Objectives
DEDALUS aims to provide a next-generation platform that offers the best of both classical and quantum worlds, effectively partitioning big data algorithms into parts solvable by classical or quantum computing, combining the advantages of both technologies.
In this direction, a cost estimator will select the most efficient computing technology for each task. Special attention will be given to mechanisms for efficient and effective transformation of quantum input (preparation) and output (measurement) to be usable by both classical and quantum algorithms, with an emphasis on translating queries and data between quantum and traditional algorithms. This cost estimator must carefully weigh access latencies, as CPUs and GPUs are typically part of the server’s architecture, while QPUs are usually accessible only via the network, as remote cloud services. An additional challenge is the wait time spent in various job queues in such configurations. Even for widely used GPUs, data transfer costs, limited memory for intermediate results, and differences in computational models make effective use of GPUs for query evaluation difficult, let alone for a newer compute model like QPUs. Moreover, while a QPU-based algorithm might be faster in theory, overhead from translation and data movement operations can result in worse performance than CPU execution. For this reason, in DEDALUS, it is essential to define new computational structures, improve them at the platform and interface level, and optimize quantum architectures to support hybrid algorithms and minimize latency between classical and quantum systems.
Overall, a new hybrid relational query optimizer, exploiting quantum advantages to search the large space of alternative execution plans and exploring indexing for leveraging quantum benefits.
Methodology
To address all the aforementioned challenges and objectives, an innovative infrastructure will be designed and implemented. The research and development methodology of the project consists of four phases, as illustrated in Figure 1.
Figure 1: Research & Development Methodoly of DEDALUS project
- Design: The design phase will include an in-depth investigation of existing hybrid architectures—including, beyond CPUs, TPUs, and GPUs, also QPUs—identifying bottlenecks in current solutions and defining the requirements for a new solution in the field. Subsequently, a hybrid architecture will be designed and proposed, focusing on novel computational structures, platform-level and interface-level enhancements to optimize classical–quantum architectures, thereby supporting the required hybrid algorithms and minimizing latency between classical and quantum systems.
- Development: The development phase will focus on distinct components, including the creation of a cost estimator to compute the cost of various operations across the available types of processing units, a data translator to handle input and output transformation to/from QPUs, and a new query optimizer that leverages quantum computing to explore state spaces beyond the reach of classical algorithms. Additionally, the feasibility of using existing or new QPU algorithms for implementing query operators will be examined to exploit the quantum advantages of the new technology. A GUI and an API will also be developed, while a quantum simulator will enable data scaling in computational environments where QPUs are not yet available, reducing both development and evaluation costs in the next phase.
- Evaluation: During the evaluation phase, the entire infrastructure will be validated and assessed in a hybrid environment that includes, among other components, access to QPUs. In addition to experiments on available QPUs, the simulator developed during the project will enable scaling experiments beyond the limitations of current QPU technology, preparing the infrastructure for future advancements in the field. Evaluation will focus not only on classical techniques but also on approaches not yet feasible with current technology (e.g., quantum indexing techniques). The results will also inform the exploitation plan, highlighting the added value of the technologies developed throughout the project.
- Exploitation: This phase spans the entire project, encompassing appropriate dissemination activities carried out through various channels by the involved partners—such as the project website, promotional materials, publications, social media, print media, postgraduate courses, STEM seminars, and related events. Following project completion, dissemination will continue through a sustainability plan and the release of services in an open-source repository. Project results will also be shared with the industry via existing channels (partners include QNT S.A., with ongoing collaborations with FORTH, IBM, and the University of Stuttgart) and will be further disseminated through a COST action in relevant fields, currently co-organized by the university and soon to be submitted.