January 2024 Summaries

Graph databases are transforming real-time fraud detection in financial institutions and e-commerce companies. As the threat landscape evolves, so must our tools to identify and prevent this new generation of fraud. PayPal's adoption of graph technology for their fraud program has inspired others to follow suit. By using graph database operations, organizations can analyze behaviors, understand subtle relationships, and mitigate fraud in rapidly changing market conditions. Graph databases like Aerospike Graph offer real-time query services that help detect anomalies in activities and behaviors associated with fraudsters. As the need for advanced fraud detection grows, incorporating graph databases into data pipelines is becoming increasingly important to keep fraudsters at bay.

Graph databases are revolutionizing fraud detection in real-time transactions by providing the ability to analyze behaviors and understand subtle relationships between data points, allowing for more effective identification and mitigation of new and evolving types of fraud. PayPal's bespoke graph solution, which predates Aerospike Graph, demonstrated the power of graph technology in detecting and preventing online fraud, with a focus on using temporal graphs to signify high fraud risk and devising a graph query service to retrieve results quickly enough to be used in real-time AI models. The use of graph databases can add context to transactions by coordinating questions about relationships between data points, such as customer behavior and transaction history, making it easier to determine the legitimacy of transactions and keep fraudsters at bay.

The new Spring Data Aerospike release 4.6.0 introduces several features and enhancements for customizing set names and using batch write operations, including the ability to provide a set name per operation, supporting batch-write operations such as deleteAllById(), saveAll(), insertAll(), updateAll(), and deleteByIds(). It also supports versioned documents with generation counts retrieved from the database records. The new features allow developers to easily specify which set they want to use, avoid potential conflicts with data from other sets, and group multiple operations into one unit for a single network trip to each database node. Additionally, it provides examples of utilizing batch-write operations in Spring Data Aerospike, including saving multiple non-versioned documents, inserting multiple new records, updating multiple existing records, and deleting multiple records in one request.

The CAP theorem states that no distributed system can be designed for both consistency and availability, but algorithms have been developed to improve system availability during common failure situations while supporting strong consistency. Strong consistency ensures all writes to a single record are applied in a specific order without re-ordering or skipping. Aerospike supports linearizable and sequential read models for strong consistency in single-record transactions. Linearizable reads guarantee the most updated version of data, while sequential reads may return stale data but are faster. Eventual consistency and strong eventual consistency apply to systems with asynchronous replication, where data updates are not synchronous between nodes. Aerospike provides strong consistency with high performance and availability during common failure scenarios.

The CAP theorem states that a distributed database system cannot be designed for both consistency and availability, but algorithms can significantly improve system availability while supporting strong consistency. Strong consistency guarantees that all writes to a single record will be applied in a specific order, ensuring that data remains consistent across all nodes in the cluster at any given point in time. There are different levels of consistency, with "strong consistency" being the highest level, which includes linearizable and sequential consistency models. Linearizable consistency is stricter, slower, but always returns the latest data, while sequential consistency is faster but may return stale data occasionally. Strong eventual consistency guarantees that every copy of a data item will end up with the same result without any data loss, but allows for stale reads. Eventual consistency allows for stale data and lost writes. Aerospike has developed strong consistency algorithms that provide high performance and 100% data availability during common failure scenarios, while preserving consistency and availability with minimal impact on performance.

In an era of signal loss, capturing attention is crucial for effective advertising and marketing. The value of focused attention cannot be overstated, as it forms the basis of data collection and subsequent value exchange between users and platforms. Marketers need to find moments that matter by integrating their products or services into customers' experiences in a compelling manner. A modern lifecycle marketing framework should leverage full potential of data, conduct media mix modeling, optimize funnel across customer lifecycle, use resources wisely, harness word-of-mouth marketing, deliver tailored content, and understand the opportunity cost of different media channels. Future-proofing AdTech and MarTech requires innovation, strategic thinking, and a deep understanding of customer behavior.

In an era of signal loss, where companies are struggling to capture attention and deliver tailored experiences, marketers and advertisers must adapt by leveraging data, optimizing their media mix, and harnessing the power of word-of-mouth marketing to maximize reach and value exchange. By integrating creative and copy into customer experiences, creating custom algorithms for funnel optimization, and using resources wisely, businesses can future-proof their AdTech and MarTech with modern lifecycle marketing frameworks.

