Power make it possible: three stories from a data scientist
Learn why IBM® Power® is the right foundation to ensure secure data, scale efficiently, and make AI computing capabilities easier
In my six years as a data scientist at IBM, I have worked on a range of projects with clients across many industries. The needs of these clients are consistent:
- Ensure secure data protection
- Become more responsive and scale efficiently
- Uncover insights from data faster with AI
Unequivocally, to help clients achieve these three objectives I recommend the new IBM Power E1080. It is the best foundation for your IT infrastructure, engineered to provide the agility to meet and exceed your goals. Here are three client experiences that explain why.
Secure data science without intervention
Four years ago, I worked on an exciting new data science project at an Australian bank. The bank wanted to use deep learning to predict the risk of customers defaulting on credit card payments. The bank had plenty of data, including transaction history for more than one million credit card holders, but like most organizations they treated customer data with the utmost care. It would take months for the data security team to even allow this data to be offloaded onto another system in their own data center for analysis and processing. Even after getting the approval, we still had strict encryption and user-access restrictions to the system the data was being processed on. This slowed our progress and we ended up undertaking a massive de-identification and transformation effort to ensure the data could be used without limitation.
When processing and analyzing data, security is of the utmost importance. The new IBM Power E1080 addresses many of these concerns enabling simplified memory encryption through end-to-end security at the processor level, with homomorphic encryption capabilities that enable data processing without decryption. The new Power server is designed to help ensure security of customer data.
Efficient scaling with hybrid cloud
While security is top of mind for most organizations, responsiveness is a close second. Responsiveness often requires architecting a hybrid cloud approach that offers the best high-performance platform for each client workload. This was why a telecommunications company I worked with needed to build a hybrid-cloud-based data science platform on top of Kubernetes and IBM Power.
The platform’s purpose was to accelerate the production of AI. The operations team and data scientists on the project had no experience with either Kubernetes or the accelerated IBM Power servers used for compute. That didn’t matter; it all just worked. The value lay in being able to build a container image that ran on both Power and x86 with Kubernetes and containers, so the data scientists could easily install their preferred tools and go to work. Throughout that first year they used the platform to develop and deploy solutions for field workforce optimization, fraud detection and development of new enterprise products.
Optimizing solution development and deployment, and saving costs ensure clients can stay competitive in today’s market. IBM Power will enable efficient scalability with consistent pay-for-use consumption across both private and public cloud, while enabling faster data processing1, 2.
Hear from Carl Jones, Vice President, Power Hybrid Cloud Software about why the Power E1080 is the perfect solution for any organization seeking a more flexible and scalable foundation.
A new type of architecture with AI built into the core
I have worked with many clients on AI projects and the missing piece of the puzzle has been the ability to run and accelerate production AI models on the enterprise systems where the data resides.
The telecommunications company, I mentioned previously, is a great example. The enterprise developed an AI model that used syslog, a standard for message logging, and Netcool alarm data to identify the root cause of an outage on their network. This model was a crucial component that enabled them to respond and fix the outages. However, the implementation still required transferring 100GBs of data to their inference platform, creating a bottleneck that reduced their ability to meet critical SLAs. The ability to have their AI model on the same system as their data would have made all the difference, affording the organization efficient, secure, and seamless AI computing capabilities.
The Power E1080 makes this possible. This new server has an AI Matrix Math Accelerator engine built into every core. AI is at the heart of this machine. The same machine that runs your enterprise systems can now accelerate your production AI models, because the AI runs where the data lives. That means lower latency inference, incremental model training, and more value from your AI.
Hear more from Dr. Silvia Mueller, Ph.D., an IBM Distinguished Engineer, FPU and Arithmetic Acceleration Engines, about how the IBM Power system was designed with client needs in mind, creating a more powerful and cost-efficient solution.
Looking ahead
As a data scientist, I foresee that the biggest challenge for most enterprise organizations will be how to effectively handle and maintain the incoming and collected data. The volume of data we create and collect is only increasing, so organizations need to process that data intelligently, automatically and at the point of collection with AI. By doing so, you can confidently keep what you need and get insights instantly. With a hybrid cloud approach, and with the right platforms and tools in the right places, your enterprise will be empowered to respond to market demands faster and with agility in an ever-accelerating age of intelligence.