Storage, infrastructure downplayed at Google Cloud Next '24
GenAI took center stage at Google Cloud Next, but a smattering of storage updates and an emphasis on AI workloads have some infrastructure experts curious about what's to come.
Google Cloud's bevvy of AI-related services and products headlined its annual Google Cloud Next '24 conference in Las Vegas this week, but the hyperscaler also brought a handful of new infrastructure releases.
Google Cloud storage offerings have expanded with Hyperdisk Storage Pools with Advanced capacity, a new service for its block storage. The offering replicates a popular on-premises infrastructure setup for cost and hardware efficiency. The vendor also debuted Axion -- a server Arm-based processor -- and the general availability of storage performance-focused VMs.
Storage experts said Google's conference embraced the cloud's popularity with developers, but it also kept customer infrastructure teams at arm's length.
Generative AI is a demanding and powerful enterprise workload once operational, according to Keith Townsend, an analyst at Futurum Group. Google's own Vertex AI efforts such as Creative Agent provide additional benefits to customers, but those same buyers still have common enterprise apps and demands to maintain.
The air has been taken out of the room for cloud infrastructure.
Keith TownsendAnalyst, Futurum Group
"The air has been taken out of the room for cloud infrastructure," Townsend said. "AI is still a workload. It's probably the most demanding workload we see, but SAP still needs to run."
Speeding up storage
Hyperdisk Storage Pools with Advanced capacity, generally available now, builds on the SSD block storage offering Google released last year to enable the pooling of storage in Google Cloud across multiple workloads.
Pooling storage resources enables larger storage reserves with thin provisioning for VMs and applications without needing to purchase additional storage, Townsend said. Historically, most hyperscalers require customers to purchase underlying storage for a VM or app instead of pooling resources the way admins have done on premises.
"This has been a longtime complaint from the enterprise," Townsend said.
This capability exists in storage vendor cloud offerings such as NetApp's OnTap, but isn't sold natively in AWS or Microsoft Azure, he said.
Google also launched the general availability of the Z3 machine series for Compute Engine, storage-optimized VMs running on Google Cloud's physical hardware. These VMs prioritize SSD cloud workloads demanding high IOPS for uses such as scale-out databases and data warehouses, according to the vendor.
Google Cloud's new Axion processors -- the platform's first Arm-based chips -- as well as the general availability of AI accelerator TPU v5p and a handful of new cloud compute instances and bare-metal offerings rounded out its infrastructure updates.
The new Axion processor for Google Cloud's hardware, held by Google Cloud CEO Thomas Kurian during his keynote at Google Cloud Next 2024 in Las Vegas, was among the few new infrastructure releases highlighted at the show.
If not for the current hype cycle surrounding AI, Townsend said, the Hyperdisk Storage Pools with Advanced capacity would have made headlines for marking a significant strategy change by Google.
Rather than attempting to lock users in to its cloud platform, Google is offering multi-cloud services and capabilities to attract new customers, he said. And having an option to pool storage rather than requiring customers to buy more is a positive trend.
"As cloud growth has slowed, they have to give up their sacred design cows to win these enterprise workloads," Townsend said. "Google has embraced open standards compared to the other cloud guys."
Creating services and capabilities for multi-cloud or hybrid cloud customers is one of the major ways Google Cloud can differentiate itself from competitors, said Simon Robinson, an analyst at TechTarget's Enterprise Strategy Group.
Google's support for open source projects such as Kubernetes container orchestration has won over buyers before, he said. Enabling infrastructure teams to work with the platform across environments, rather than exclusively within it, could win additional customers.
"I don't think they're as embedded in the core [enterprise] infrastructure compared with AWS and Microsoft Azure," Robinson said. "Developers are where it's at. They have huge influence on the direction, pace and extension of a cloud platform."
All AIs on me
Still, Google Cloud Next's primary focus remained firmly on AI and related offerings. But developments surrounding the technology are laying the foundation for additional capabilities and a stronger partner ecosystem including for storage, said Sid Nag, an analyst at Gartner.
"[Google] finds connective tissue across different services," Nag said. "[Like] what are the resources needed to support that workload [and] how do you address storage for it to be an asset."
Google Cloud's Vertex AI offerings might be among the more technologically advanced out there, but haven't gained as much enterprise traction as Microsoft Copilot and the Amazon Bedrock managed service, said Brent Ellis, an analyst at Forrester Research.
"They haven't broken through in the way OpenAI [has with] Microsoft and, to a lesser degree, AWS has," Ellis said. "[Google] doesn't get to live on their own little island. They get adoption by partnering with other technology organizations."
Generative AI workloads will eventually become commoditized and made more available to enterprise IT teams, according to Steve McDowell, founder and lead analyst at NAND Research. Generative AI creation will require hybrid cloud implementations, he said, demanding faster storage through an R&D investment by platform holders.
"Cloud storage has never been known for or marketed for performance, but AI changes that equation," McDowell said. "The cloud guys will have to stand up and deliver."
Tim McCarthy is a news writer for TechTarget Editorial covering cloud and data storage.
