Posted on

Software Is About Storytelling

Software engineering is more a practice in archeology than it is in building. As an industry, we undervalue storytelling and focus too much on artifacts and tools and deliverables. How many times have you been left scratching your head while looking at a piece of code, system, or process? It’s the story, the legacy left behind by that artifact, that is just as important—if not more—than the artifact itself.

And I don’t mean what’s in the version control history—that’s often useless. I mean the real, human story behind something. Artifacts, whether that’s code or tools or something else entirely, are not just snapshots in time. They’re the result of a series of decisions, discussions, mistakes, corrections, problems, constraints, and so on.  They’re the product of the engineering process, but the problem is they usually don’t capture that process in its entirety. They rarely capture it at all. They commonly end up being nothing but a snapshot in time.

It’s often the sign of an inexperienced engineer when someone looks at something and says, “this is stupid” or “why are they using X instead of Y?” They’re ignoring the context, the fact that circumstances may have been different. There is a story that led up to that point, a reason for why things are the way they are. If you’re lucky, the people involved are still around. Unfortunately, this is not typically the case. And so it’s not necessarily the poor engineer’s fault for wondering these things. Their predecessors haven’t done enough to make that story discoverable and share that context.

I worked at a company that built a homegrown container PaaS on ECS. Doing that today would be insane with the plethora of container solutions available now. “Why aren’t you using Kubernetes?” Well, four years ago when we started, Kubernetes didn’t exist. Even Docker was just in its infancy. And it’s not exactly a flick of a switch to move multiple production environments to a new container runtime, not to mention the politicking with leadership to convince them it’s worth it to not ship any new code for the next quarter as we rearchitect our entire platform. Oh, and now the people behind the original solution are no longer with the company. Good luck! And this is on the timescale of about five years. That’s maybe like one generation of engineers at the company at most—nothing compared to the decades or more software usually lives (an interesting observation is that timescale, I think, is proportional to the size of an organization). Don’t underestimate momentum, but also don’t underestimate changing circumstances, even on a small time horizon.

The point is, stop looking at technology in a vacuum. There are many facets to consider. Likewise, decisions are not made in a vacuum. Part of this is just being an empathetic engineer. The corollary to this is you don’t need to adopt every bleeding-edge tech that comes out to be successful, but the bigger point is software is about storytelling. The question you should be asking is how does your organization tell those stories? Are you deliberate or is it left to tribal knowledge and hearsay? Is it something you truly value and prioritize or simply a byproduct?

Documentation is good, but the trouble with documentation is it’s usually haphazard and stagnant. It’s also usually documentation of how and not why. Documenting intent can go a long way, and understanding the why is a good way to develop empathy. Code survives us. There’s a fantastic talk by Bryan Cantrill on oral tradition in software engineering where he talks about this. People care about intent. Specifically, when you write software, people care what you think. As Bryan puts it, future generations of programmers want to understand your intent so they can abide by it, so we need to tell them what our intent was. We need to broadcast it. Good code comments are an example of this. They give you a narrative of not only what’s going on, but why. When we write software, we write it for future generations, and that’s the most underestimated thing in all of software. Documenting intent also allows you to document your values, and that allows the people who come after you to continue to uphold them.

Storytelling in software is important. Without it, software archeology is simply the study of puzzles created by time and neglect. When an organization doesn’t record its history, it’s bound to repeat the same mistakes. A company’s memory is comprised of its people, but the fact is people churn. Knowing how you got here often helps you with getting to where you want to be. Storytelling is how we transcend generational gaps and the inevitable changing of the old guard to the new guard in a maturing engineering organization. The same is true when we expand that to the entire industry. We’re too memoryless—shipping code and not looking back, discovering everything old that is new again, and simply not appreciating our lineage.

Posted on

FIFO, Exactly-Once, and Other Costs

There’s been a lot of discussion about exactly-once semantics lately, sparked by the recent announcement of support for it in Kafka 0.11. I’ve already written at length about strong guarantees in messaging.

My former coworker Kevin Sookocheff recently made a post about ordered and exactly-once message delivery as it relates to Amazon SQS. It does a good job of illustrating what the trade-offs are, and I want to drive home some points.

