In our previous introduction to quantum computer technology, we explained how quantum technologies obey the laws of quantum mechanics, which in turn, describe the physics of microscopic phenomena such as manipulating single atoms and single photons. The primary quantum phenomena used in quantum technologies are superposition, entanglement, and no cloning. Thanks to superposition, the quantum bits in a quantum computer can take the 0 and 1 values or a superposition of them, and therefore a quantum computer can process many operations at the same time.
For quantum computing, Ericsson Research has identified several potential use cases in telecom:
- physical layer processing of the user data plane in the RAN (quantum Fourier transform and quantum linear solver)
- clustering for automatic anomaly detection in network design optimization project (quantum K-means algorithm)
- prediction of the quality of user experience for video streaming based on device and network level metrics (quantum support vector machine)
- database search at the data management layer (Grover’s algorithm)
The physical layer processing functions, such as modulation schemes, waveform, frame structure, reference signals, multi-antenna transmission and channel coding have very tight requirements on latency and thus they are most likely executed on local hardware (as opposed to a cloud-based hardware). For example, the inverse fast Fourier transform and discrete Fourier transform, DFT (classical counterparts of iQFT and QFT) are part of the waveform generation and thus executed in a local accelerator. Therefore, this algorithm cannot be processed in the cloud. Nonetheless, algorithms that control performance of the RAN and perform data analytics are not time-critical and can be executed in a cloud environment. Examples of these algorithms are database search, K-means, support vector machines, the classical counterparts of Grover’s algorithm, quantum K-means and quantum support vector machine algorithms.
In our previous introduction to quantum computing algorithms, we explore all of these algorithms and illustrate how the current hardware limits the number of qubits, gates, and operations to perform (as qubits are connected in a specific way only) and that the circuit depth is critical when running on near-term noisy Q machines.
Let us now evaluate the possible cloud deployments or local deployments from the technology maturity point of view.
When can we expect to see quantum RAN deployments? And where?
A quantum chipset for 50-100 qubits based on superconducting qubits measures a few square centimeters at most but since superconducting conditions are only achieved at temperatures around absolute zero (0 K = -273 C = -479 F), today the chipsets can only operate within refrigerators. This renders the current footprint up to . Due to this large footprint, we believe that early deployments of quantum computers for the RAN will be cloud-enabled quantum accelerators that reside in a centralised location (e.g. a data center) and are accessed through cloud (see Figure 1), quantum computer-as-a-service.
In this scenario, the control and management functions in the RAN needs to be virtualised and then could be executed in the quantum processor. We estimate that these deployments could begin to take place in 5-10 years when quantum computers are expected to count on 100 logical qubits (the estimated number of qubits required to execute the Quantum Fourier transform with input vectors of 4000 elements).
As the technology matures and quantum-chips are available in more compact form-factors (within the next 10 years), they could be deployed closer to user premises. Thus, in this second scenario, the quantum processor would be collocated with the baseband unit but in this case, the processor would be cloud-enabled to target acceleration of virtualized RAN functions. In this scenario, the quantum accelerator is part of a distributed quantum computing system.
Finally, when technology allows for miniaturization of both the chipset (see an example by Intel) and the refrigerating technologies, the quantum processor/computer could be locally deployed in the digital unit to substitute the current accelerator. In this architecture, the Q processor or accelerator is physically placed in the digital unit and the purpose is to accelerate data plane functions (such as Fourier transform) that currently cannot be virtualized due to latency requirements. In this scheme, the quantum processor would not be cloud-enabled, instead, it would be managed by the Host CPU in the digital unit. We estimate local deployments of quantum computers in a time frame of +15 years when either cold Q chipsets and cryo-electronics are miniutarised or room-temperature quantum chips are available. Note that time estimates are based on the extrapolation of roadmaps from leading vendors behind Quantum Computers.
Figure 1: Quantum cloud enabled deployment. Read more about RAN architecture to find out more about each of the specific functions.
The challenges ahead
Quantum computing is just one of the many functions towards the development of a quantum network that will deliver the quantum Internet, but it still has many challenges ahead. The most significant challenges that academia and industry need to address are:
- the development of error-correcting codes for error-free quantum computing
- the building of architectures and interfaces between quantum computers and communication systems
- the development of reliable quantum memories
- the development of quantum programming languages, compilers and middle-ware stack
This series of blogs is an early assessment of how the quantum computers could potentially be deployed in the telecom infrastructure when quantum technologies are available off-the-shelf and how we need to redefine the algorithms so they benefit from the computational power advantage that quantum is said to offer.
This is a syndicated post. Read the original post at Source link .