Chapter 26: Quantum Computing – Architectural Implications

Welcome to Last Minute Lecture!

This free chapter overview is designed to help students review and understand key concepts.

These summaries supplement not replaced the original textbook and may not be redistributed or resold.

For complete coverage, always consult the official text.

Welcome to the Deep Dive.

Today we're looking right at the horizon, maybe even a bit beyond it.

Exploring the future of software architecture, specifically chapter 26 on quantum computing.

Yeah, it's a fascinating topic.

The source material kicks off with this great analogy comparing quantum computers today to the right flyer back in 1903.

Right, barely functional, maybe a bit clunky, but clearly signaling something huge is coming.

Exactly.

And the mission for us today is to unpack what that something huge means for architects, especially those building systems now.

Because the timeline is actually shorter than you might think, isn't it?

We're talking practical quantum computers, potentially within what, five to ten years?

That's the projection.

And if you think about enterprise systems, they often have lifespans measured in tens of years.

So systems designed today,

well, they're likely going to encounter this quantum shift midlife.

It's not just about speed, though.

That's a common misconception, right?

It's not just a faster CPU.

Oh, absolutely not.

Quantum computers don't just do classical calculations faster.

They perform computations that classical computers fundamentally cannot do efficiently using principles from quantum physics.

Like that Google announcement from 2019.

That really caught people's attention.

It did.

They claimed their quantum computer did a specific calculation in 200 seconds and their estimate for the most powerful classical supercomputer, something like 10 ,000 years.

10 ,000 years versus 200 seconds.

That's hard to even wrap your head around.

But OK, was that a bit of a cherry pick problem?

I mean, what does that raw power actually mean for the typical systems we build?

That's the critical point.

Quantum computers won't be universally better.

They're basically irrelevant for, say, your standard transaction processing, database lookups, payroll systems.

So no quantum powered smartphones anytime soon?

Highly unlikely.

No QPUs on your desk or in your pocket.

They excel at very specific complex problems, things involving

combinatorics, optimization, areas where classical computers just hit a wall.

OK, so a specialized tool for specialized problems.

And presumably, like with classical computing, we'll eventually get higher level languages and abstractions so architects don't need a PhD in physics.

That's the hope, yes.

Future abstractions should hide a lot of the deep physics.

But understanding the architectural impact now is crucial.

All right, let's dive into the fundamentals then.

The qubit, that's the core unit, right?

It is.

The qubit is the quantum part to the classical bit.

And right now, the state of the art is, well, maybe a few hundred qubits in the best machines.

Still early days.

And architecturally, how does this fit in?

Does the quantum processing unit, the QPU, replace the CPU?

No, think of it more like a specialized coprocessor, similar to how a GPU works today.

The QPU handles the quantum calculations, but it's controlled by and communicates with a classical CPU.

And that communication between the CPU and QPU that happens using regular classical bits.

Yes, the interface is classical.

The quantum weirdness happens inside the QPU.

OK, let's contrast them.

A classical bit is simple.

It's either a zero or a one.

That's it.

Reading it doesn't change it.

What about the qubit?

The qubit is different.

It's defined by probabilities.

Specifically, the probability of measuring a one, the probability of measuring a zero, and a third property called phase.

Probability.

So it's not definitively zero or one until you look.

Exactly.

If there's a non -zero chance of finding either zero or one, we say the qubit is in a state of superposition.

It holds the potential for both outcomes simultaneously.

We hear superposition a lot, but you mentioned measurement.

The catch is huge, architecturally speaking.

Measuring a qubit is destructive.

The moment you observe it, that superposition, that potential collapses.

It instantly becomes either a definite zero or a definite one.

So the act of reading it fundamentally changes it.

Precisely.

The original probabilistic state is gone, replaced by the single outcome you measured.

You can't peak without consequence.

And mathematically, the square of the probability amplitudes for zero and one must always add up to one.

That destructive measurement sounds like it would cause problems for things like, say, making a backup, which leads to the no copying rule.

Absolutely.

You can't simply copy the full quantum state of a qubit, a classical copy involves reading the value and then storing it somewhere else.

But if reading destroys the quantum state...

Then you can only copy the collapsed result, the zero or one you measured, not the original superposition with its probabilities and phase.

The full potential can't be duplicated directly.

This really messes with traditional ideas about data redundancy or sharing state.

Okay, that's a major constraint.

Before we get to how we do move quantum information, what about that third thing you mentioned, the phase?

If it doesn't affect the O1 probabilities, what's it for?

The phase is like another dial you can turn.

It's an angle usually between zero and two pi radians.

Algorithms can manipulate this phase, often to kind of mark or influence qubits during a calculation, even if it doesn't change the final measurement probabilities directly.

It's another tool in the quantum toolbox.

So manipulating these fragile states must require very specific kinds of operations then.

Definitely.

Unlike many classical operations, most quantum operations have to be invertible.

You need to be able to mathematically run them backward and recover the input from the output.

Except for that destructive read operation we just talked about.

Right.

Red is the big exception.

But other common single qubit gates include things like NOT, which flips the amplitudes between zero and one, Z, which adds pi to the phase, and crucially, the Hadamard gate, or 8A.