In the article, Kevin shows how FIFO delivery is really only meaningful when you have one single-threaded publisher and one single-threaded receiver. Amazon’s FIFO queues allow you to control how restrictive this requirement is by applying ordering on a per-group basis. In other words, we can improve throughput if we can partition work into different ordered groups rather than a single totally ordered group. However, FIFO still effectively limits throughput on a group to a single publisher and single subscriber. If there are multiple publishers, they have to coordinate to ensure ordering is preserved with respect to our application’s semantics. On the subscriber side, things are simpler because SQS will only deliver messages in a group one at a time in order amongst subscribers.

Amazon’s FIFO queues also have an exactly-once processing feature which deduplicates messages within a five-minute window. Note, however, that there are some caveats with this, the obvious one being duplicate delivery outside of the five-minute window. A mission-critical system would have to be designed to account for this possibility. My argument here is if you still have to account for it, what’s the point unless the cost of detecting duplicates is prohibitively expensive? But to be fair, five minutes probably reduces the likelihood enough to the point that it’s useful and in those rare cases where it fails, the duplicate is acceptable.

The more interesting caveat is that FIFO queues do not guarantee exactly-once delivery to consumers (which, as we know, is impossible). Rather, they offer exactly-once processing by guaranteeing that once a message has successfully been acknowledged as processed, it won’t be delivered again. It’s up to applications to ack appropriately. When a message is delivered to a consumer, it remains in the queue until it’s acked. The visibility timeout prevents other consumers from processing it. With FIFO queues, this also means head-of-line blocking for other messages in the same group.

Now, let’s assume a subscriber receives a batch of messages from the queue, processes them—perhaps by storing some results to a database—and then sends an acknowledgement back to SQS which removes them from the queue. It’s entirely possible that during that process step a delay happens—a prolonged GC pause, crash, network delay, whatever. When this happens, the visibility timeout expires and the messages are redelivered and, potentially, reprocessed. What has to happen here is essentially cooperation between the queue and processing step. We might do this by using a database transaction to atomically process and acknowledge the messages. An alternative, yet similar, approach might be to use a write-ahead-log-like strategy whereby the consuming system reads messages from SQS and transactionally stores them in a database for future processing. Once the messages have been committed, the consumer deletes the messages from SQS. In either of these approaches, we’re basically shifting the onus of exactly-once processing onto an ACID-compliant relational database.

Note that this is really how Kafka achieves its exactly-once semantics. It requires end-to-end cooperation for exactly-once to work. State changes in your application need to be committed transactionally with your Kafka offsets.

As Kevin points out, FIFO SQS queues offer exactly-once processing only if 1) publishers never publish duplicate messages wider than five minutes apart and 2) consumers never fail to delete messages they have processed from the queue. Solving either of these problems probably requires some kind of coordination between the application and queue, likely in the form of a database transaction. And if you’re using a database either as the message source, sink, or both, what are exactly-once FIFO queues actually buying you? You’re paying a seemingly large cost in throughput for little perceived value. Your messages are already going through some sort of transactional boundary that provides ordering and uniqueness.

Where I see FIFO and exactly-once semantics being useful is when talking to systems which cannot cooperate with the end-to-end transaction. This might be a legacy service or a system with side effects, such as sending an email. Often in the case of these “distributed workflows”, latency is a lower priority and humans can be involved in various steps. Other use cases might be scheduled integrations with legacy batch processes where throughput is known a priori. These can simply be re-run when errors occur.

When people describe a messaging system with FIFO and exactly-once semantics, they’re usually providing a poor description of a relational database with support for ACID transactions. Providing these semantics in a messaging system likely still involves database transactions, it’s just more complicated. It turns out relational databases are really good at ensuring invariants like exactly-once.

