This post follows up to the latest entry in the Tech Talk series with Gestalt IT's Russ White in which we discussed the fallout from the Meltdown and Spectre vulnerabilities disclosed earlier this year by Intel.
In his last entry, Russ discussed how composable systems like Liqid's can help reduce surface threat areas by electrically isolating workloads. Traditional static compute solutions often lead to intermingling of compute processes that often have varying security requirements, increasing overall data vulnerability. Composable solutions, by contrast, eliminate the need to right-size a system to meet the application demands, allowing applications to perform at optimum levels while enabling reallocation of unused resources. This approach reduces overall infrastructure requirements while minimizing the threat surface by allowing IT administrators to electrically isolate sensitive applications from outside compute processes.
InsideHPC (February 2018) reported that future CPU designs will by necessity take a "safety first approach," favoring security over performance, at least in the next generation of processors designed to address security issue at the core. Several approaches, such as adopting blockchain-oriented capabilities, are being explored as a foundation for a more transparent, secure architecture. These and other future solutions to the Spectre and Meltdown vulnerabilities will introduce data performance challenges in next-gen chips, which are at odds with increasing workload and application demands. Over the next decade, AI-driven applications will be processing enormous volumes of IoT data in both consumer and industrial environments at edge computing facilities. Major slow-downs in processing power involved with securing compute resources and protecting digital assets must be offset.
GPUs can absorb some of the processing burden from CPUs and are being used to handle an increasing set of data-intensive, mission-critical applications. Infrastructure leveraging composable solutions, which can intelligently assign GPUs and other hardware assets to where they are needed in the system, will further offset performance challenges of increased CPU-level security while minimizing the economic impact of using these additional resources.
The Liqid Team appreciates the opportunity to engage in this discussion with Russ and the GestaltIT team. We are wrapping up on this topic, but we'll be back soon with another Tech Talk addressing issues related to two events we'll be attending later this month: The Open Compute Project Summit on March 20-21 and the GPU Technology Conference, March 26-29. Come by the Liqid booth to learn more about the composable future of IT.