Conf42 Quantum Computing 2022 - Online

Quantum communication challenges in scaling quantum computers

Video size:


To scale quantum computers, the communication between them is key. The benefits of quantum computation are realized when that computation occurs in a distributed manner - this requires a quantum network to join those computational elements together, placing the prospect of quantum communication as a critical piece of the puzzle. However long-distance quantum communication is raising a number of questions that are not yet answered.

There are different modes of operations that a quantum network could have, each of them requiring quantum repeaters to guarantee end-to-end communication. All these modes come with their challenges in terms of communication rate, quantum memory management or cost of the network. This leads to additional challenges to address such as the reliability of information exchange and the integrity of end-to-end transport, as well as considerations around the engineering approach to quantum networking.

In this session, we will explore the role of quantum communication in getting quantum computers to scale meaningfully, the related challenges the industry needs to overcome and how Cisco is approaching these issues through close collaboration with selected academic partners.


  • Sam Samuel from Cisco Systems will give a talk on quantum communication challenges in scaling quantum computers. It'll go from basic quantum computing and how we can apply it in a variety of ways. And we eventually look at the problems that if we choose to go that direction, we will face.
  • There's an increasing amount of investment going into quantum computing. The total investment is over a billion dollars now. It includes quantum computing, quantum software and also quantum communication and quantum Internet type things. This talk will look at that particular area and more.
  • Quantum computing is very different from classical computing. A qubit has the superposition of zero and one. The other thing that quantum computing uses an awful lot is entanglement. Quantum computing works off quantum circuits that are probabilistic in nature.
  • Nevin's law is different from Moore's law in the sense that it's a double exponential rise in potency. As the potency of quantum computers increases, the likelihood of a quantum computer being able to break ciphering of current networking starts to go up.
  • In order to make teleportation work, we really need to extend how we do these teleportation, and we do that by successively swapping entanglement on a communication link. In order for us to increase distance, to increase the distance over which we're able to teleport, we have to combine successive segments together with quantum regenerators.
  • In order for us to do communications, we have to have qubits that have high fidelity and high purity. We also want techniques available to us that allow us to compensate for loss and decoherence. Network synchronization is also another problem. But the community, the academic community, and also the startup community are starting to address these problems.
  • How can I universally interpret programs for classical quantum communication? How can quantum data centers be virtualized in terms of resource and the management of those resources? And how can I find a cost effective way extending the rate and range of a which quantum communication can take place?