I’ve picked on Kafka a bit in the past, especially with the exactly-once announcement, but my issue is not with Kafka itself. Kafka is a fantastic technology. It’s well-architected, battle-tested, and the team behind it is talented and knows the space well. My issue is more with some of the intangible costs associated with it. The same goes for similar systems (like exactly-once FIFO SQS queues). Beyond just the operational complexity (which Confluent is attempting to tackle with its Kafka-as-a-service), you have to get developer understanding. This is harder than it sounds in any modestly-sized organization. That’s not to say that developers are dumb or incapable of understanding, but the fact is your average developer is simply not thinking about all of the edge cases brought on by operating distributed systems at scale. They see “exactly-once FIFO queues” in SQS or “exactly-once delivery” in Kafka and take it at face value. They don’t read beyond the headline. They don’t look for the caveats. That’s why I took issue with how Kafka claimed to do the impossible with exactly-once delivery when it’s really exactly-once processing or, as I’ve come to call it, “atomic processing.” Henry Robinson put it best when talking about the Kafka announcement:

If I were to rewrite the article, I’d structure it thus: “exactly-once looks like atomic broadcast. Atomic broadcast is impossible. Here’s how exactly-once might fail, and here’s why we think you shouldn’t be worried about it.” That’s a harder argument for users to swallow…

Basically “exactly-once” markets better. It’s something developers can latch onto, but it’s also misleading. I know it’s only a matter of time before people start linking me to the Confluent post saying, “see, exactly-once is easy!” But this is just pain deferral. On the contrary, exactly-once semantics require careful construction of your application, assume a closed, transactional world, and do not support the case where I think people want exactly-once the most: side effects.

Interestingly, one of my chief concerns about Kafka’s implementation was what the difficulty of ensuring end-to-end cooperation would be in practice. Side effects into downstream systems with no support for idempotency or transactions could make it difficult. Jay’s counterpoint to this was that the majority of users are using good old-fashioned relational databases, so all you really need to do is commit your offsets and state changes together. It’s not trivial, but it’s not that much harder than avoiding partial updates on failure if you’re updating multiple tables. This brings us back to two of the original points of contention: why not merely use the database for exactly-once in the first place and what about legacy systems?

That’s not to say exactly-once semantics, as offered in systems like SQS and Kafka, are not useful. I think we just need to be more conscientious of the other costs and encourage developers to more deeply understand the solution space—too much sprinkling on of Kafka or exactly-once or FIFO and not enough thinking about the actual business problem. Too much prescribing of solutions and not enough describing of problems.

My thanks to Kevin Sookocheff and Beau Lyddon for reviewing this post.

Posted on

Are We There Yet: The Go Generics Debate

At GopherCon a couple weeks ago, Russ Cox gave a talk titled The Future of Go, in which he discussed what the Go community might want to change about the language—particularly for the so-called Go 2.0 milestone—and the process for realizing those changes. Part of that process is identifying real-world use cases through experience reports, which turn an abstract problem into a concrete one and help the core team to understand its significance. Also mentioned in the talk, of course, were generics. Over the weekend, Dave Cheney posted Should Go 2.0 support generics? Allow me to add to the noise.

First, I agree with Dave’s point in that the Go team has two choices: implement templated types and parameterized functions (or some equivalent thereof) or don’t and own that decision. Though I think his example of Haskell programmers owning their differences with imperative programming is a bit of a false equivalence. Haskell is firmly in the functional camp, while Go is firmly in the imperative. While there is overlap between the two, I think most programmers understand the difference. Few people are asking Haskell to be more PHP-like, but I bet a lot more are asking Elm to be more Haskell-like. Likewise, lots of people are asking Go to be more Java-like—if only a little. This is why so many are asking for generics, and Dave says it himself: “Mainstream programmers expect some form of templated types because they’re used to it in the other languages they interact with alongside Go.”

But I digress. The point is that the Go team should not pay any lip service to the generics discussion if they are not going to fully commit to addressing the problem. There is plenty of type theory and prior art. It’s not a technical problem, it’s a philosophical one.

That said, if the Go team decides to say “no” to generics and own that decision, I suspect they will never fully put an end to the discussion. The similarities to other mainstream languages are too close, unlike the PHP/Haskell example, and the case for generics too compelling, unlike the composition-over-inheritance decision. The lack of generics in Go has already become a meme at this point.

Ian Lance Taylor’s recent post shows there is a divide within the Go team regarding generics. He tells the story of how the copy() and append() functions came about, noting that they were added only because of the lack of generics.

The point I want to make here is that because we had no way to write a generic Vector type with an Append method, we wound up adding a special purpose language feature to implement it. A language that supported parameterized types with methods would not have required a special built-in function that only works with slices. An append operation makes sense for other sorts of data structures, such as various kinds of linked lists. The built-in append function can not be used for them.