What does Hadamard do?

Head E is often used to create superposition.

It takes a definite state, like a zero, and puts it into a perfect 50 -50 mix, equal probability of measuring zero or one.

It's a fundamental way to inject quantum potential into the system.

Okay, so we can manipulate single qubits, but the real power, the stuff that makes quantum computing potentially revolutionary, comes from making qubits interact, right?

Things like entanglement.

Exactly.

The primary two qubit operation is called controlled NOT or COD.

It's simple conceptually.

One qubit acts as a control.

If the control qubit is one, it flips the state of the second target qubit.

If the control is zero, it does nothing to the target.

So CNOT is the mechanism.

But what about entanglement itself?

The sources call it strange and wondrous with no classical equivalent.

And it really is.

Entanglement is a unique quantum connection between two or more qubits.

If two qubits are entangled, their fates are linked.

No matter how far apart they are meters, kilometers, theoretically light years, if you measure one, you instantly know the state of the other.

Instantly.

Regardless of distance.

Instantly.

If you measure the first qubit and get a zero, you are guaranteed the entangled partner, wherever it is, will also yield a zero, if measured.

Same for one.

Their outcomes are perfectly correlated, faster than light communication would allow Wow.

Okay.

That sounds useful.

Especially considering we just established you can't copy a qubit state directly because of the no -copying theorem.

Precisely.

So if you can't copy, how do you move quantum information from point A to point B?

The answer is this almost sci -fi sounding thing.

Quantum teleportation.

Teleportation.

Like Star Trek.

Sort of.

But only for quantum spates, not people.

It's the workaround for the no -copy rule.

It allows you to transfer the state of a but it requires destroying the original state at A.

And it relies on that entanglement property?

It does.

You need three qubits involved.

Let's call them C, the payload qubit whose state you want to send, located at A, and two other qubits, A and B, which are entangled with each other, one at location A, one at location B.

Okay, so A and B are entangled across some distance.

Zeeler is also at A.

What happens next?

You perform a joint measurement on the bits of information.

Just two regular bits.

Yep, just two classical bits.

These are then sent over a conventional communication channel like the internet or a radio wave from location A to location B.

Critically, that measurement process at A destroys the original state of C and breaks the entanglement connection for qubit A.

So the original is gone, but B is still over at the destination.

Exactly.

Using the two classical bits received from A, the person at location B can perform a specific operation on their qubit, B.

And this operation perfectly reconstructs the original state of Evgenyso onto qubit B.

The state has been teleported.

That's ingenious.

And the security aspect here seems really interesting.

If someone eavesdrops on the communication channel, what do they get?

They only get those two classical bits, which by themselves tell the eavesdropper absolutely nothing about the complex quantum state that was actually transferred.

It's inherently secure against that kind of interception.

Which has huge implications for future network protocols.

The source mentioned HTTPQ.

Right.

MIST, the National Institute of Standards and Technology, is already looking into quantum resistant cryptographic protocols.

HTTPQ is conceptualized as a potential successor to HTTPS, designed to be secure even against attacks from future quantum computers.

And the push is to get these adopted before the quantum threat fully materializes.

Absolutely.

Because we know that quantum computers pose a serious, almost existential threat to our current security infrastructure.

Especially things based on hashing and prime factorization.

Let's talk about passwords first.

Hashing is used to store them securely, but quantum computers are good at reversing hashes.

They are significantly better, thanks to something called Grover's algorithm.

It doesn't break hashing instantly, but it provides a quadratic speedup.

Quadratic speedup?

What does that mean in practical terms?

Think of it like this.

If a classical computer needs n steps to brute force a hash,

a quantum computer using Grover's algorithm needs roughly the square root of n steps.

So if cracking a hash classically takes, say, a million years, quantum might do it in roughly a thousand years.

If it takes a year classically.

A quantum computer might do it in weeks, maybe less than a month.

Okay, that makes a lot of currently protected password data suddenly look very vulnerable.

It does.

But the even bigger threat is to public key cryptography.

Like RSA, which underpins secure online transactions, banking, communication,

everything.

And that relies on the difficulty of factoring large numbers, specifically the product of two very large prime numbers, PEQ.

Correct.

Classically, factoring PEQ is an incredibly hard problem.

It gets exponentially harder as the numbers get bigger.

But not for a quantum computer.

Not with Shor's algorithm.

Shor's algorithm can factor PEQ incredibly efficiently.

The running time scales roughly with a logarithm of the number of bits in PEQ.

It basically makes the core difficulty assumption of RSA, well, invalid.

So Shor's breaks RSA, Grover's weakens hashing.

This really underscores why architects need to be thinking about post -quantum cryptography now.

Without a doubt.

The transition needs to happen proactively.

Okay, shifting from breaking things to building things.

What about the positive applications, like in science or machine learning?

What's holding those back?

Is it just the number of qubits?

Qubit count and quality are part of it.

But a major missing piece, according to the source, is something called quantum random access memory, or QRAM.

QRAM.

So quantum RAM.

What's special about it?

