Oracle’s new Exadata architecture targets efficiency, cost

by Pelican Press
10 views 8 minutes read

Oracle’s new Exadata architecture targets efficiency, cost

Oracle on Thursday launched Exadata Exascale, a new architecture for the cloud designed to provide improved performance for all Oracle Database workloads at significantly reduced costs.

Exadata was first introduced well over a decade ago and has since been the infrastructure for Oracle Database, a relational database for enterprise data storage. It has since been updated many times to evolve as data management and analytics have evolved.

Exadata Exascale similarly represents an evolution to meet modern data management and analytics needs, which increasingly includes both traditional AI as well as generative AI.

Toward that end Exadata Exascale provides improved performance for the vector processing workloads that are frequently used to discover relevant data for AI models and applications and feed the retrieval-augmented generation pipelines used to update and train such models and applications.

Exadata Exascale is immediately available on Oracle Cloud Infrastructure (OCI) as part of the Exadata Database Service on Exascale Infrastructure and Oracle Database 23ai. Availability in Exadata Cloud@Customer, OCI Dedicated Region and in multi-cloud environments is still to come.

Given the improved performance and potential for substantial cost savings due to its cloud-native architecture, removal of certain infrastructure requirements and pay-per-use cost structure, Exadata Exascale is a major release for Oracle, according to Holger Mueller, analyst at Constellation Research.

“It is [significant],” he said. “It allows Oracle customers to scale up and down their database resources as they need it — in the cloud — for the first time. Exadata now benefits from the overall OCI innovations [enabling users] to bring compute power to the data as needed and get charged only for that.”

Based in Austin, Texas, Oracle is a tech giant that along with competitors such as AWS, Google and Microsoft provides a large swath of data management and analytics tools. Among its database offerings are Oracle Database and Autonomous Database.

Autonomous Database is a fully managed service that removes an enterprise’s administrative responsibility while Oracle Database can be deployed as either a fully managed service or under an enterprise’s own control.

In May, Oracle released Database 23ai, adding improved vector search capabilities to help customers augment data with generative AI within their database rather than require them to export their data to AI platforms to apply generative AI to their data. More recently, in June, the tech giant launched HeatWave GenAI, an update to its HeatWave database aimed at enabling enterprises to combine proprietary data with generative AI tools.

New capabilities

Exadata Exascale represents a new generation of Oracle’s Exadata Cloud Infrastructure, including a new architecture that better handles hyperscale workloads — often to develop AI models and applications — that lead to lower costs when charged on a usage basis.

Those savings can be as much as 95% compared to current costs for some customers, according to Kothanda Umamageswaran, Oracle’s senior vice president of Exadata and scale-out technologies.

Previously, a dedicated infrastructure was required to use Exadata, costing customers $10,800 per month. With Exadata Exascale, an entry-level environment is $357 per month. In addition, the price of running the Exadata Database Service on Exascale infrastructure with a bring-your-own license is $1.77 per hour, down from $15.81 per hour.

Beyond reduced pricing, performance improvements achieved through features such as AI Smart Scan and intelligent storage cloud also contribute to lower costs for customers.

AI Smart Scan is an AI capability that offloads high volume data and compute-intensive vector search workloads to the Exascale intelligent storage cloud where workloads can run up to 30 times faster than in Oracle Database, according to the tech giant. That offload, in turn, allows customers to run potentially thousands of concurrent vector searches.

Intelligent storage cloud, meanwhile, automatically distributes databases across all available storage servers by using AI Smart Scan to make use of thousands of CPU cores that increase the speed of database queries.

As a result of the improved efficiency and cost savings resulting from Exadata Exascale’s architecture, Stephen Catanzano, an analyst at TechTarget’s Enterprise Strategy Group, also said Oracle’s new release is notable.

“What makes it a big deal is its ability to combine Exadata’s well-established performance and reliability with the elasticity and cost-efficiency of the cloud,” he said. “This integration allows for hyper-elastic scaling, pay-per-use economics and substantial infrastructure cost reductions.”

Beyond reducing costs for existing customers, one potential benefit of Exadata Exascale is that it has the potential to attract small- and medium-sized organizations that previously couldn’t afford Oracle Database, Catanzano continued.