My biggest complaint on the matter at this point in the language’s life is the doublethink. That is, the mere existence of channels, maps, and slices seems like a contradiction to the argument against generics. The experience reports supporting generics are trivial: every time someone instantiates one of these things. The argument is that having a small number of built-in generic types is substantially different than allowing them to be user-defined. Allowing generic APIs to be strewn about will increase the cognitive load on developers. Yet, at the same time, we’re okay with adding new APIs to the standard library that do not provide compile-time type safety by effectively subverting the type system through the use of interface{}. By using interface{}, we instead push the responsibility of type safety onto users to do it at runtime in precarious fashion for something the compiler could be well-equipped to handle. The argument here is what is the greater cognitive load? More generally, it’s where should the responsibility lie? Considering Go’s lineage and its aim at the “ordinary” programmer, I’d argue safety—in this case—should be the language’s responsibility.

However, we haven’t fully addressed the “complex generic APIs strewn about” argument. Generics are not the source of complexity in API design, poorly designed APIs are. For every bad generic API in Java, I’ll show you a good one. In Go, I would argue more complexity comes from the workarounds to the lack of generics—interface{}, reflection, code duplication, code generation, type assertions—than from introducing them into the language, not to mention the performance cost with some of these workarounds. The heap and list packages are some of my least favorite packages in the language, largely due to their use of interface{}.

Another frustration I have with the argument against generics is the anecdotal evidence—”I’ve never experienced a need for generics, so anyone who has must be wrong.” It’s a weird rationalization. It’s like a C programmer arguing against garbage collection to build a web app because free() works fine. Someone pointed out to me what’s actually going on is the Blub paradox, which Paul Graham coined.

Graham considers the hierarchy of programming languages with the example of “Blub”, a hypothetically average language “right in the middle of the abstractness continuum. It is not the most powerful language, but it is more powerful than Cobol or machine language.” It was used by Graham to illustrate a comparison, beyond Turing completeness, of programming language power, and more specifically to illustrate the difficulty of comparing a programming language one knows to one that one does not.

Graham considers a hypothetical Blub programmer. When the programmer looks down the “power continuum”, he considers the lower languages to be less powerful because they miss some feature that a Blub programmer is used to. But when he looks up, he fails to realise that he is looking up: he merely sees “weird languages” with unnecessary features and assumes they are equivalent in power, but with “other hairy stuff thrown in as well”. When Graham considers the point of view of a programmer using a language higher than Blub, he describes that programmer as looking down on Blub and noting its “missing” features from the point of view of the higher language.

Graham describes this as the “Blub paradox” and concludes that “By induction, the only programmers in a position to see all the differences in power between the various languages are those who understand the most powerful one.”

I sympathize with the Go team’s desire to keep the overall surface area of the language small and the complexity low, but I have a hard time reconciling this with the existing built-in generics and continued use of interface{} in the standard library. At the very least, I think Go would be better off with continued use of built-in generics for certain types instead of adding more APIs using interface{} like sync.Map, sync.Pool, and atomic.Value. Certainly, I think the debate is worth having though, but the hero-worship and genetic-fallacy type arguments do not further the discussion. My gut feeling is that Go will eventually have generics. It’s just a question of when and how.

Posted on

You Cannot Have Exactly-Once Delivery Redux

A couple years ago I wrote You Cannot Have Exactly-Once Delivery. It stirred up quite a bit of discussion and was even referenced in a book, which I found rather surprising considering I’m not exactly an academic. Recently, the topic of exactly-once delivery has again become a popular point of discussion, particularly with the release of Kafka 0.11, which introduces support for idempotent producers, transactional writes across multiple partitions, and—wait for it—exactly-once semantics.

Naturally, when this hit Hacker News, I received a lot of messages from people asking me, “what gives?” There’s literally a TechCrunch headline titled, Confluent achieves holy grail of “exactly once” delivery on Kafka messaging service (Jay assures me, they don’t write the headlines). The myth has been disproved!

First, let me say what Confluent has accomplished with Kafka is an impressive achievement and one worth celebrating. They made a monumental effort to implement these semantics, and it paid off. The intention of this post is not to minimize any of that work but to try to clarify a few key points and hopefully cut down on some of the misinformation and noise.

