Why Intel for Compute-Intensive Workloads on Google Cloud?

What’s the Challenge?

  • Massive data growth, new frontiers in artificial intelligence (AI) and machine learning (ML), high-performance compute (HPC) workloads: all of these and more are driving an insatiable demand for more compute power in the cloud.
  • Some organizations need compute-optimized instances with near real-time performance to meet their business requirements. Other organizations have a wide range of compute-intensive workloads—and while they may not need real-time performance, they do need instances that are designed from the ground up to process data efficiently.
  • Customers need help choosing the Google Cloud machine type that best enables them to balance performance and cost considerations in order to be efficient and cloud agile.

 

What’s the Solution?

Google Cloud C2 and N2 Instances Enabled by Intel® Architecture

Google Cloud and Intel have been co-innovating to create cloud computing environments that can meet a wide range of compute needs. Google Cloud instances enabled by Intel® architecture include the general-purpose N2 instance, as well as the compute-optimized C2 instance. The Intel® hardware and software behind these machine types can increase compute performance and operational efficiency, and can help transform Google Cloud customers’ businesses into digital powerhouses. Using the combination of Google Cloud and Intel® technology, your end customers can process more data in less time, make faster decisions and create better experiences for their customers.

C2: Near Real-Time Performance for the Most Demanding Workloads

With the highest per-core performance on Google Cloud1, C2 instances based on 2nd generation Intel® Xeon® Scalable processors are ideal for multi-player gaming, HPC, video encoding and real-time web serving. The instances can provide up to 3.8 GHz all-core frequency with Intel® Turbo Boost Technology enabled.

With Intel® Deep Learning Boost, Google Cloud customers can take advantage of built-in AI acceleration. These technologies help customers’ AI applications to deliver results faster.

Google Cloud customers can accelerate their digital transformation with transparent access to the underlying Intel hardware, enabling them to fine-tune HPC workloads.

N2: A Flexible, Powerful Choice for Most Workloads

General-purpose workloads include databases, web applications, mobile gaming and dev/test environments.

You can maximize operational efficiency by helping Google Cloud customers configure an exact fit for every workload. N2 instances can have from two to 80 vCPUs with 640 GB of total memory (depending on the exact N2 machine type).

Compared to an N1 instance, Google Cloud customers can expect the following performance improvements with an Intel® hardware-powered N2 instance:

Accelerate Digital Transformation with Artificial Intelligence and Machine Learning Solutions Optimized for Intel® Technology

Workloads such as image and video analysis, recommender systems, natural language processing (NLP) and other analytics applications benefit immensely from Intel technology available on N2 and C2 instances.

Google Cloud deep-learning virtual machines (VMs) and containers are optimized for 2nd generation Intel Xeon Scalable processors and configured to support common AI frameworks like TensorFlow and PyTorch out of the box. Customers do not have to waste time customizing or configuring these frameworks for optimal performance, enabling them to focus on the AI itself.

Optimized Software from Intel Can Help Google Cloud Customers Increase Operational Efficiency

  • Software optimizations from Intel can help Google Cloud customers increase performance and achieve better operational efficiency.
  • Intel® Advanced Vector Extensions 512 (Intel® AVX-512), an acceleration feature of Intel® Xeon® processors, is available on both C2 and N2 instances.
  • Optimized application performance on Intel Xeon Scalable processors gives Google Cloud customers a performance advantage when they choose Intel technology-powered Google Cloud solutions.
  • Descartes Labs used Intel compilers to optimize its global food supply analytics application for Intel AVX-512, running on Google Cloud. The optimizations lowered costs by 40 percent for certain use cases8
  • Sony Imageworks takes advantage of Intel AVX-512 to optimize the firm’s OpenShadingLanguage (OSL) to speed up rendering, enabling artists to create content faster9.

Google Cloud and Intel Alliance: Drive Continual Innovation

Google and Intel’s close and long-standing relationship can help you provide better enterprise support, accelerate customer success and drive strong customer retention. Get in touch with your Intel account representative to access:

  • Take advantage of the N2/C2 trial program through Intel-funded GCP credits, details located at: go/intel-compute-trial.
  • Technical support from Intel’s engineers.
  • Opportunities to work closely with system integrators (SIs) that are intimately familiar with Intel technology. Intel can help you connect with an SI that is the best fit for your customer.
  • Assistance with proof of concepts (PoCs) to help align customers’ needs with the right combination of Intel technology.
  • Marketing support for your customers’ PoCs to help develop compelling proof points and content that will help fuel sales and marketing efforts.

If you are interested in working with Intel to accelerate your sales cycle, contact your account team.

Take advantage of the N2 / C2 trial program through Intel-funded GCP credits, details located at: go/intel-compute-trial.

Notices and Disclaimers

Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.

Performance results are based on testing as of dates shown in configurations and may not reflect all publicly available updates. See backup for configuration details. No product or component can be absolutely secure.

Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy.

Intel technologies may require enabled hardware, software or service activation.

Your costs and results may vary.

Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.

Product and Performance Information

4Comparing N1-Highmem-32 instance vs. N2-Highmem-32 instance for OLTP HammerDB 3.2. Testing by Intel on 8/11/20. Cloud: GCP; instance: N1 Highmem 32 (Intel); region: GCP US Central 1 Iowa; CPU: Intel Skylake; memory capacity/instance: 208 GB; storage: Persistent Disk SSD 500 GB; network BW: 32 Gbps; storage BW: not specified; OS: Ubuntu 20.04; kernel: 5.4.0-1021-gcp; workload and version: MySQL 8.0.20; compiler: NA; libraries: NA; other SW: OLTP HammerDB 3.2 Testing by Intel on 8/10/20. Cloud: GCP; instance: N2 Highmem 32 (Intel); region: GCP US Central 1 Iowa; CPU: Intel Cascade Lake; memory capacity/instance: 256 GB; storage: Persistent Disk SSD 500 GB; network BW: 32 Gbps; storage BW: not specified; OS: Ubuntu 20.04; kernel: 5.4.0-1021-gcp; workload and version: MySQL 8.0.20; compiler: NA; libraries: NA; other SW: OLTP HammerDB 3.2
5Comparing N1-Standard-32 instance vs. N2-Standard-32 instance for Kmeans. Testing by Intel on 4/9/20. Cloud: GCP; instance: n1-standard-32 (Intel); region: GCP US Central 1a Iowa; CPU: Intel Broadwell; memory capacity: 128 GB; storage: Pd-ssd, 6x300 GB; network BW: not specified; storage BW: not specified; OS: Ubuntu 19.10; kernel: 5.3.0-1011-gcp; workload and version: Hibench v7.0; compiler: NA; libraries: NA Testing by Intel on 4/9/20. Cloud: GCP; instance: n2-standard-32 (Intel); region: GCP US Central 1a Iowa; CPU: Intel Cascade Lake; memory capacity: 128 GB; storage: Pd-ssd, 6x300 GB; network BW: not specified; storage BW: not specified; OS: Ubuntu 19.10; kernel: 5.3.0-1011-gcp; workload and version: Hibench v7.0; compiler: NA; libraries: NA
6"The new general-purpose machine types (N2) offer greater than 20% price-performance improvement compared with first generation N1 machines." Hanan Yousef, Product Manager of Compute Engine, Google Cloud. Learn more: https://cloud.google.com/blog/products/compute/expanding-virtual-machine-types-to-drive-performance-and-efficiency
9 https://cloud.google.com/customers/sony-pictures-imageworks