This transcript was autogenerated. To make changes, submit a PR.
Hello, everybody. I'm Sam Samuel from Cisco Systems. I'm here to give you a talk on quantum communication challenges in scaling quantum computers. This is a general talk. It'll cover a fair amount of ground. It'll go from basic quantum computing and how we can apply it in a variety of ways, whether it has poses threats for the way we currently use classical networks, and whether the same principles used in quantum computing can be applied to quantum communication, and whether that would be a beneficial direction to go in. And we eventually look at the problems that if we choose to go that direction, we will face. And these, of course, give you an idea of what the situation could be, or how the ability or the interaction between quantum computing and quantum communication can lead to it in the future and give you some idea of what that would be. So it's a longish talk, hopefully be interesting for you, and so we'll dive right in. So the first question we really want to ask is, why is it that there's a fair amount of interest in quantum computing today? And I guess the best way of answering that right now is to sort of look at the investment levels from the venture community, investment community into quantum technologies of any sort. And this chart here shows us that there's an increasing amount of investment going into this particular area. The total investment is over a billion dollars now, although this chart here is over two years old. So I'm assuming that the total modes investment is well beyond a billion and a half dollars that we managed to count in this particular chart. And that investment is spread over three things. It's spread over quantum computing, and it's spread over quantum software. And then it also includes quantum communication and quantum Internet type things. It seems natural that the way that the investment is spread is kind of weighted towards quantum computing, because without that, none of the other two really kind of make sense. And it's kind of easy to see that without a quantum computer having quantum software to run, then it won't be at all useful. So my interest is more to do with the quantum communication and quantum Internet side of things. And this talk will look at that particular area and more interest. But the point is, all aspects of quantum computing or quantum technology are being investigated, invested in recently. And this talk is trying to give you an idea of where we are in the state of all of that, really. So with that in mind, then, the first question we have to ask is, what is quantum computing? Quantum computing is very different from classical computing. In classical computing, we're quite used to dealing with bits a bit is zero or one. In quantum computing, we are dealing with a quantum bit, a qubit. And a qubit is different from a classical bit in the sense that it has the superposition of zero and one, meaning that the same bit represents zero and one simultaneously. And if we have a register of two bits, and that register will then represent, because of superposition, all possible combinations of terms and ones in that register. So, in other words, it has the entire gamut of all possibilities, these in one throw, if you will. The other thing that quantum computing uses an awful lot is entanglement. So, entanglement, in this case is where two particles are highly correlated, and it's very difficult to break apart an individual state between two particles that are entangled. So when it's difficult to break apart the state, then that means that the states are pretty much highly correlated, that we can't divide them apart. So it becomes very difficult for us to look into a quantum system and get a definite answer out of it. And entanglement is the means behind which it makes it very difficult for us to go in and inspect an individual particle. We have to sort of look holistically at the overall system. Quantum computing works off quantum circuits that are probabilistic in nature. So, in contrast to classical networking, classical networkings tend to be very network circuit, sorry, quantum circuits, as opposed CTO classical circuits. Classical circuits tend to be very deterministic in nature, that you apply the same computation to it, you get the same answer in quantum circuits. That may not necessarily be the case. The outcome is very probabilistic in nature. So these three things, superposition, entanglement, and probabilistic computation, come together when we do quantum computing. And the final figure in this particular slide shows that a classical circuit takes a predefined path through a computational, rather sequential, not predefined, because I know you can do conditional statements inside a program. But the point is that it's a very sequential way that the program executes, whereas in a quantum system, it's very different because of superposition, we're able to explore the entirety of the state space of the problem that the algorithm is addressing. And in so doing, as we explore the entirety of the state space, we can then come to a resolution of the answer quite quicker, more effectively than we could do otherwise. And that's the benefit that quantum computing brings. So how can we apply that? And so the next few slides look at what the possible implications are of having such a computer. So a computer that's able to explore the entirety of state space kind of changes, the way we look at the potency of calculations that such a computer would have. So this slide here has a table on it, and the table kind of indicates the generations that a particular technology would have. So operations zero, one through all the way to n. We're quite used to measuring the potency of calculations in a silicon based system, a classical based system, through Moore's law. So in other words, we expect every 18 months or so for a silicon based processor to double in its potency over time. And that's what the top line of this particular table shows. The second line, the double exponential line, is showing something called Nevin's law. So Nevin's law is different from Moore's law in the sense that it's a double exponential rise in potency. So it has a different sequence. So it's not a case of doubling every generation. It's a case of double doubling every generation. So rather than 1248, we have 2416, 256. So we can see that if we have a computer which has that kind of potency or that kind of