they tend to focus on the prototyping and deployment phases of AI applications. Unfortunately, the underlying data infrastructure is often overlooked and leads to underperforming applications,” said Adrian Soars, chief technology officer at Hazelcast. “The latest features of Hazelcast Platform empower developers and data scientists to integrate performance, resilience, and data consistency into their AI applications from the start. Even better, Hazelcast Platform provides low-latency storage and compute in a single platform, thus significantly lowering the total cost of ownership, especially post-deployment.” 

Eliminating the gap between consistency and performance

Data consistency is at the core of Hazelcast Platform 5.4, which powers some of the biggest payment processing workloads in the world. The platform ensures all nodes in a cluster deliver a single view of data for applications, eliminating the potential for corrupted or failed transactions. The key challenge with a strongly consistent, distributed architecture is overcoming the performance impact of data replication across the cluster, which is necessary for enabling resilience. Hazelcast Platform ensures strong data consistency while still retaining a high-performance advantage.

Internal benchmarks on a minimal cluster show that on strongly consistent data, Hazelcast Platform delivered over 10x the read and write performance requirements of our leading payment processing customers.

Hazelcast solves the challenges of data consistency, performance, and scale with the inclusion of several features:

• Advanced CP Subsystem delivers an accurate, up-to-date view of data for all client requests for key/value data structures (“maps”) in a distributed system. Combined with TPC, this makes for an efficient and predictable approach to ensuring strong data consistency and unencumbered performance at scale. 
• Thread-per-core architecture fully utilizes all cores in a modern CPU. While Hazelcast already delivers industry-leading performance, TPC can increase Hazelcast Platform throughput by an additional 30% on large workloads with a high number of clients. Organizations with improved hardware optimization can process large data loads in sub-milliseconds by fully exploiting the computation power resident in their network.
• Tiered Storage is a companion to Hazelcast’s unique distributed fast data store architecture. Tiered storage enables customers to scale storage processing for intensive AI/ML workloads in a single, integrated environment. This flexibility eliminates the need to offload to disk-based systems, a complex route that hits performance.

According to Gartner², “Organizations are increasing their use of machine learning inferencing, business rule processing, and other analytics as nodes within the flow of ESP applications to support sophisticated and powerful decision intelligence. We expect this trend to accelerate in concert with the overall expansion of real-time analytics and AI in mainstream companies.”

Delivering Value to Customers

Many Global 2000 brands trust Hazelcast, with two-thirds of the world’s transactions passing through its Hazelcast-powered applications. With its market-leading technology and the rapid adoption of real-time and AI applications, companies using the Hazelcast Platform can simplify and accelerate application modernization – at a lower total cost of ownership – and achieve results such as: 

• Processing more than $1 billion (USD) per minute for a global financial institution;
• Boosting loan conversion rates by 400% through personalized offers in the moment of need;
• Avoiding losses of $100 million a year with improved fraud detection; and
• Reducing operational costs by 33% by optimizing the use of equipment and resources in real time.

Customers often use Hazelcast Platform to support AI/ML deployments for real-time applications, including payments, fraud detection, trade monitoring, in-game betting, IoT edge computing, real-time offers, and Customer 360. 

Resources 

• Announcing Hazelcast Platform 5.4 Release [Blog]
• Hazelcast Training: Strong Data Consistency [Training]
• Embracing the Demands of an AI-Centric Future with Hazelcast Platform 5.4 [Webinar]

1. Beena Ammanath et al., Thriving in the era of pervasive AI: Deloitte’s state of AI in the enterprise, 4th edition, Deloitte, 2021. 
2. Gartner, Inc. “Market Guide for Event Stream Processing” by W. Roy Schulte, Pieter den Hamer, Ehtisham Zaidi. May 15, 2023.

About Hazelcast

The world’s leading companies trust Hazelcast and its unified real-time data platform for mission-critical and data-intensive workloads that require high performance, resilience, and scale. Customers such as JPMorgan Chase, Volvo, New York Life, Target, and Domino’s rely on the Hazelcast Platform for application modernization initiatives and AI/ML deployments.

Hazelcast enables companies across every industry to drive new revenue, mitigate risk, and improve operational efficiencies – at a low total cost of ownership. Top banks and credit card companies rely on Hazelcast for payments, processing billions of dollars per second of domestic and cross-border payments, avoiding $100 million per year in fraud loss. Manufacturing, shipping and services companies rely on Hazelcast to optimize the utilization of equipment and resources in real-time to save 33% in operational costs. 

The Hazelcast unified real-time data platform uniquely combines a distributed compute engine and fast data store into a single runtime to process real-time data, enrich it with historical context and enable real-time ML inference.

To join our community of CXOs, architects, and developers, visit www.hazelcast.com.