Google Cloud is looking to address the very specific requirements of its customers with a series of updates to its workload-optimized compute infrastructure portfolio.

The new virtual machine offerings, announced at Google Cloud Next ‘24, span a wide range of use cases, including general-purpose workloads, power-efficient workloads and specialized applications such as SAP deployments and high-performance databases.

In a blog post, Google Cloud’s director of product management Salil Suri explained that the rapid pace of innovation today has resulted in customers having some very specific needs when it comes to compute. Fortunately for them, Google Cloud has recognized these demands and adopted the concept of “workload optimization” as one of the main guiding principles of its infrastructure platform.

“We engineer golden paths from silicon to the customer workload, using a combination of purpose-built infrastructure, prescriptive architectures, and an open ecosystem to deliver workload-optimized infrastructure,” he said.

Fourth-Generation C4 VMs for high-performance workloads

The latest general-purpose processors in Google Cloud’s VM portfolio include the new C4 VMs, which are powered by 5th Generation Intel Xeon central processing units and enabled by Titanium, a system of purpose-built, custom silicon and multiple tiers of scale-out offloads that provides increased performance by freeing up CPU resources.

Suri said the C4 VMs are fine-tuned to provide high performance that’s paired with a “controlled maintenance experience” to support more flexibility for mission-critical workloads. He claims that the new family delivers 19% superior price-performance compared to alternative offerings from other cloud providers, and 25% superior price-performance compared to the company’s previous-generation C3 VMs. In addition, it delivers 80% better CPU responsiveness for real-time workloads, such as high-frequency trading and video gaming, leading to faster trades and a smoother gameplay experience.

These capabilities mean the C4 VMs are an ideal fit for multiple workloads, including database, caches, network appliances, high-traffic web servers, online gaming, data analytics, media streaming and real-time inference, Suri said.

Of course, the new C4 VMs are not just more powerful, but also more functional, with new features designed to minimize the impact on workloads when customers are performing infrastructure maintenance. “With C4, the vast majority of updates to the host operating system and instance are conducted using a combination of hitless upgrades and live migration, with near-zero customer impact,” Suri said. “And for workloads or times when you need longer notifications and more direct control over maintenance events, C4 includes enterprise-grade controls to help you achieve these goals.”

The C4 VM instances are available in preview in Google Compute Engine and Google Kubernetes Engine. There are 24 size options, scaling up to 192 virtual CPUs and 1.5 terabytes of DDR5 memory, with support for Hyperdisk up to 500,000 input/output operations and Tier 1 networking at up to 200 gigabytes per second. Customers can choose from high-cpu (2GB/vCPU), standard (4GB/vCPU) and high-mem (8GB/vCPU) configurations.

New N4 VMs for flexible, cost-optimized workloads

As for the N4 VMs, they’re designed to support more flexible workloads with greater cost optimization. They do this through a more efficient architecture of streamlined features, shapes and enhanced Dynamic Resource Management capabilities, Suri said.

The N4 family, generally available today, delivers some impressive gains, with Suri claiming an 18% improvement in price-performance for most workloads, rising to a 39% improvement for key workloads such as MySQL databases. What’s more, they also support custom machine types, a key differentiator for Google Cloud that allows customers to optimize costs by eliminating the need to pay for fixed-sized instances.

Suri said the N4 VMs can cater to a wide range of workloads that don’t require peak processing power at all times, such as developer and test environments, virtual desktops, microservices, business intelligence applications, code repositories, data analytics, storage and archiving, medium-traffic web servers and more.

Once more, customers can choose from highcpu (2GB/vCPU), standard (4GB/vCPU), and high-mem (8GB/vCPU) configurations, with up to 80 vCPUs and 640 GB of DDR5 memory. The NV VMs also support up to 160,000 Hyperdisk IOPs and 50 Gbps instance networking.

C3, X4, Z3 and Arm Axion instances

In addition to the C4 and N4 VMs for general-purpose workloads, Google Cloud announced a range of more specialized VMs, including the native bare-metal C3 machine types, which are an extension of the C3 Machine Family that launched last year. These, Suri said, are aimed at customers with specialized workloads such as hypervisors, those that aren’t supported in virtual environments, and those with specialist performance or licensing restrictions.

The C3 bare-metal instances provide direct access to the CPUs and memory resources of the underlying server, as well as Intel’s integrated CPU features, such as the Data Streaming Accelerator and In-Memory Analytics Accelerator, Suri said.

As for the X4 memory-optimized bare metal instance family, these are meant to cater to compute and block storage workloads, packaging up to 1,920 cVPUs and 32TB of DDR5 memory. They’re available in preview now in the U.S., Europe, Middle East and Africa and Asia-Pacifc regions with three configurations. They’re backed by Google Compute Engine’s Memory Optimized 99.95% Single Instance service-level agreement, providing the reliability required for business-critical workloads such as SAP databases.

“SAP Enterprise Cloud Services has been evaluating and exploring the integration of X4 and C4 Metal systems for RISE with SAP, and has been impressed with its reliability, performance, and ease of management,” said Lalit Patil, chief technology officer for RISE with SAP.

Elsewhere, Google Cloud is expanding its VM portfolio with the new Z3 instances, which are said to be its first storage-optimized VMs. They’re designed to provide more consistent and reliable performance, with predictable maintenance and reduced total cost of ownership for storage dense workloads that require solid-state drives. According to Suri, they support the highest IOPS for storage among all of the instances offered by the top cloud providers, meaning they are supposedly superior to equivalent instances offered by rivals such as Amazon Web Services Inc. and Microsoft Corp.

Finally, Google Cloud is introducing a new Arm-based CPU called the Google Axion Processor, which delivers industry-leading performance and efficiency by combining Google’s silicon expertise with Arm’s most advanced core designs.

The result is a new family of Arm-based instances that deliver 30% better performance than previous Arm VMs, and up to 50% superior performance and 60% energy-efficiency compared to existing x86-based VMs, Suri said.

The Axion instances will become available later this year, but Google Cloud has already begun using them to support its own internal workloads, including services such as BigTable, Spanner, BigQuery and Blobstore, as well as the YouTube Ads platform.

Image: BrianPenny/Pixabay