“Exactly-once delivery” is a poor term. The word “delivery” is overloaded. Frankly, I think it’s a marketing word. The better term is “exactly-once processing.” Some call the distinction pedantic, but I think it’s important and there is some nuance. Kafka did not solve the Two Generals Problem. Exactly-once delivery, at the transport level, is impossible. It doesn’t exist in any meaningful way and isn’t all that interesting to talk about. “We have a word for infinite packet delay—outage,” as Jay puts it. That’s why TCP exists, but TCP doesn’t care about your application semantics. And in the end, that’s what’s interesting—application semantics. My problem with “exactly-once delivery” is it assumes too much, which causes a lot of folks to make bad assumptions. “Delivery” is a transport semantic. “Processing” is an application semantic.

All is not lost, however. We can still get correct results by having our application cooperate with the processing pipeline. This is essentially what Kafka does, exactly-once processing, and Confluent makes note of that in the blog post towards the end. What does this mean?

To achieve exactly-once processing semantics, we must have a closed system with end-to-end support for modeling input, output, and processor state as a single, atomic operation. Kafka supports this by providing a new transaction API and idempotent producers. Any state changes in your application need to be made atomically in conjunction with Kafka. You must commit your state changes and offsets together. It requires architecting your application in a specific way. State changes in external systems must be part of the Kafka transaction. Confluent’s goal is to make this as easy as possible by providing the platform around Kafka with its streams and connector APIs. The point here is it’s not just a switch you flip and, magically, messages are delivered exactly once. It requires careful construction, application logic coordination, isolating state change and non-determinism, and maintaining a closed system around Kafka. Applications that use the consumer API still have to do this themselves. As Neha puts it in the post, it’s not “magical pixie dust.” This is the most important part of the post and, if it were up to me, would be at the very top.

Exactly-once processing is an end-to-end guarantee and the application has to be designed to not violate the property as well. If you are using the consumer API, this means ensuring that you commit changes to your application state concordant with your offsets as described here.

Side effects into downstream systems with no support for idempotency or distributed transactions make this really difficult in practice I suspect. The argument is that most people are using relational databases that support transactions, but I think there’s still a reasonably large, non-obvious assumption here. Making your event processing atomic might not be easy in all cases. Moreover, every part in your system needs to participate to ensure end-to-end, exactly-once semantics.

Several other messaging systems like TIBCO EMS and Azure Service Bus have provided similar transactional processing guarantees. Kafka, as I understand it, attempts to make it easier and with less performance overhead. That’s a great accomplishment.

What’s really worth drawing attention to is the effort made by Confluent to deliver a correct solution. Achieving exactly-once processing, in and of itself, is relatively “easy” (I use that word loosely). What’s hard is dealing with the range of failures. The announcement shows they’ve done extensive testing, likely much more than most other systems, and have shown that it works and with minimal performance impact.

Kafka provides exactly-once processing semantics because it’s a closed system. There is still a lot of difficulty in ensuring those semantics are maintained across external services, but Confluent attempts to ameliorate this through APIs and tooling. But that’s just it: it’s not exactly-once semantics in a building block that’s the hard thing, it’s building loosely coupled systems that agree on the state of the world. Nevertheless, there is no holy grail here, just some good ole’ fashioned hard work.

Special thanks to Jay Kreps and Sean T. Allen for their feedback on an early draft of this post. Any inaccuracies or opinions are mine alone.

Posted on

Smart Endpoints, Dumb Pipes

I read an interesting article recently called How do you cut a monolith in half? There are a lot of thoughts in the article that resonate with me and some that I disagree with, prompting this response.

The overall message of the article is don’t use a message broker to break apart a monolith because it’s like a cross between a load balancer and a database, with the disadvantages of both and the advantages of neither. The author argues that message brokers are a popular way to pull apart components over a network because they have low setup cost and provide easy service discovery, but they come at a high operational cost. My response to that is the same advice the author puts forward: it depends.

I think it’s important not to conflate “message broker” and “message queue.” The article uses them interchangeably, but it’s really talking about the latter, which I see as a subset of the former. Queues provide, well, queuing semantics. They try to ensure delivery of a message or, more generally, distribution of work. As the author puts it: “In practice, a message broker is a service that transforms network errors and machine failures into filled disks.” Replace “broker” with “queue” and I agree with this statement. This is really describing systems like RabbitMQ, Amazon SQS, TIBCO EMS, IronMQ, and maybe even Kafka fits into that category.