The article outlines the process of establishing connectivity between two Aerospike clusters using the Cross Data Center Replication (XDR) feature to seamlessly transfer data from a source to a destination cluster, particularly within a Kubernetes environment. It addresses the challenge of network visibility by placing a proxy server in front of a private Kubernetes destination cluster. The setup involves installing the Kubernetes Operator, configuring both source and destination clusters, and setting up an XDR proxy to manage data replication efficiently while resolving potential update conflicts. The guide includes practical steps for implementing this configuration, including the use of Kubernetes commands and YAML files to create and manage the required resources, ensuring that data is successfully replicated across clusters. Additionally, it demonstrates the use of Aerospike's benchmarking tool to test data flow and highlights the scalability of XDR proxies, emphasizing the system's reliability and resilience in mitigating cluster unavailability risks through asynchronous data replication.

At Aerospike Summit 2023, a discussion on the power of graph databases took place between Aerospike Product Manager Ishaan Biswas and two fellow graph experts, George Anadiotis and Marko Rodriguez. They delved into topics such as understanding graph databases, the architecture of Aerospike Graph, analyzing the current graph market, adopting a graph data model, evaluating graph databases, scaling graph databases, and more. The key takeaways from their discussion include the importance of leveraging connections in graph databases, the differences between SQL and graph data models, criteria for evaluating graph databases, and how Aerospike Graph handles challenges like the supernode problem to provide efficient performance at large scale.

Aerospike Graph was discussed at Aerospike Summit 2023 by Ishaan Biswas, George Anadiotis, and Marko Rodriguez, who talked about the power of graph databases, their current market growth, and how to evaluate them. The three experts defined graph databases as a solution for handling large amounts of connected data and enabling developers to leverage the graph data model to address complex computing challenges. They also discussed various use cases such as identity graphs in AdTech, customer 360, and real-time fraud prevention. Key takeaways include adopting a paradigm shift from SQL to a graph data model, evaluating commercial solutions based on performance, scalability, cost, analytics, developer-friendliness, operational agility, and community support. Additionally, the experts highlighted challenges such as scaling graph databases, managing bidirectional pointers between vertices on different machines, and establishing in-sync cleanup mechanisms for smoother system performance.

Charles Schwab modernized its database architecture for real-time intraday operations by adopting Aerospike as its intraday system of record. The company faced challenges with their existing database architecture, including limited scalability and availability issues. To address these issues, they established new architectural principles such as a single operational data store, no local caching, maximum code reuse, and CLR/JVM focus. By adopting Aerospike, Charles Schwab achieved 99.99% availability during intraday hours and can handle about 150 TB in each zone with 75 billion daily transactions.

Charles Schwab modernized its database architecture to support real-time intraday operations, adopting Aerospike as its system of record, to achieve high availability and scalability for its $7.7 trillion assets. The company's goal was to build better, faster, and cheaper solutions, addressing challenges such as core trading applications residing on the mainframe, limited scalability, and constraints on latency and throughput. By implementing a single operational data store, no local caching, maximum code reuse, CLR/JVM focus, and adopting Aerospike, Charles Schwab achieved data consistency, high performance, and 99.99% availability during intraday hours. The transformation has enabled the company to handle massive transactions, with 75 billion daily transactions and a throughput of 1.1 million reads per second.

The global energy consumption of computers is rapidly increasing due to the widespread adoption of Artificial Intelligence (AI) and Machine Learning (ML). In fact, the carbon footprint of cloud computing now surpasses that of the airline industry. Training several AI models can release up to 626,000 pounds of carbon dioxide, which is nearly five times the emissions of an average American car over its entire lifecycle. Technologists have a significant responsibility for ensuring the sustainability of the software and systems they employ. Sustainable AI is not only beneficial for the environment but also good for business, as companies that reduce carbon emissions can earn more profit than those that don't. Efficient database technology like Aerospike can significantly shrink server footprints, reducing latency and boosting throughput while lowering total cost of ownership (TCO). IT decision-makers must factor in the environmental impact when selecting a technology stack to help create a sustainable future.

The global energy consumption of computers is increasing, with experts predicting a significant spike by 2030, driven primarily by the growing adoption of Artificial Intelligence (AI) and Machine Learning (ML). The carbon footprint of cloud computing now surpasses that of the airline industry, raising concerns about its environmental impact. Researchers have found that training AI models can unleash substantial carbon emissions, highlighting the need for sustainable AI solutions. Technologists have a significant role to play in driving sustainability, and there are methods to achieve this without sacrificing performance, such as efficient coding practices and optimized data centers. Sustainable AI is not only good for business but also essential for addressing global warming, with companies committing to net-zero emission targets and earning significant profits by doing so.