doubling law associated with it, then it can be very potent, become very potent in a short number of generations. These double exponential law was discovered by a guy called Hartman. Nevin and Hartman realized that in his experiments, in order to prove the potency of the quantum computer that he had, he had to go and check it against a classical computer. And these classical computers were running out of potency far faster than he expected. When he looked into it, it was a double exponential law. Now, the interesting thing with that is that if you have algorithms, such as factorization algorithms, which can be optimized for quantum computers, then you start to see the potential that this kind of computation has versus classical computation. And if you have a means of factorizing numbers very, very quickly, then you have a means of actually breaking ciphers in classical networks, which is what we're going to show in the next slide. So if we have a quantum computer that has a double exponential rise in its potency in the number of qubits it's able to employ against a problem, then one of those problems could be the factorization of keys in a network. And so if we apply that to the current cipher algorithms that are in networks today, we can see that a quantum computer has the potential to break the cipher of certain key lengths. And this is what this chart is showing. And it also shows that there comes a point when Nevin's law starts to accelerate the potential for a quantity computer to break the ciphering in current networks. Now, this may seem a little bit down in the sense that, oh, the world will collapse. That's not, strictly speaking true, because, of course, the very clever people at NIST have run competitions to ensure that we are slightly ahead of the curve in terms of the new quantum ciphering, or post quantum rtography, rather, that would be applied to combat these techniques. In terms of computation. Though, if we go to the next slide, we really have to realize that the certain kinds of algorithms are more suited to breaking certain kinds of key. So keys which are PKI in nature are impacted by Shaw's algorithm keys, which are more symmetric in nature. Symmetric ciphering that's going on cannot be addressed as easily by Shore's algorithm. They can be addressed by a different algorithm called Grover's algorithm. And these algorithms are not as effective in breaking longer and longer keys. And so that's what this chart shows. So, long story short, as the potency of quantum computers increases, then the likelihood of a quantum computer being able to break ciphering of current networking starts to go up. But that data isn't necessarily being broken on the fly. It's because the data has been siphoned off from the network and is stored somewhere where a bad actor can take his time over trying to break the key if he has a quantum computer at his disposal. So that's one application of quantum computers. But of course, we can apply quantum computing in different directions, and there are other sorts of very complex problems that need solving, and there are other np hard problems out there that are in need of the attention of something like a quantum computer. And so we would like to be able to increase the potency of the computers that address that particular problem. So it becomes natural to state the following obvious point, that you can always make a more powerful computer by joining more powerful computers together. And this is also true for quantum computing. So if we're able to join quantum computers together to make a larger quantum computer, then we can solve more complex problems more effectively and probably quicker as well. So I want to turn this talk really now onto how we would do that, how we would try and join computers together. But before we do that, it's probably worth exploring what we mean by this. So the idea of having a parallel execution path in a program has been out there for quite some time. So back when I was at university in the late 1980s, early 1990s, there was an awful lot of work going on in parallel computing. So we were able to get an application, break the application down into a set of parallel chunks, and then try and speed up the computation by applying the parallelism to that particular application or the algorithm that we're looking at, we want to do it slightly differently for quantum computing. And the reason for it is that not all quantum computers are good at solving every single problem. In other words, there's no such thing as a general quantum computer. Yet these computers tend to be very specialized and look at either annealing like problems or optimization like problems, or they're very good at factorizing, but not necessarily a mixture of both. Of course, over time, the situation may actually change. But it's important for us to understand that not only as we present an application or a program to a network, we want to make sure that that application is broken down and distributed to the appropriate quantum computer. So that's another reason why we would like to have networking involved in this kind of evolution. So in order for us to communicate, it's probably worth reexamining some of the principles that we looked at in terms of quantum computing and apply them here a little bit more. We obviously want to ensure that a network is secure. So if we're able to communicate, and we want to make sure that the network is itself not able to be eavesdropped. Interestingly enough, if we're working in terms of quantum state and entanglement, it's impossible to clone an unknown quantum state. In other words, we cannot copy quantum state if we don't know what these quantum state is. So that kind of makes the prospect of having a secure quantum network in the future kind of appealing. If we were able to take advantage of this particular theorem, the no cloning theorem, communication would also require entanglement. So we mentioned entanglement in the past, very briefly, in passing. It's the same entanglement here. And this is where essentially two particles have overlapping wave functions. And as a result of the overlapping nature of the wave functions, then these is impossible for us to distinguish the individual state of each particle, and therefore the states are entangled. These interesting thing behind that entanglement is that, and we probably heard this in the popular scientific press, as Einstein's spooky action in these distance is that if we apply