Military Technology 05/2021

62 · MT 5/2021 C4ISR Forum are orders of magnitude ahead of task parallelism alone. Data is stored across multiple nodes for parallel processing, with most frequently used data pre-loaded into RAM to eliminate input/output bottlenecks. Ideal for supporting ABMS and JADC2 goals, Kinetica works with cloud services, such as Microsoft Azure Government cloud and Amazon Web Services GovCloud, to connect sensors from all military branches – Air Force, Army, Marine Corps, Navy and Space Force – into a single net- work. It is currently used in both defence and enterprise applications by a diverse range of customers, including the US Air Force, NORAD, USPS, Citibank, Telkomsel, MSI, OVO, and Softbank, among others. Deploying a Vectorised Database at the Tactical Edge Because it runs on industry standard GPUs and CPUs, a vectorised database such as Kinetica can be integrated into a deployable rugged system, hosted on tactical servers. For example, Kinetica can be de- ployed near the tactical edge, on a mobile ground platform like a HMMWV, to locally process sensor data. The position of enemy vehicles correlated with a historical timespan can be determined in a command post setting or on deployed communications vehicles, then distributed wirelessly to dismounted soldiers or commanders sitting in a command post vehicle. By eliminating the need to send data back to the command centre for processing, tactical decision-making in the field is greatly enhanced – and ensures analytic processing, COP and decision support are availa- ble, even in D-DILL or connectivity-denied environments, in which tactical organisations are unable to rely on communications links to large cloud infrastructure. An ideal tactical server for hosting Kinetica is the PacStar 453 from Curtiss-Wright Defense Solutions, an NVIDIA GPU-enhanced deployable, rugged, small form factor server. This best-in-class solution combines rugged environmental protection with a small footprint, enabling deploy- ment of Kinetica in small vehicles, or even backpack-based deployments. The PacStar 453 lies at the heart of the PacStar Tactical Fusion System (TFS), a COTS-based, modular, tactical and expeditionary, rugged pro- cessing and data distribution node for sensor-enabled platforms. TFS combines the PacStar 453 module with additional network route/switch, storage and wireless modules to provide a single compact solution for cloud connectivity and replication, as well as enabling local processing of large datasets in realtime. This type of tactical, expeditionary, rugged data centre is capable of hosting mission command, cloud/storage, sensor fusion, AI and analytics applications. The combination of the new class of vectorised database processing and analysis, hosted in rugged size-, weight- and power-optimised com- puting and networking hardware, packaged for deployment in harsh en- vironments, can deliver the secure, high-speed data replication capability needed to realise the DoD’s vision of JADC2 application, such as POL data processing and mobile cloud computing. cyberattacks. The three main goals of these capabilities are to enable warfighters to see themselves; to see the battlespace; and to understand the battlespace. To see themselves, warfighters need a user interface that provides visibility of friendly operations in cyberspace, including network and information system status, configuration and topology, key terrain in cyberspace, and risk. Key elements are network and computer security significant activities (SIGACTs), such as denial of service and significant security information and events. Seeing the battlespace involves correla- tion of enemy activities in cyberspace with friendly activities, to give the commander the ability to see the battlespace from the perspective of the cyberspace domain. The user interface should provide visibility of enemy operations in cyberspace, including enemy cyber forces and misinforma- tion activity. SIGACTs should include enemy jamming and loss of network components. To understand the battlespace, the user should be able to comprehend and appreciate the meaning of previously unseen events in cyberspace, so they can proactively plan support for multi-domain op- erations. At this level, information is incorporated from the social layer, including personal and organisational information, as well as information from networks and infrastructure in the area of interest (AOI). User-defined criteria will allow the user to create SIGACTs that are specific to the unit or situation. The Benefits of Vectorisation Vectorisation is ideal for IoT analytics, because it excels at rapid- ly and cost-effectively analysing streams of data, handling geospatial, time series, and graph analytics. One example of a vectorised data- base is Kinetica, which can quickly ingest and correlate airborne objects across sensors, building feature-rich entities, allowing military operators to deepen their data analysis capabilities and increase their situation- al awareness by combining functions currently performed by multiple isolated systems into a cloud platform, producing real-time intelligence for leadership. To rapidly process huge data sets, Kinetica runs on parallel process- ing systems based on powerful devices, such as NVIDIA GPUs, and vectorised CPUs, such as the latest Intel processors, like SkyLake or above, that support AVX512 floating point instruction set capability. What makes this type of solution different from previous C2 systems is that it is predicated on data-level parallelism that processes all of the data at once with a single instruction. This form of parallel processing can be over a thousand times faster than traditional task-level parallelism. Purpose-built to leverage parallel computing power, Kinetica optimally routes query processing in each node across CPUs and GPUs for fastest results. While not a traditional SQL database, Kinetica uses the industry standard Structured Query Language that gives SQL its name to query and analyse billions of rows of data in a matter of microseconds. Kinetica is purpose built for the unique requirements and demands of analysing data across space and time. At the core of Kinetica is a distributed, co- lumnar, memory-first database designed for workloads that combine task-level and data-level parallelism to achieve performance gains that The PacStar 453 GPU-enabled server provides tactical POL processing. PacStar’s TFS – a small form factor, all-in-one tactical fusion system.