Quantum Computing and storage

The OVH cloud is offering quantum computing-as-a-service (QaaS) using the Quandela Belanos photonic quantum computer interfaced to traditional (classical) object storage.

Quandela is a France-based, European startup, founded in 2017, focussed on photonic computing. Its technology uses single photons as qubits, units of quantum computing, and Belanos is its 12 qubit (quantum bit) quantum computer. It also produces the MosaiQ 6 to 24 qubit system and Canopus, a 24 qubit offering. An early version of the MosaiQ system was installed at an OVHcloud datacenter in 2023. OVH is now offering Belanos on a pay-as-you-go basis.

Miroslaw Klaba, R&D Director at OVHcloud, said: “We are delighted to deliver on the promise of the Quantum Platform by adding a second reference quantum computer, Belenos from the French company Quandela. The quantum revolution accelerates and OVHcloud is taking its part as the European Cloud leader within the ecosystem.”

Belanos workloads and their data are stored on current, bit-based, object storage, not qubit-based storage.

This distinction is important as there is no qubit-based storage, and that is because the very act of reading or sensing a qubit’s value destroys that value. In that sense, qubit storage is effectively impossible.

In the classical, digital bit, world a bit has a physical instantiation, such as the electrical charge in a DRAM cell, the magnetic polarity of the grains in an area of a disk drive’s magnetic recording medium, or electric current flow through semiconductor channels in NAND. Detecting the bit’s value, binary 1 or zero, does not change the bit’s value; it persists when it is read or sensed.

This is not the case with qubits; quantum bits. They are instantiated in physical phenomena at the quantum mechanics level as the quantum state of a physical 2-level (or effectively 2-level) quantum-mechanical system. Such a system obeys the laws of quantum mechanics and has two distinguishable basis states (usually labeled |0⟩ and |1⟩), described as superposition, and can function as a physical qubit.

Quantum mechanics applies at the level of atoms, electrons and sub-atomic particles, and photons. This is a weird world in which an item’s state can be entangled with another item so that one of the pair’s state influences the other’s, irrespective of distance. Items can act as both particles and waves, and aspects of an item, such as its energy and angular momentum, don’t come in continuous values but in discrete packets 

Quantum computing manipulates qubits to perform calculations and the qubit instantiations are said to cohere if they retain their states through processing and decohere if they do not. When a qubit’s value is sensed in some way then its quantum state, its superposition, is destroyed; it decoheres. For example, in a photonic quantum computer, a photon is moved into an interferometer or single-photon detector where it interacts with molecular-level or macroscopic level and (larger objects) and loses its superposition and entanglement state.

We are simplifying things here but, effectively, this means you cannot suspend a quantum computing system’s state and store it over time. Researchers are looking at the concept of QRAM (Quantum Random Access Memory) for short-term persistence, but it’s mostly theoretical and experimental.

All this means that a quantum computer cannot have a workload fed into it in the same way that an x86 processor, with its DRAM, gets its applications or containers or whatever. The digital bit world is continuous and interchangeable across its various instantiations; SRAM, DRAM NAND, HDD, tape, etc. The qubit world is not, being limited, effectively, to what’s going on inside a quantum computer.

It means that quantum computing operates inside a classical computing environment and there has to be a translation between the digital bit world and the qubit world.

To run a workload on Belanos you start with it and its data set being stored on a classical computing and storage system. This has then to be encoded into a photonic quantum computing state and sent to the QPU (Quantum Processing Unit). This can be done using a Jupyter notebook - open-source, browser-based environment - and Quandela’s Perceval, its open-source photonic computing SDK. 

Effectively you prepare the input state for a photonic circuit and then it’s used to create a set of photons and send them into the Belanos QPU in a single shot, at the end of which the exit photons and their modes are detected and measured. Multiple shots, tens of thousands or more, are typically run as the output of a run is a set of probabilistic measurements. Each shot is stateless and its data re-encoded for each run. These results are post-processed in classical (binary bit-based) computers.

Although quantum computing can be very fast, total job time is much slower as the QPU has to be book-ended by classical computing input and output processes. Where a classical computer could take years to run a particular workload, like RSA decryption, a QPU could possibly run it in seconds or minutes and that's what is energizing QPU research. 

OVH and Quandela say Belanos can be used “to experiment with new algorithms in innovative domains such as image sorting and generation, accelerated AI calculus, or quantum machine learning (QML).” This is not for general business use. The two companies say there are “New use cases in the fields of electromagnetic simulation, structural mechanics, engine combustion, material simulation, meteorology, and earth observation also benefit from the advances in quantum computing.” 

Think of them actually saying “may” also benefit. The best known application for quantum computing is RSA encryption breaking which has generated a rise in post-quantum cryptography, see Cohesity, Commvault, NetApp and Quantum for example. 

But no quantum computer has actually broken any RSA encryption method, it being a theoretical capability only.

There is more infomation about OVHCloud's quantum computing here.

Bootnote

A Quandela quantum computing glossary is a useful resource.