“[This makes] high-performance database capabilities accessible to organizations of all sizes,” he said.

Umamageswaran similarly said that one benefit of the new architecture and its cost efficiency is that it makes Oracle Database accessible to a broader audience.

More than three-quarters of Fortune’s Global 100 companies use Exadata, including more than half who use it in the cloud, according to Umamageswaran. Cost efficiency opens its use to a more diverse customer base.

“Exadata Exascale supports organizations of all sizes,” Umamageswaran said. “Exascale’s low cost opens it up to Fortune Global 2000-type companies, entry-level enterprises and SMBs to use for smaller workloads.”

In addition to AI Smart Scan, intelligent storage cloud and pay-per-use pricing, Exadata Exascale also includes the following:

  • Intelligent online transaction processing that enables communication between servers to foster scaling across Exascale Virtual Machine clusters.
  • Intelligent analytics to unload data-intensive SQL queries to the Exascale intelligent storage cloud for scaling.
  • Data cloning using the Exascale intelligent storage cloud to make development, test or recovery data immediately available with the same performance and scale as the source data.

Combined, Exadata Exascale’s capabilities now differentiate Oracle Database from relational database competitors such as Microsoft SQL Server and Google Cloud Spanner, according to Catanzano.

“While other cloud database providers offer scalable and efficient solutions, Oracle’s approach of integrating Exadata’s advanced database capabilities with cloud elasticity is distinctive,” he said. “The specific features like AI Smart Scan, intelligent storage cloud with RDMA [remote direct memory access] capabilities, and database-aware intelligent clones set it apart from competitors.”

That differentiation could be short-lived, however, given that Oracle’s peers are similarly working to improve the efficiency and cost-effectiveness of their databases, Catanzano added.

Mueller likewise said that Exadata Exascale represents differentiation for Oracle. However, , he added that he’s not seeing much innovation from Oracle’s relational database competitors meaning that Oracle’s advantage could last for at least a little while.

“The competition is not doing anything,” Mueller said. “SQLServer [from Microsoft] has not shown any innovation. IBM has done nothing for DB2. And on the cloud database front, HeatWave has out-innovated all the cloud database startups.”

While lowering costs to potentially attract new customers was one motivation for developing Exadata Exascale, Oracle recognized that increasing development of AI capabilities — fueled by the surging interest in generative AI –necessitated a new database architecture, according to Umamageswaran.

Vector processing, in particular, is a high-volume, compute-intensive workload that has become a critical capability over the past year.

“New cloud paradigms have emerged since Exadata was initially developed, and requirements for scalable GenAI workloads — such as vector processing — have risen in importance,” Umamageswaran said. “Accordingly, we wanted to enhance the capabilities of Exadata’s scale-out storage, VMWare clusters and low-latency networking to address a broader range of customers.”

Plans

Going forward, Exadata Exascale will be the underlying architecture for all Oracle Database services in OCI, according to Umamageswaran.

Regarding further product development, Mueller said Oracle should look to make more of its data stack autonomous by using machine learning to automate routine tasks without needing humans  to be part of the process. The tech giant provides Autonomous Database, but other data management and analytics tools require human involvement.

“My long-term critique [about Oracle] has been that it is great to have an autonomous database, but it is only one part of the stack,” Mueller said. “The whole stack needs to be autonomous, and Oracle is slow at enabling that.”

Catanzano, meanwhile, suggested that Oracle could do more to integrate with other vendors’ platforms, particularly those that provide AI and machine learning tools.

While Oracle offers a broad array of data management, analytics and AI tools, many enterprises prefer to mix and match tools from various vendors rather than use tools from only one provider. That enables them to both develop a data stack that best suits their needs as well as avoid the potential pitfalls of vendor lock-in.

“It would be beneficial to see Oracle enhance integrations with other AI and machine learning platforms to streamline workflows for data scientists and developers,” Catanzano said. “Additionally, further advancements in automation, security and ease of migration from other systems would help organizations transition more seamlessly to Exadata Exascale.”

Eric Avidon is a senior news writer for TechTarget Editorial and a journalist with more than 25 years of experience. He covers analytics and data management.

 



Source link

#Oracles #Exadata #architecture #targets #efficiency #cost

You may also like