We know customers have varying workloads composed of many services and components with different technical requirements, and a “one size fits all” approach prevents customers from scaling while trying to balance performance and cost.
With that in mind, last month at Google Cloud Next ‘24 we announced C4 and N4, two new machine series in our general-purpose VM portfolio. C4 and N4 are the first machine series from a leading cloud provider to leverage the latest 5th generation Intel Xeon processors (code-named Emerald Rapids), and are powered by Google’s Titanium, a system of purpose-built microcontrollers and tiered scale-out offloads.
With their combination of industry-leading performance and cost-efficiency, C4 and N4 provide solutions for all of your diverse general-purpose workload needs. C4 tackles demanding workloads that benefit from industry-leading performance, while N4’s price-performance gains and flexible configurations, like custom shapes and extended memory, let you choose various combinations of compute and memory to optimize costs and reduce resource waste for the rest of your workloads. Whether your workload needs unwavering performance for mission-critical tasks or prioritizes adaptability and cost optimization, C4 and N4 are designed to meet all your general-purpose computing needs. The N4 machine series is generally available today and the C4 machine series is available now in preview on Compute Engine and Google Kubernetes Engine (GKE).
N4: Price-performance gains and flexible shapes
The N4 machine series is built from the ground up for price-performance gains and cost optimization through an efficient architecture of streamlined features, shapes, and next-generation Dynamic Resource Management. N4 enables you to lower your TCO for various workloads compared to previous generation N-Family instances, with up to 18% better price-performance than N2 instances1 and up to 70% better price-performance than N1 instances.2 When compared to N2 across key workloads, N4 delivers up to 39% better price-performance for MySQL workloads3 and up to 75% better price-performance for Java applications4.
Relative to comparable offerings from other leading cloud providers, N4 delivers up to 90% better price-performance on MySQL workloads5 and up to 75% better price-performance on Java applications6.
With N4, you also get maximum flexibility to support evolving workload needs with custom shapes, which give you granular control of resources. Custom shapes allow you to only pay for what you use so you don’t have to overbuy any vCPU or memory, adding a cost optimization lever that is currently only available on Google Cloud. Custom shapes also make it easy to optimize for unique workload patterns with nonstandard resource needs and move from on-prem to cloud or from another cloud provider to Google seamlessly, without any reconfiguration. Compared to similar offerings from other leading cloud providers, N4 also offers 4x more predefined shapes at much larger sizes up to 80 vCPUs and 640GB of DDR5 memory to serve a wider variety of workloads. N4’s predefined shapes are available in high-cpu (2GB/vCPU), standard (4GB/vCPU), and high-mem (8GB/vCPU) configurations.
“At Bluecore, we really care about cost savings. N4 is great for our dev/test and benchmarking workloads, and it met our expectations in every way. We observed around a 66% performance uplift from N1, and we will definitely adopt N4.” – Michael Hurwitz, Software Architect, Bluecore
N4 also offers up to 4x the standard networking bandwidth (50 Gbps standard) compared to competing cloud providers and up to 160K IOPS with Hyperdisk Balanced7. N4 is suited for most general-purpose workloads that need a balance of performance and cost efficiency, such as medium-traffic web and application servers, dev/test environments, virtual desktops, microservices, business intelligence applications, batch processing, data analytics, storage and archive, and internal applications. To learn more, visit N4 overview.
Next-generation Dynamic Resource Management
Google’s next-generation Dynamic Resource Management (DRM) on N4 and Titanium enhances existing optimization mechanisms to improve price-performance, lower costs, and serve an expanded set of applications. Dynamic Resource Management is a Google-specific technology that underpins our global infrastructure, servicing workloads such as Search, Ads, and YouTube. Dynamic Resource Management on N4 brings proven reliability, efficiency, and performance at scale, and the latest generation of Dynamic Resource Management leverages Titanium to accurately predict and reliably deliver the required performance for your workloads. To learn more, visit Next-generation dynamic resource management overview.
C4: Industry-leading performance and advanced features
Equipped with the latest-generation compute and memory, Titanium network and storage offloads, and advanced performance and maintenance features, C4 is engineered to support your most demanding workloads. C4 delivers up to 20% better price-performance relative to comparable offerings from other leading cloud providers8, and up to 25% better price-performance across key workloads compared to the previous-generation C3 VM9.
Relative to comparable offerings from leading cloud providers, C4 delivers up to 65% better price-performance for Redis workloads10 and up to 90% better price-performance for MySQL workloads11.
With Titanium, C4 also offers up to 80% better CPU responsiveness compared to previous generations for real-time workloads, including high-frequency trading and gaming, resulting in faster trades and a smoother gaming experience12. The high-performance of C4 is a great fit for demanding databases and caches, network appliances, high-traffic web and application servers, online gaming, analytics, media streaming, as well as real-time CPU-based inference, with Intel AMX.
C4 also offloads virtual storage processing to the Titanium adapter to improve infrastructure performance, lifecycle management, reliability, and security. With Titanium, C4 offers scalable, high-performance I/O up to 500k IOPS and 10 GB/s throughput on Hyperdisk and up to 200 Gbps of networking bandwidth13. C4 also offers enhanced maintenance controls for greater predictability and control for planned maintenance, with less frequency and disruption. C4 instances scale up to 192 vCPUs and 1.5TB of DDR5 memory. It is available in high-cpu, standard, and high-mem configurations. C4 is available now in Preview on Compute Engine and Google Kubernetes Engine (GKE) customers at goo.gle/C4Preview.
C4 and N4: Better together
We know that selecting the right VM is essential for your success. With C4 and N4, you get tailored solutions for all your general-purpose workloads, so you can lower the total cost of running your business without compromising on performance or workload-specific requirements.
“We are excited about Google Cloud’s new C4 and N4 machines. The new C4 VMs show 35% performance boost for real-time ML threat detection and N4 VMs provide 20% higher throughput for Dynamic Analysis of Malware samples at a lower cost vs. the previous generation.” – Xu Zou, VP of Product, CDSS, Palo Alto Networks
Empower your business with VMs that grow with you. When you use C4 and N4 together, you get cost-effective solutions that prioritize performance and reliability, as well as the flexibility to mix and match machine instances based on your different workloads.
“We’re excited to partner with Google Cloud on the N4 and C4 machine series powered by the latest 5th Gen Intel Xeon processors. These instances deliver up to 2.5x higher AI inference performance than previous generations, using Intel AMX. Users will have a choice between flexibility, cost optimization and exceptional performance for their unique workload requirements.” – Suzi Jewett, General Manager – Intel Xeon Products, Intel Corporation
N4 is generally available now in us-east1 (South Carolina), us-east4 (Virginia), us-central1 (Iowa), europe-west1 (Belgium), europe-west4 (Netherlands), and asia-southeast1 (Singapore). C4 is available now in Preview on Compute Engine and Google Kubernetes Engine (GKE) customers at goo.gle/C4Preview.
1, 2 Google Cloud Internal Data, as of March 21, based on planned listed prices.
3, 7, 9, 13 Google Cloud Internal Data, as of March 2024, based on planned list prices.
4, 5, 6, 8, 10, 11 Google Cloud Internal Data, as of May 2024, based on planned list prices.
12 Google Cloud Internal Data, as of March 19, 2024, based on Google internal benchmark to measure vCPU interruption time, i.e. ‘jitter’