The Future of Cloud: A View from Google Cloud Next ‘19

Thousands of IT professionals will gather to learn about the latest cloud technologies at Google Cloud Next '19 this week. At last year’s event, I wrote about some of the upcoming projects our teams were developing, and it’s terrific to see that work presented to the industry today.

At Intel’s Data-Centric Summit in August 2018, Intel’s Navin Shenoy handed Google Cloud’s Vice President of Engineering, Bart Sano, the first production unit of Intel® Optane™ DC Persistent Memory.  In October, Google was the first large cloud service provider to offer previews of this transformative technology. Since then, the two companies have made tremendous strides in optimizing hardware and software to deliver infrastructure innovation to customers. Google Cloud recently announced three new virtual machine (VM) instances based on the 2nd Gen Intel Xeon Scalable Processors. We’ll be discussing these instances in-depth at this year’s Google Cloud Next event.

Cloud Instance Optimization

Google Cloud’s new memory-optimized M2 instances run on new, 2nd generation Intel® Xeon® Scalable processors and are designed for the most memory-intensive applications, such as SAP HANA. M2 instances are available for early testing with up to 12 terabytes of memory, with plans for future instances up to 18 terabytes of memory.  Google Cloud’s Memory Optimized VMs also support configurations featuring Intel Optane DC persistent memory for better restart times and TCO improvements.  At Google Cloud Next’19, we’ll demonstrate how GCP instances with Intel Optane DC persistent memory offers a more affordable alternative to DRAM, and enable new capabilities like 12X faster startup time in Google Cloud.1

Google Cloud’s new compute-optimized C2 instances are also available on the latest 2nd generation Intel Xeon Scalable processors, and are designed for best performance per instance. With turbo frequencies up to 3.8GHz, these new instances can power the most demanding and performance-hungry applications.  In fact, Google Cloud has stated these new C2 instances provide over 40% more performance compared to current GCP VMs. We plan to demonstrate extreme online gaming workloads on C2 instance platforms at the event.

Finally, we will also showcase 2nd generation Intel Xeon Scalable processors coming to GCP’s general-purpose instances with highly-competitive  price performance, as well as flexible configurations for high utilization and workload customization.

Artificial Intelligence Advancement

Following extensive collaboration, it’s a great pleasure to announce that Intel solutions are now available on the Google AI Hub. All the instances I’ve mentioned above use 2nd generation Intel Xeon Scalable processors with Intel® Deep Learning Boost (Intel® DL Boost), our integrated AI inference acceleration technology.  When combined with optimized software, these new processors enable up ta 14X higher AI inference performance than the first-generation Xeon Scalable processors at their launch in July 2017.2

A New Solution for Hybrid Cloud

This week, we also announced a strategic collaboration with Google to create a new hardware and software solution for on/off-prem hybrid clouds.  This new reference architecture will bring together Google’s Kubernetes Engine stack and Intel’s latest Xeon Scalable platform and enable customers to enjoy a common operating environment across their public and private cloud infrastructure.  This should make workloads seamlessly portable, and reduce workload validation effort and cost by consolidating into a single environment.  Google and Intel will work together to bring this optimized architecture to market as an Intel® Select Solution later in 2019, available from leading server manufacturers and system integrators.

The collaboration between Intel and Google Cloud will put amazing new capabilities at your fingertips, and give you an express-lane to take advantage of 2nd generation Xeon Scalable processors, Intel Optane DC Persistent Memory, and the latest in hybrid cloud solutions.

For the latest news about Intel and GCP, visit our collaboration page, or check out a recent Chip Chat podcast with Google’s Paul Nash.


1 All measurements based on comparing Google Cloud n1-ultramem 160 vCPUs, 3844 GB of memory, 4 Intel® Xeon® Processor E7-8880 v4 (Broadwell) CPUs and DDR4 memory versus Google Cloud new instances: 2nd Generation of Intel Xeon Scalable Processors Platinum, 96vCPUs, ~5.5TB AEP and 1.4 DRAM done in Q42018

2 14x inference throughput improvement vs baseline:  Tested by Intel as of 2/20/2019. 2 socket Intel® Xeon® Platinum 8280 Processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode: 0x200004d), Ubuntu 18.04.1 LTS, kernel  4.15.0-45-generic, SSD 1x sda INTEL SSDSC2BA80 SSD 745.2GB, nvme1n1 INTEL SSDPE2KX040T7 SSD 3.7TB, Deep Learning Framework: Intel® Optimization for Caffe version: 1.1.3 (commit hash: 7010334f159da247db3fe3a9d96a3116ca06b09a) , ICC version 18.0.1, MKL DNN version: v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a,  model: https://github.com/intel/caffe/blob/master/models/intel_optimized_models/int8/resnet50_int8_full_conv.prototxt, BS=64, DummyData, 4 instance/2 socket, Datatype: INT8 vs Tested by Intel as of July 11th 2017: 2S Intel® Xeon® Platinum 8180 CPU @ 2.50GHz (28 cores), HT disabled, turbo disabled, scaling governor set to “performance” via intel_pstate driver, 384GB DDR4-2666 ECC RAM. CentOS Linux release 7.3.1611 (Core), Linux kernel 3.10.0-514.10.2.el7.x86_64. SSD: Intel® SSD DC S3700 Series (800GB, 2.5in SATA 6Gb/s, 25nm, MLC).Performance measured with: Environment variables: KMP_AFFINITY='granularity=fine, compact‘, OMP_NUM_THREADS=56, CPU Freq set with cpupower frequency-set -d 2.5G -u 3.8G -g performance. Caffe: (http://github.com/intel/caffe/), revision f96b759f71b2281835f690af267158b82b150b5c. Inference measured with “caffe time --forward_only” command, training measured with “caffe time” command. For “ConvNet” topologies, dummy dataset was used. For other topologies, data was stored on local storage and cached in memory before training. Topology specs from https://github.com/intel/caffe/tree/master/models/intel_optimized_models (ResNet-50),. Intel C++ compiler ver. 17.0.2 20170213, Intel MKL small libraries version 2018.0.20170425. Caffe run with “numactl -l“.