People are easily seduced by “fat” middleware—systems with more features, more capabilities, more responsibilities—because they think it makes their lives easier, and it might at first. Pushing off more responsibility onto the infrastructure makes the application simpler, but it also makes the infrastructure more complex, more fragile, and more slow. Take exactly-once message delivery, for example. Lots of people want it, but the pursuit of it introduces a host of complexity, overhead (in terms of development, operations, and performance), and risk. The end result is something that, in addition to these things, requires all downstream systems to not introduce duplicates and be mindful about their side effects. That is, everything in the processing pipeline must be exactly-once or nothing is. So typically what you end up with is an approximation of exactly-once delivery. You make big investments to lower the likelihood of duplicates, but you still have to deal with the problem. This might make sense if the cost of having duplicates is high, but that doesn’t seem like the common case. My advice is to always opt for the simple solution. We tend to think of engineering challenges as technical problems when, in reality, they’re often just mindset problems. Usually the technical problems have already been solved if we can just adjust our mindset.

There are a couple things to keep in mind here. The first thing to consider is simply capability lock-in. As you push more and more logic off onto more and more specialized middleware, you make it harder to move off it or change things. The second is what we already hinted at. Even with smart middleware, problems still leak out and you have to handle them at the edge—you’re now being taxed twice. This is essentially the end-to-end argument. Push responsibility to the edges, smart endpoints, dumb pipes, etc. It’s the idea that if you need business-level guarantees, build them into the business layer because the infrastructure doesn’t care about them.

The article suggests for short-lived tasks, use a load balancer because with a queue, you’ll end up building a load balancer along with an ad-hoc RPC system, with extra latency. For long-lived tasks, use a database because with a queue, you’ll be building a lock manager, a database, and a scheduler.

A persistent message queue is not bad in itself, but relying on it for recovery, and by extension, correct behaviour, is fraught with peril.

So why the distinction between message brokers and message queues? The point is not all message brokers need to be large, complicated pieces of infrastructure like most message queues tend to be. This was the reason I gravitated towards NATS while architecting Workiva’s messaging platform and why last month I joined Apcera to work on NATS full time.

When Derek Collison originally wrote NATS it was largely for the reasons stated in the article and for the reasons I talk about frequently. It was out of frustration with the current state of the art. In my opinion, NATS was the first system in the space that really turned the way we did messaging on its head (outside of maybe ZeroMQ). It didn’t provide any strong delivery guarantees, transactions, message persistence, or other responsibilities usually assumed by message brokers (there is a layer that provides some of these things, but it’s not baked into the core technology). Instead, NATS prioritized availability, simplicity, and performance over everything else. A simple technology in a vast sea of complexity (my marketing game is strong).

NATS is no-frills pub/sub. It solves the problem of service discovery and work assignment, assumes no other responsibilities, and gets out of your way. It’s designed to be easy to use, easy to operate, and add minimal latency even at scale so that, unlike many other brokers, it is a good way to integrate your microservices. What makes NATS interesting is what it doesn’t do and what it gains by not doing them. Simplicity is a feature—the ultimate sophistication, according to da Vinci. I call it looking at the negative space.

The article reads:

A protocol is the rules and expectations of participants in a system, and how they are beholden to each other. A protocol defines who takes responsibility for failure.

The problem with message brokers, and queues, is that no-one does.

NATS plays to the strengths of the end-to-end principle. It’s a dumb pipe. Handle failures and retries at the client and NATS will do everything it can to remain available and fast. Don’t rely on fragile guarantees or semantics. Instead, face complexity head-on. The author states what you really want is request/reply, which is one point I disagree on. RPC is a bad abstraction for building distributed systems. Use simple, versatile primitives and embrace asynchrony and messaging.

So yes, be careful about relying on message brokers. How smart should the pipes really be? More to the point, be careful about relying on strong semantics because experience shows few things are guaranteed when working with distributed systems at scale. Err to the side of simple. Make few assumptions of your middleware. Push work out of your infrastructure and to the edges if you care about performance and scalability because nothing is harder to scale (or operate) than slow infrastructure that tries to do too much.