Extreme Cloud Centre of Excellence (ECCoE)
Extreme Cloud is a ‘catch-all’ term to describe the category of intensive workloads for solving complex problems in a variety of industry sectors. Typical workloads include Computational Fluid Dynamics (CFD), seismic analysis, data analytics, Machine Learning model training, media rendering and risk modelling. Extreme Cloud refers not only to established High-Performance Computing / High Throughput Computing (HPC/HTC) but also to data-intensive workloads, Artificial Intelligence (AI) and Machine Learning (ML) initiatives. Demand for Extreme Cloud workloads is estimated to grow to tens of billions of dollars over the next few years.
HPC and HTC environments typically require a large capital programme of works to establish a data centre and the purchase of thousands of cores of compute power to run the required workloads. Companies are then tied into this investment for a minimum period in order to realise the depreciation of their assets. Ongoing costs include power, cooling, maintenance, security and employing a specialised team to run the HPC/HTC facility. That investment means an organisation has a fixed number of cores available to carry out HPC/HTC workloads. When demand is low, utilisation is low. When demand is high, users have to wait until cores become available which hampers innovation and ultimately could impact the profitability of a company.
Market demand for HPC is estimated to grow to US$45 billion by 2022.
The good news is that Cloud offers a different way of thinking about Extreme workloads. By avoiding the large capital up-front investment, companies can track Moore’s law ensuring that for every doubling in computer power, costs remain the same. Another way of thinking about this is costs will halve every two(2) years for the same workload. Cloud allows for the ultimate flexibility. When demand is low, compute can be decommissioned to reduce costs. When demand is high, compute can be provisioned as required so users do not have to wait.
The Extreme Cloud Centre of Excellence (ECCoE) will help organisations understand how Cloud can be used to augment or replace their existing HPC/HTC facility. Appsbroker does this by providing: a set of benchmarks for standard HPC/HTC workloads; an engineering service to assist an organisation prove out their workload on Google Cloud with the latest Intel CPU architectures; and clear articulation of the business case and ROI for adopting a Cloud centric approach. A week long engagement can be booked in Appsbroker’s state of the art customer facility to accelerate Google Cloud adoption for your organisation.
Benefits of ECCoE (Extreme Cloud Centre of Excellence)
The Intel® Xeon® scalable Platform with AVX512, MPI and the Intel® compiler chain make for the best performance on any equivalent CPU architecture
The variable nature of Extreme Computing means that it is well suited to the Google Cloud per second billing model and running these workloads in the cloud avoids minimum spend commitments and long-term hardware contracts.
Large upfront expenses are replaced with low, variable payments that only apply to what you use. Even greater savings are available using preemptible VM’s - short-lived compute instances suitable for batch jobs and fault-tolerant workloads.
Core and memory densities on the latest Intel CPU architectures provide the best price / core
Appsbroker provides a platform and tooling to make HPC/HTC consumption on Google Cloud a frictionless experience
Proving Out the Customer Solution
Over the course of a week-long engagement, Appsbroker will work with customers on a chosen Extreme workload in order to prove out the following:
Optimal way to run your workload on Google Cloud and optionally in a hybrid on-premise / Cloud configuration
Your workload running on Google Cloud and Intel Skylake
A business case that articulates the ROI that Google Cloud & Intel brings to HPC/HTC in your organisation
Next steps and indicative costs to establish this capability in your organisation
The elastic compute power available via Google Cloud is ideally suited to Extreme workloads and varying demands. Google invest billions every year on Security, Data centres, renewable energy sources and the latest technology to make their Cloud services robust to Internet threats and scalable to meet their customer’s needs. There is no lock in with Google. Google believe customers will continue to use their platform because of their services, flexibility and competitive pricing.
Intel’s Xeon® processors based on the new Intel® Xeon® scalable Platform, offer significant per-core performance enhancements and engineering advancements (AVX 512) that support a wide range of workloads. Google Cloud is first to market with the latest Intel® Xeon® scalable Platform and via the Google / Intel strategic alliance, will continue to be the first Cloud hyperscaler able to offer the latest CPU technology for Extreme Cloud workloads.