or we influence one of the entangled particles, the other entangled particle almost responds immediately. And so you get something, a situation like this, whereas we apply a force on one, the other one responds. Now, at this point, people tend to get a bit confused and say, well, that's breaking the laws of the speed of light and so on and so forth. And that's not, strictly speaking true, because for us to understand the impact of that, we have to take a measurement of the particle, send that measurement over CTO, the destination or the other half of the entangled pair, and then recover what the message was. And so, at that point, it all slows down to the speed of light, or slightly below the speed of light. So we don't actually break any laws of physics here, even though it can be an interesting thing to try and prove CTO yourself, that's not happening. So, 2d concepts here. And then the question then becomes, can we use at least the entanglement to be able to communicate over a distance? And this is where teleportation come in. So, teleportation, at its very essence, is that we have an entangled pair. So we have managed to overlap the wave functions of two photons. Say we've distributed them to the end of a communication link, and now we want to introduce to that communication link a particle that has state on it, and we want to transmit that state across the network. So, how do we do that? So, at the bottom of this slide here, we have this qubit, this matter qubit that we wish to transmit across the network. We need to entangle that matter qubit with the communication qubit on the left hand side of the chart here. So, in order to do that, we take a bell state measurement of the interaction between the two particles. That produces a measurement result. We send that measurement result over to the counterpart pair of the entangled pair at the other end of the communication link. We apply a localized operation to it, and then that localized operation recovers what the original matter qubit was as it entered the system. And we get these teleported qubit materializing at the far end of it. So we've not actually transmitted the state across the network at all. And hence, that's why it teleports itself. And so we have a mechanism, fundamentally, that we can use for communication across any kind of distance, and that's teleportation. Now, in order to make teleportation work, we really need to extend how we do these teleportation, and we do that by successively swapping entanglement on a communication link. So, here we have three maximum entangled pair segments, pair a, pair b, pair c, and what we're going to do is entangle pair a with pair b and have an entanglement between a and b, and then do the same again to get entanglement between c. And then we have an entanglement end to end entanglement across the entire link. So, in order to do that we apply the same principles we did in the previous slide. In other words, we do an entanglement swapping measurement between the far pair of a and the near pair of b. We send that measurement to the far pair of b. We do a local operation on it. And then that means that a and B, or the near pair of a and the far pair of B, are now entangled. And these we perform the same measurement again between the far pair of B and the near pair of C. We do an entanglement swapping measurement there. We send the measurement result over to the far end of C. We do a local operation there. And these, as a result of that, we have entanglement between a and C. So, to communicate now over that entirety of that link, we present the matter qubit to a, and we do the same performance again. We take a bell of state measurement of the result of that interaction, and we send that over to c, and we recover at c by doing a local operation, the original matter cupid that was presented at a. So that's how we would extend distance on such a system. The interesting thing is, while it appears that we have the answer in our hands, the truth of the matter is that in order for us to be able to entangle over a distance, we realize there's a limit to how far we can send an entangled pair and get away with. Rather. I think, quite honestly, my interest in quantum communication was actually through reading this particular paper here that I'm quoting by Suraf Kumar, and he kind of indicated that in order for us to generate entanglement at a certain rate, that rate will only propagate a certain distance. So that means we have to be able to regenerate entanglement if we wish to communicate over larger and larger distances. So this brings the idea of a quantum regenerator. So, in order for us to increase distance, to increase the distance over which we're able to teleport these, we have to chain successive segments together, and we do that chaining with quantum regenerators. Now, that's well and good, but that's kind of like a point to point thing. And really, we want to get to a network where we're able to direct how and where the entanglement goes. So we're able to pick out a source and a destination at will and ensure that the source and destination are entangled. And in order to do that, we need to be able to have at our disposal an adequate distribution rate of entanglement. In other words, if we have successful communication, we have to be able to ensure we keep distributing the entanglement to the ends, and therefore, we're able to do this module transportation that we mentioned in the previous slide. But also the swapping is also a means of us being able to route entanglement around the network. And that's kind of the critical thing, although it overlooks the fact I've not mentioned it specifically here. But quantum memory starts to play a very important part in that type of activity inside a network. So if we're able to generate entanglement at a sufficiently high enough rate, commensurate with the communication needs of a quantum network, and we're able to do entanglement swapping in memory, then we're able to manipulate entanglement around a network, and we're able to target a source and a sync. So that's what this slide is basically saying. So what are the challenges associated with that? Well, the challenges are essentially twofold. There are qubit challenges, and then there are networking challenges. I'll address the qubit challenges first, because we are dealing with particles, and these particles are very fragile in nature in the sense that they start to interact with their environment, or