The cloud has created a “new normal” in how compute resources are delivered and consumed, creating new businesses and streamlining IT services for enterprises. At Intel’s Data Centric Innovation Summit, we shared how the Cloud Service Provider (CSP) segment is leading the wave of data-centric computing, pushing the limits of compute, memory, storage, and network connectivity.
Intel’s CSP segment revenue doubled over the last four years with a compound annual growth rate of 30% from 2014 to 2017 (source: Intel). Looking forward, strong growth is expected to continue as we explore the confluence of cloud architectures innovations, the growth of the “cloud edge”, cybersecurity, network transformation, and acceleration of AI.
CSPs are not a monolithic block with uniform needs. Each has their own business priorities, unique workloads, and datacenter design points, so they “personalize” their infrastructures to differentiate their services. This diversification drives us to engage deeply with our customers, going “under the hood” with them to identify particular hardware and software optimization points that will increase efficiency and expose the full performance of the Intel® platform to offer the best user experience. We use this type of engagement model with the both the largest CSPs and smaller organizations looking to grow their footprint in the market. It’s exciting to engage with them as they build out their own unique capabilities and approach to the market, creating a diverse range of choices for end users.
Intel is in a strong and unique position to partner with CSPs worldwide, not only to drive the best performance and TCO for their infrastructure, but to engineer and deliver new services that deliver user value.
Transformative Platform Technologies Optimized for Cloud
One factor in our success in the cloud is our ability to customize CPUs to the specific needs of individual customers and their workloads. In the past five years, our volume of CPUs that have been customized to a CSP’s unique needs has increased from 19% in 2014 to 50% in 2017 (source: Intel). These customized CPUs are tailored for specific operation profiles, differentiated by power, thermals, core-count, I/O, and other features in response to the customer’s needs. The optimization does not stop at the component level; it expands across our broad portfolio of compute, storage, memory, and networking products, and up and down the cloud stack.
“We have been working with Intel for over a decade on highly custom silicon that enables us to deliver a choice of high performance compute options for our customers. AWS continues to innovate by offering new Intel® Xeon® Scalable processor based EC2 instances like M5, R5, C5, and most recently the z1d instance that offers sustained all-core turbo speeds of up to 4.0 GHz, the fastest of any cloud instance.”
– Amazon Web Services
Toutiao is one of the fastest growing digital content providers in China, providing a compelling news headline app serving hundreds of millions of people. A key success factor driving this traffic is Toutiao’s personalized content recommendation engine based on machine learning and infrastructure optimized for the best user experience. Toutiao’s next-generation data-driven infrastructure is based on Intel® Xeon® Scalable processors with an innovative platform design philosophy, achieving balanced system design with Intel's high-speed Ethernet and Intel DC SSDs to fuel the performance scale required for Toutiao’s users.
Another example of the power of partnership is Oath Inc., a subsidiary of Verizon Communications and the parent of its many digital media divisions, including Yahoo!, TechCrunch, and Tumblr. Intel and Oath collaborated on an ambitious cloud modernization and optimization plan, spanning hardware infrastructure, data center design, and software optimization. Nothing was left unexamined. In addition, Intel and Oath worked together on a customized, configurable CPU that was optimal for their services and, when deployed across their data centers, reduced their overall TCO.
Solution Optimization and Differentiation
In addition to customizing at the CPU level, we work with CSPs to ensure their software is getting the most out of the Intel® platform’s features by exposing them through our extensive software tools and libraries.
One example is Tencent, where we helped enhance the WeChat* user interface through voice recognition, and convert user voice into text for hands-free input. Through improving the Long-Short-Term Memory (LSTM) model on the latest Intel® Xeon® Scalable processor, Intel helped Tencent WeChat accelerate voice recognition performance by 16X, and enabled Tencent to get full performance from AI on their workloads.
Furthermore, Amazon Web Services launched its Deep Learning AMIs that come with TensorFlow* 1.6 built with Intel® Advanced Vector Instructions (Intel® AVX) and Intel® Math Kernel Library for Deep Neural Networks (Intel® MKL-DNN) to optimize deep learning training performance. With these optimizations, training a ResNet-50 benchmark with synthetic ImageNet dataset in Intel® Xeon® Scalable processors-based AWS C5 instance was 7.4X faster than training on the stock TensorFlow 1.6 binaries.
Helping Accelerate the Cloud Marketplace
At Intel, we want to help a wide range of CSPs build the most optimized cloud infrastructure. The Intel® Cloud Insider Program offers training, testing, and community support resources to help CSPs build an industry-leading cloud infrastructure. We just launched the Intel® Cloud Insider Advanced Technology Sandbox program globally, offering easy, early access to the pre-release hardware in compute, storage, network interface controllers, and more.
With over a decade of engagement with CSPs in driving infrastructure optimization, it’s fantastic to be recognized by our customers for our joint work. At the Google Cloud Partner Summit in July, Intel received the Google Innovative Solution in Infrastructure Award for our work in 2017, and we look forward to even more fruitful collaboration this year to deliver SAP HANA* in-memory database services equipped with Intel® Optane™ DC persistent memory through Google Cloud.
Intel’s investment in a silicon foundation across compute, memory, storage, and network, and a system’s approach to workload optimized infrastructure has created unmatched innovation for the drivers of data-centric infrastructure capability. The diversification in cloud requirements, services, and capabilities is more complicated than ever before. When we collaborate with CSPs, we look at the system and not just silicon as the design point and we use our hardware and software engineers to unlock more value and differentiation.
To learn more about Intel resource that can help Cloud Service Providers, visit us here.
Check out all the news from the Intel Data Centric Innovation Summit here.