Trinity College Dublin’s Dan Kilper and University of Arizona’s Saikat Guha discuss the quantum cloud and how it could be achieved.
Quantum computing has been receiving a lot of attention in recent years as several web-scale providers race towards so-called quantum advantage – the point at which a quantum computer is able to exceed the computing abilities of classical computing.
Large public sector investments worldwide have fuelled research activity within the academic community. The first claim of quantum advantage emerged in 2019 when Google, NASA and Oak Ridge National Laboratory (ORNL) demonstrated a computation that the quantum computer completed in 200 seconds and that the ORNL supercomputer verified up to the point of quantum advantage, estimated to require 10,000 years to complete to the end.
Roadmaps that take quantum computers even further into this regime are advancing steadily. IBM has made quantum computers available for online access for many years now and recently Amazon and Microsoft started cloud services to provide access for users to several different quantum computing platforms. So, what comes next?
The step beyond access to a single quantum computer is access to a network of quantum computers. We are starting to see this emerge from the web or cloud-based quantum computers offered by cloud providers – effectively quantum computing as a service, sometimes referred to as cloud-based quantum computing.
This consists of quantum computers connected by classical networks and exchanging classical information in the form of bits, or digital ones and zeros. When quantum computers are connected in this way, they each can perform separate quantum computations and return the classical results that the user is looking for.
Quantum cloud computing
It turns out that with quantum computers, there are other possibilities. Quantum computers perform operations on quantum bits, or qubits. It is possible for two quantum computers to exchange information in the form of qubits instead of classical bits. We refer to networks that transport qubits as quantum networks. If we can connect two or more quantum computers over a quantum network, then they will be able to combine their computations such that they might behave as a single larger quantum computer.
Quantum computing distributed over quantum networks thus has the potential to significantly enhance the computing power of quantum computers. In fact, if we had quantum networks today, many believe that we could immediately build large quantum computers far into the advantage regime simply by connecting many instances of today’s quantum computers over a quantum network. With quantum networks built, and interconnected at various scales, we could build a quantum internet. And at the heart of this quantum internet, one would expect to find quantum computing clouds.
At present, scientists and engineers are still working on understanding how to construct such a quantum computing cloud. The key to quantum computing power is the number of qubits in the computer. These are typically micro-circuits or ions kept at cryogenic temperatures, near minus 273 degrees Celsius.
While these machines have been growing steadily in size, it is expected that they will eventually reach a practical size limit and therefore further computing power is likely to come from network connections across quantum computers within the data centre, very much like today’s current classical computing data centres. Instead of racks of servers, one would expect rows of cryostats.
‘Quantum computing distributed over quantum networks has the potential to significantly enhance the computing power of quantum computers’
Once we start imagining a quantum internet, we quickly realise that there are many software structures that we use in the classical internet that might need some type of analogue in the quantum internet.
Starting with the computers, we will need quantum operating systems and computing languages. This is complicated by the fact that quantum computers are still limited in size and not engineered to run operating systems and programming the way that we do in classical computers. Nevertheless, based on our understanding of how a quantum computer works, researchers have developed operating systems and programming languages that might be used once a quantum computer of sufficient power and functionality is able to run them.
Cloud computing and networking rely on other software technologies such as hypervisors, which manage how a computer is divided up into several virtual machines, and routing protocols to send data over the network. In fact, research is underway to develop each of these for the quantum internet. With quantum computer operating systems still under development, it is difficult to develop a hypervisor to run multiple operating systems on the same quantum computer as a classical hypervisor would.
By understanding the physical architecture of quantum computers, however, one can start to imagine how it might be organised to support different subsets of qubits to effectively run as separate quantum computers, potentially using different physical qubit technologies and employing different sub-architectures, within a single machine.
One important difference between quantum and classical computers and networks is that quantum computers can make use of classical computers to perform many of their functions. In fact, a quantum computer in itself is a tremendous feat of classical system engineering with many complex controls to set up and operate the quantum computations. This is a very different starting point from classical computers.
The same can be said for quantum networks, which have the classical internet to provide control functions to manage the network operations. It is likely that we will rely on classical computers and networks to operate their quantum analogues for some time. Just as a computer motherboard has many other types of electronics other than the microprocessor chip, it is likely that quantum computers will continue to rely on classical processors to do much of the mundane work behind their operation.
With the advent of the quantum internet, it is presumable that a quantum-signalling-equipped control plane might be able to support certain quantum network functions even more efficiently.
Fault tolerance and quantum networks
When talking about quantum computers and networks, scientists often refer to ‘fault-tolerant’ operations. Fault tolerance is a particularly important step toward realising quantum cloud computing. Without fault tolerance, quantum operations are essentially single-shot computations that are initialised and then run to a stopping point that is limited by the accumulation of errors due to quantum memory lifetimes expiring as well as the noise that enters the system with each step in the computation.
Fault tolerance would allow for quantum operations to continue indefinitely with each result of a computation feeding the next. This is essential, for example, to run a computer operating system.
In the case of networks, loss and noise limit the distance that qubits can be transported on the order of 100km today. Fault tolerance through operations such as quantum error correction would allow for quantum networks to extend around the world. This is quite difficult for quantum networks because, unlike classical networks, quantum signals cannot be amplified.
We use amplifiers everywhere in classical networks to boost signals that are reduced due to losses, for example, from traveling down an optical fibre. If we boost a qubit signal with an optical amplifier, we would destroy its quantum properties. Instead, we need to build quantum repeaters to overcome signal losses and noise.
‘Together we have our sights set on realising the networks that will make up the quantum internet’
If we can connect two fault-tolerant quantum computers at a distance that is less than the loss limits for the qubits, then the quantum error correction capabilities in the computers can in principle recover the quantum signal. If we build a chain of such quantum computers each passing quantum information to the next, then we can achieve the fault-tolerant quantum network that we need. This chain of computers linking together is reminiscent of the early classical internet when computers were used to route packets through the network. Today we use packet routers instead.
If you look under the hood of a packet router, it is composed of many powerful microprocessors that have replaced the computer routers and are much more efficient at the specific routing tasks involved. Thus, one might imagine a quantum analogue to the packet router, which would be a small purpose-built quantum computer designed for recovering and transmitting qubits through the network. These are what we refer to today as quantum repeaters, and with these quantum repeaters we could build a global quantum internet.
Currently there is much work underway to realise a fault-tolerant quantum repeater. Recently a team in the NSF Center for Quantum Networks (CQN) achieved an important milestone in that they were able to use a quantum memory to transmit a qubit beyond its usual loss limit. This is a building block for a quantum repeater. The SFI Connect Centre in Ireland is also working on classical network control systems that can be used to operate a network of such repeaters.
Together we have our sights set on realising the networks that will make up the quantum internet.
Dan Kilper is professor of future communication networks at Trinity College Dublin and director of the Science Foundation Ireland (SFI) Connect research centre.
Saikat Guha is director of the NSF-ERC Center for Quantum Networks and professor of optical sciences, electrical and computer engineering, and applied mathematics at the University of Arizona.