they're able to interact with their environment quite easily. So as a result, there's going to be errors in the way that these qubits behave. And that's not unnatural, of course. So that means that if we don't have the right interactions when we set up the quantum circuit, then that means we're going to have programming errors in initiating the states. And that's one set of challenges that we're going cto face. In other words, how faithfully can we program something these the qubits themselves are likely to interact with the environment that they're in, causing decoherence between the entangled particles. And that's actually quite an important one in the sense that we have to be very careful how we design these circuits to ensure that we minimize that kind of exterior influence onto the qubits that form part of the circuit. And that usually means we have to reduce these noise in the circuit by actually reducing the temperature of the circuit. And that's something which we'll come to in a minute. But at least for communications, we communicate a photon, and photons tend to be very much at root temperatures. So that's not necessarily a problem for quantum communications, but they are a problem if you want to go from quantum, a photon for communication into a matter qubit for storage and memory, say. And so there's a problem there. These particles that we're dealing with these qubits that we're dealing with have a lifetime, a shelf life, if you will. They don't necessarily last forever. In fact, they don't. The coherence times that we're seeing, in other words, the time at which we can guarantee the entanglement is not necessarily a long time. Obviously, the techniques are improving, and material size is improving all the time, and we can extend the lifetime of the coherence that these entangled particles have. But the fact of the matter is, they have a shelf life, and we have to ensure that all operations are done before the shelf life expires, if you will. The other sorts of problems that we have at this level are operation error. So is the gate, or is the device faithfully acting on the qubit in the way we expect it to be? And again, because of the interaction with the environment, that may necessarily be the truth, not because it doesn't want to behave properly. It's because that's the way physics works at that level. And then there are other issues as well with communication qubits that are well known and have been known for some time, and that is that photons can be absorbed by the fiber that they're traveling down. And so we're going to get photon losses. And that can also mean that we are losing state. And if we lose one half of entangled pair, then essentially we have a problem because we don't have an entangled pair anymore, and we have to go and regenerate and so on and so forth. That causes problems. Now, these are the kinds of challenges that we face at the qubit level, but it's also important to know that we're looking at the health of that qubit, too. The health of entanglement that that qubit has. And that's usually referred to in terms of fidelity. And fidelity is a measure between zero and one that sort of indicates how good something is, how accurate, or how well matched the integral qubits happen to be. And so it's not only that we have these errors, but we also have to make sure that before we start any of these operations, the qubit that we're presenting, CTO, the system, has a high fidelity to it as well. So that's another issue that we have to try and cover and produce mechanisms to ensure that we have high fidelity qubits available to us. So these are the qubit challenges from a networking perspective, there are a different set of challenges, one we've kind of covered already, which is in order for us to do communications? Well, we have to have qubits, photon photonic qubits, that have high fidelity and high purity. So if we're going to pass them through successive gates, or if we're going to send them over any distance, we really have to ensure that they're very healthy before they start their journey. And that's what this high fidelity impurity really means. We also want techniques available to us that allow us to compensate for loss and decoherence. And one of those techniques is quantum error correction. If we're able to get good quantum error correction and that prospect against decoherence, it also protects against quantum noise. But the problem here is that quantum error correction requires us to have many more qubits available to us, and so the overall population has to increase in order for us to take advantage of that. So that generation of qubits that we had before starts to play a more important part if we wish to go down this quantum error correcting route. Network synchronization is also another problem that we have to have. We have synchronization in networks today, but these level of synchronization that we're after is far tighter than we have in current networks, and it needs to be tight, because we have to ensure that when we generate entanglement, that the wave functions, we can measure accurately or determine accurately when particles are entangled. And that requires us to have a high degree of precision in the timing of the network. And so also the lifetime of the qubit starts at the moment it's entangled. And so we have to know how much longer we have left on some of these, on these qubits in order for us to complete the operations that we need within the specified amount of time. So it kind of starts to impact these way we schedule, as well as well as acknowledging the overall health through having an accurate measure of decoherence time on such particles as well. So those are the network challenges. So, is there a pragmatic way forward here? Interestingly enough, there is a paper that came out in 2016 by Saramin Marlutheran. I hope I pronounced his name correctly, but he and his colleagues produced an excellent paper that indicated that there's a generational shift. And being engineers and scientists, of course, we magically christened these 1st, second and third generations, which is quite novel of us. But the point is that each generation starts to improve the efficacy of the overall system by introducing quantum error correction in two directions. Either it's used to compensate for loss, or it's used to compensate for error, operational error. So focus on loss and operational error, and depending on how we apply, it depends on which one of