Mission-critical edge processing
The HPE strategy evolves around the idea that corporations will use a mix of on-prem private cloud architectures along with a hybrid cloud model to achieve optimal results for a reasonable investment. Another tenet to the strategy is that everything will be understood through artificial intelligence and its components of machine and deep learning. Finally that a large portion of computing will be done at the edge. By the edge, HPE means near the creation of an event or transaction. Processing at the edge might eliminate the need to move the data, or at least most of it, back to the data center since it would have already been processed. Only aggregate data and outliers would make the long and somewhat expensive trip back to the data center. By way of example, I was reading of video ‘smart cities’ surveillance systems that were setup in China a few years ago. The raw video footage was 457PB per day. Imagine moving all that information every day and analyzing it. By the use of edge systems to capture, aggregate and determine requirements for sending information upstream, this was reduced to 76PB per day. That’s still an amazing amount to transfer but dropping from 457PB to 76PB is an impressive saving. This means systems at the edge would need to support transactional input from existing devices along with many new streams coming from IoT devices. Additionally they would need to have an ability for both OLTP/transaction processing along with OLAP/analytical processing to effectively process and analyze at the edge. Analytic code for such things as fraud, preventive maintenance, security and many other machine learning based algorithms would be dropped onto these edge systems after being developed on high-end machine learning systems at the core, such as the HPE Apollo and Cray systems. Ideally the edge system would have a mixed workload capability to balance high-priority streaming traffic from backend analytics. As more and more critical processing is distributed to edge systems, the more failure becomes intolerable, expensive and in many cases dangerous. Will the edge require mission-critical systems? NonStop with its known capability for high transaction analytical processing (HTAP) and its demonstrated mixed workload capability (ZLE) makes it an ideal fit as a mission-critical edge server. HPE has a series of edge systems called Edgeline. A virtual NonStop can run in a number of HPE Edgeline systems. We have booted on the 4000 (NonStop in 1U at the edge) but I believe we are favoring the newer 8000 edgeline. Bringing complete fault tolerance to the edge. Not bad.