Cassandra is the “finest f*cking database for Gen AI,” says DataStax CEO 

Chet Kapoor, CEO of DataStax, an organization that gives a cloud database primarily based on open supply Apache Cassandra, boasted at a convention in Silicon Valley yesterday that Cassandra is the “finest f*cking database for Gen AI.”

Kapoor’s remarks, made whereas talking on stage on the Linux Basis occasion, Dev.AI, with 700 attendees, come at a time when there’s a full-on race by new startups and incumbents to seize the mantle of management within the quick rising space of Gen AI. It’s also time when many enterprise manufacturers utilizing the expertise are deciding which expertise suppliers they’ll use.

Whereas a lot consideration has gone to the competitors between large-language mannequin suppliers, like OpenAI, Anthropic, Google (Gemini) and Meta (Llama), one other extremely aggressive space is the databases that end-user corporations will use to retailer and retrieve information used for LLM functions.

Throughout his keynote, Kapoor gave a number of the reason why DataStax Cassandra database is doing properly in opposition to others. Cassandra is already one of the crucial dependable operational databases broadly utilized by enterprise corporations, it boasts some early buyer instances of corporations really deploying generative AI at scale, and its expertise prowess in key areas related to generative AI proceed to provide it a leg up in opposition to key rivals like MongoDB and Pinecone, Kapoor mentioned.

It’s additionally price noting that DataStax is contemplating going public quickly, and Kapoor has an curiosity in making some noise. In June of final 12 months, DataStax raised $115 million on a $1.6 billion valuation. The corporate has not launched any monetary information, however Kapoor acknowledged in an interview that DataStax is on the shortlist of corporations that banks would need to take public subsequent 12 months. 

Listed here are the explanations behind Kapoor’s bullishness:

Cause 1: Cassandra is already one of the crucial broadly used and dependable operational databases 

Kapoor’s feedback additionally come at a time when the big cloud corporations like Microsoft and Amazon have been asserting that their cloud choices, which embrace integrations with their very own databases, are one of the best positioned to carry out generative AI duties. They’ve been encouraging customers to consolidate on their platforms, and aggressively eradicating obstacles which have prevented customers up to now from doing that, together with advanced extract, rework and cargo (ETL) work that has saved information siloed.

Nonetheless, these cloud corporations have provided too many particular person databases to customers over the previous decade, within the effort to supply the specialised options for every use case a buyer has, Kapoor mentioned. “There’s one to go to the lavatory within the morning,” Kapoor joked, “after which one for the afternoon, and one for the night.” However generative AI has caught these cloud corporations unexpectedly: Enterprise CIOs now need their information built-in right into a single database to permit Gen AI apps to question the information extra simply and effectively, Kapoor mentioned. 

And right here Cassandra has a bonus as a result of it is among the extra widespread “operational” databases, whereas most of Microsoft and Amazon’s databases have been centered on analytical workloads, primarily for enterprise intelligence functions. Whereas they can be utilized for operational workloads for generative AI functions, it might turn out to be very costly, as a result of they aren’t optimized for that. 

DataStax has spent a number of time centered on worth for efficiency, for instance, Kapoor and Chief Product officer Ed Anuff defined in a follow-up interview with VentureBeat after Kapoor’s remarks. Consequently, Cassandra is the most well-liked for Fortune 500 corporations, which ship information at scale. Cassandra boasts 90 % of these corporations as prospects, Anuff mentioned. For instance, Netflix makes use of it for his or her film metadata, Fedex makes use of it for monitoring packages, Apple makes use of it for his or her  iTunes, iMessages and iCloud app information, and retailers like House Depot use it for his or her websites.

As these huge corporations construct new AI apps, they’re snug with the monitor file they’ve with Cassandra, and so are prone to proceed to consolidate round that, Anuff mentioned. Furthermore, Microsoft and Amazon have realized they should provide option to prospects. Amazon, for instance, affords a aggressive operational database, DynamoDB, but it surely additionally affords customers the power to simply use Cassandra inside its cloud constellation. In that means, Cassandra additionally affords prospects a technique to keep away from lock-in with a selected cloud vendor, Anuff mentioned.

