Sydney is the home to one of five Dell EMC AI Experience Zones which are intended to smooth the path to adopting artificial intelligence (AI) technologies for businesses across the Asia Pacific.
All five of the new Dell EMC zones are in APJ — in addition to Sydney, Bangalore, Seoul, Singapore are host cities, with Tokyo opening its doors next month. The ‘Zones’ are based in existing Dell EMC Customer Solution Centres.
They are a multipurpose initiative, combining a showcase for the vendor’s products with meeting and training facilities. As well as promoting Dell’s reference architecture for AI, they will host educational initiatives for technical and business leaders, as well as allow the vendor’s customers to work on prototypes and proof-of-concepts using Dell-run infrastructure, either in person or remotely.
The concept builds on Dell EMC’s three-year experience with its HPC and AI Innovation Lab in Austin, Texas — a facility that incorporates a 13,000‑square‑foot data centre, including the Zenith cluster, which ranks 396 in the TOP500 list. It also draws on the experiences of the technology company’s customers when it comes to adopting AI solutions, said Andrew Underwood, Dell EMC’s APJ chief technology officer for HPC and AI.
The aim is to help customers get to grips with the infrastructure necessary for processing AI workloads, such as image classification, natural language processing, sentiment analysis, and topic classification, he said.
“These are the types of workloads that generate huge amounts of data — and the data sets are so big that no individual human, let alone the combined intellect or processing power of every human on the planet could actually work through this data in a timely manner to be able to undertake these tasks,” the CTO said.
AI using massively parallel processing “can actually complete trillions of trillions of trillions of tasks in a single second,” he added. “This is where we see that AI is fundamentally going to be leveraged today as well as the next decade.”
Underwood said that 64 per cent of CIOs in a 2018 global Dell EMC survey said that AI was important to their digital transformation strategy over the next three years.
Dell EMC has sought to simplify AI adoption by designing a modular, scalable and efficient architecture, which is showcased at the zones, the CTO said. It’s based on the vendor’s PowerEdge C6420 and R740 PowerEdge compute nodes, which integrate Intel’s Cascade Lake CPUs, Dell EMC’s Isilon F800 NAS, and its PowerSwitch S3048F and PowerSwitch S5232F for networking.
“Beyond infrastructure we’ve also validated a software architecture stack and services and support architecture stack that we see our customers are investing in when it comes to artificial intelligence” Underwood added. It includes a range of optimised libraries and deep-learning frameworks, as well applications including SLURM, HDF5, Spark, JupyterHub and Jupyter Notebook, to help customers immediately start using the Zone’s lab with “no down time and no requirement to install their own packages”.
Dell employs Bright Cluster Manager and Kubernetes for managing the infrastructure layer at the Experience Zone. Underwood said that for customers’ sysadmins and other IT leaders, Bright will offer insight into the hardware resources employed by a particular AI workload.
“As customers come into the Experience Zone and they start to leverage the infrastructure, we can potentially give them reports that actually say ‘Well, you might need a 100Gbit — or you might need 25Gbit, that’s more than sufficient, or this number of GPUs was sufficient for training your algorithm.”
It also gives data scientists an opportunity to optimise their software algorithms and reduce model training time, he added.
Engagement length will depend on where a customer is in terms of AI adoption, Underwood said. “We do know have a lot of customers that have a team of data scientists [who] will obviously look to be spending maybe two to four hours in the AI Experience Zone and getting an understanding from that,” the CTO said.
“We also have some customers who are interested in AI, want to understand the impact of AI, but aren’t necessarily ready to adopt artificial intelligence. They’ll actually be looking to attend things such as our executive briefing sessions on how they can manage an AI project, or identify an AI project, or even participate in our AI masterclass.”
“We also know there are some customers who could spend up to a week doing different types of prototyping to evaluate different techniques,” he added.