Conceptually, regular RAM takes a specific address, a classical number, and gives you the data stored there.

QRAM is theorized to take a superposition of addresses as input.

A superposition of addresses.

Yeah, and it would return the superposition of the corresponding data stored at those addresses.

This capability is crucial for many quantum algorithms, especially in machine learning, where you want to process vast data sets simultaneously.

But it doesn't exist yet.

It's purely theoretical right now.

And there's a significant theoretical challenge.

The number of physical components needed seems to scale linearly with the number of memory locations you want to access in superposition.

Retrieving a superposition of a million locations might require something proportional to a million physical devices, which could make it impractical for very large scale QRAM.

So QRAM is a major hurdle for quantum big data.

Assuming we overcome it, or for problems that don't need massive QRAM, what algorithms show promise?

The source mentions matrix inversion.

Yes.

Solving systems of linear equations represented as Axe -Bibo is fundamental in science, engineering, and especially machine learning.

The HHL algorithm, named after Harrow, Hasidim, and Lloyd, offers a potential quantum speedup for this.

But again, there are catches, right?

It's not a simple plug -and -play replacement for classical solvers.

Correct.

The HHL algorithm has some important constraints.

First, you need efficient quantum access to the vector dollar, which brings us back to needing QRAM or some similar mechanism if buys is large.

Okay, QRAM dependency.

What else?

Second, the matrix doll itself needs to have certain properties.

It generally needs to be sparse, meaning most of its entries are zero, and well -conditioned, meaning it's not close to being singular.

And the third constraint sounds tricky, the result itself.

Yeah, the solution vector six dollars doesn't just pop out as a list of numbers.

It appears encoded in the amplitudes of the quantum state in superposition.

You get this sort of quantum representation of the answer.

Extracting the specific numerical values you want often requires additional techniques, like amplitude amplification, which uses that phase property we talked about earlier.

So powerful, but with significant prerequisites and challenges in getting the input in and the output out, it paints a picture of quantum computing being powerful but quite particular.

Very much so.

It's not a Swiss army knife.

Given all these complexities, what are the potential application areas people are actively researching beyond the code breaking we've discussed?

Well, major players like IBM are exploring a wide range.

Cybersecurity is the definite, as we've seen, but also things like discovering new materials for better batteries, drug discovery and development, optimizing complex systems like financial models or traffic flow,

improving weather forecasting, and enhancing AI and machine learning algorithms.

That's quite a list, but how much of that is reality versus, say, hopeful speculation?

The source is pretty clear that outside of cybersecurity, most of these are still in the realm of feverish research or speculation.

They haven't yet led to broadly applicable proven results that outperform classical methods at scale.

We're still very much in that Wright Brothers phase.

Okay, so bringing it back to the architect listening to this.

We're in 1903, technologically speaking, but we expect something like Moore's Law to kick in for qubits eventually, leading to exponential growth.

What should they be doing now?

The advice boils down to a few key things.

First, just pay attention.

Stay updated on the hardware progress qubit counts, error correction, coherence times, and algorithm refinements, especially for things like HHL or quantum simulation if they're relevant to your domain.

Tactical advice one, keep learning.

What's two?

Tactical advice two is about isolation.

Look at your current systems.

Identify the components most likely to be impacted or even completely replaced by quantum capabilities down the line.

Think cryptography modules, complex optimization solvers, maybe parts of your data analysis pipeline.

And isolate them how?

Design them with really clear, stable interfaces.

Treat them like black boxes.

Right.

The goal is to make it easier to swap them out later for a quantum -powered service or library without having to re -architect the entire system.

Build for future replacement.

Good advice.

Isolate to insulate.

And the third piece.

Tactical advice three is focus squarely on security readiness.

Keep a close watch on the development and standardization of post -quantum cryptography.

Understand the migration paths.

You need to know when the tipping point arrives, when your current encryption effectively becomes worthless, and be ready to switch to protocols like the eventual HTTPQ or similar standards.

So pay attention, isolate vulnerable parts, and prepare for the security transition.

It sounds quite defensive though.

Is there an offensive angle, an opportunity here?

Absolutely.

The preparation shouldn't just be about mitigating threats.

Think about the possibilities.

We talked about entanglement, enabling instant correlation regardless of distance.

Imagine designing communication networks or distributed systems that leverage that.

What new architectures become possible if information transfer in some sense becomes instantaneous.

That's a mind -bending thought.

Building systems that exploit rather than just defend against quantum phenomena.

Exactly.

So to recap the key takeaways for First, quantum computers offer speed ups not by being fast or classically, but by using fundamentally different physics, superposition, entanglement, destructive measurement, which unfortunately makes them very good at breaking current encryption.

And second,

architects need to prepare now.

Not necessarily by implementing quantum algorithms today, but by isolating the system components most likely to change, especially security, and closely monitoring the progress towards practical quantum computing and standardized post -quantum protocols like HDTBQ.

It's a proactive stance for a future that might arrive sooner than we think.

And if that idea of instant communication networks sounds a bit far -fetched, well, so did flying machines once upon a time.

A perfect closing thought.

Thank you for joining us for this deep dive into the quantum future of architecture.