these generations you have to be. So, to put it into perspective. So even though the paper was written in 2016, where are we today? Today we are still scratching around at the first generation quantum repeaters, which means that we still have a very long way to go before we get to second and third operations, where we have a lot more reliability into the overall communication system. But the community, the academic community, and also the startup community are starting to address these problems, and this is starting to produce some results that I think are looking good for how quantum networking will evolve over the next few years or so. We know that there's a roadmap, if you will, in terms of generational shifts for a quantum repeater that we need in order to endtoend distance over which we can communicate. That means that the prospect in the future is kind of interesting for us to keep plowing on in this particular direction. And if we do plow on and we're able to solve these problems, then the prospect of getting to a quantum Internet starts to look kind of appealing. Now, I need to put this into context. To get to a quantum Internet is not like, oh, we solve something tomorrow, and then in two years time we'll have a viable commercial product that you can go and buy, put into your network. Everything's great. It's not quite like that. I think we are talking maybe ten years or so before we start to get to a reliable network at reasonable entanglement rates that can start to teleport at reasonable rates. Then after that, there's going to be a few years of development before we get CTO, a quantum Internet where we can teleport with high fidelity over any indicated pair that wish to communicate over such a network can do so. So it's a long way to go for that, but we understand the processes around it. We understand the sorts of architectures that may evolve to cost effectively address that area, which is all good news. And the technology and the progress being made by various starts and very academic bodies are moving along at a feral pace, which kind of gives me hope that we should be able to do this in a reasonable amount of time. Reasonable here, being an ex wireless guy, I tend to think in ten year generational shift. So ten years or so, I think we should see staggering progress from where we are today. So, wrapping things up. So what's the possible end state and problems if we solve them, or problems yet to be solved? If we carry on down this path, what would this idealized future look like? So these diagram, this stick diagram here that I drew kind of indicates where I think these is going. And some of the problems that I've mentioned in the qubit network are not necessarily the problems I'm mentioning here. So an idealized future is something like this, that we have a person who is interested in writing a program for a very complex problem. So he writes, this program. This program does not necessarily have any knowledge of quantum networking in it whatsoever. The program, or the application is then presented to a virtualized quantum data center. Right? And the first thing that this realized quantum data center modes is it breaks down that program, these application, and it compiles it and manages it such that it then breaks it down into these components that are required by the classical computing to work on that particular program, and also break down the applicable parts of that particular program that are requiring the attention of a quantum computer. So that compilation is one problem area that needs to be solved. How can I universally interpret programs for classical quantum communication? These mix of them, if we're able to break it down, then the next natural thing is I need to be able to do the networking between quantum computers inside the same data center. That's like a very different. Well, I wouldn't say it's very different. It is networking, but it's probably a very different networking, as we can use direct transmission between these machines, more than likely. But the point is, as we try to extend distance, there's more than likely a knee point in the behavior that requires us to change from one kind of network into a different kind of network, and which brings in the second problem that we have, and that is to find a cost effective way extending the rate and range of a which quantum communication can take place. So that's the, if I'm able to extend the range of a which I can get quantum computers to talk cto each other, that means I can then extend the number of quantum computers that talk to each other, and then I get a more quantum powerful quantum computer, and so on and so forth, which is the aim of that one. The final one is, and I think this is the one that's going to. Well, actually, I think a lot of these things are going to employ my colleagues for a very long time. But the final problem that's left these is how can quantum data centers be virtualized in terms of resource and the management of those resources? And so I think the first problem, the universal interpretation, and then the virtualization of quantum data centers, both of these, I think, are going to occupy my colleagues that are doing quantum computer science for many years to come. Whereas the networking pieces of it where the internal networking of a data center, quantum data center, rather, or the external extension of a quantum data center, will be occupying me for quite some time. So I've gone through quantum computing and its application to quantum communication. I've taken you through a whole bunch of stuff. I've taken you through a journey through the impact of quantum computers onto current cryptography that we use in networks today and the potential threat that there is. I've taken you through how that same principles can be applied for quantum communication and the benefits that could bring. I've indicated to you some of the problems that we'll face on that particular journey, and also indicated to you exactly how an idealized future may appear for us and what the kind of problems are, if we're trying to address or aim to, towards that idealized future. Anyway, I hope you've enjoyed the talk, I hope you found it informative, and I look to hearing your questions. Thanks very much.

Sam Samuel

CTO @ Mobility Business Group, Cisco

Sam Samuel's LinkedIn account

Awesome tech events for

Priority access to all content

Video hallway track

Community chat

Exclusive promotions and giveaways