Cause 2: DataStax has prospects that really “deploy” generative AI 

Kapoor cited 9 corporations which have deployed generative AI on DataStax’s Astra DB database, the cloud database-as-a-service primarily based on Cassandra. Whereas many enterprise corporations are experimenting like loopy with generative AI, few have moved to precise manufacturing at scale, out of considerations round issues like security and reliability. Certainly, the strain within the business has risen markedly: The potential for generative AI could also be enormous, however most distributors of the expertise agree they’re ready for patrons to start out spending actual income, which can come subsequent 12 months when corporations transfer to manufacturing in a severe means.

The DataStax prospects with deployed LLMs embrace:

• Physics Wallah, an Indian on-line training platform, that serves 6 million customers with a multi-modal (textual content, photos, and audio) giant language model-driven bot. The corporate moved to deployment in 55 days, Kapoor mentioned.
• Skypoint, a Portland-based Gen AI healthcare supplier for seniors and care suppliers, that makes use of an LLM to offer customized remedies and interactions. Kapoor mentioned Astra DB helps release 10+ hours every week for medical doctors to deal with affected person care. 
• Others embrace Hey You, Reel Star, Arre, Hornet, Restworld, Sourcetable, and Concide.

Kapoor mentioned these corporations are a part of a fast-moving class of small and medium sized enterprise (SMB) which might be capable of transfer extra rapidly, whereas enterprise corporations are slowed by having to observe extra rules and avoiding issues of safety in generative AI that embrace its tendency to hallucinate. 

Cause 3: DataStax’s Cassandra tech prowess beats others on key LLM benchmarks

Kapoor mentioned DataStax’s Astra’s vector search choices carry out higher and are extra related than these of opponents. Vector search is a key requirement for generative AI databases, since that’s how an AI utility interprets a consumer’s question in pure language to seek for textual content or different information in an organization’s database related to that question. DataStax benchmarked its JVector vector search expertise in opposition to a main vector database competitor, Pinecone, and located the JVector outcomes are 16 % extra related than Pinecone’s. Kapoor mentioned that’s an enormous distinction, contemplating how vital it’s to get the fitting reply. A 3rd social gathering vendor can be releasing the total efficiency benchmarking report in a number of days, Kapoor mentioned, however he confirmed slide of among the outcomes (under). The benchmarking additionally confirmed Datastax having superior throughput, or potential to course of extra transaction requests per unit of time, than each Pinecone and MongoDB.

He mentioned Astra DB is the one database that may make vectorized information obtainable with zero latency, together with indexing, ingestion and querying.

Kapoor: “This Gen AI wave goes to be sooner than any frickin’ factor we’ve seen”

Kapoor mentioned that GenAI adoption will occur a lot sooner than earlier expertise revolutions, because it builds on vital foundations which might be already there, resembling internet, cell and cloud applied sciences.

He mentioned that the “actual enjoyable” will begin subsequent 12 months with extra transformative and revenue-oriented use instances, together with individuals utilizing LLMs as “brokers.” These brokers enable LLMs to do extra than simply reply questions and make suggestions, he mentioned, as a result of they’ll orchestrate extra advanced duties. Materials income from generative AI deployments will present up within the second quarter of subsequent 12 months, with “extra sizable” numbers hitting by the top of the 12 months, when use instances in areas like retail and journey ramp up, Anuff mentioned.

Whereas Kapor and Anuff had been wanting to level out some great benefits of Cassandra, they conceded that the broader database sector goes to see a elevate from generative AI. The vector database searches that gen AI apps carry out use 8 instances the storage and about 10 instances the compute than different database workloads use, Anuff mentioned. “That’s a part of the explanation why you see the entire cloud suppliers, and the entire database suppliers wanting that enterprise,” he mentioned. “If AI functions turn out to be a giant deal, they’ll be the first progress driver for each non-public and public database corporations for simply the following 5 years.”