> “WE’RE NOT NETFLIX!” I finally snapped. “Netflix has 500 engineers. We have 4. Netflix has dedicated DevOps teams. We have one guy. Netflix has millions of users. We have 50,000.”
Then
> Lesson 5: The Monolith Isn’t Your Enemy
> A well-structured monolith can:
> Scale to millions of users (Shopify, GitHub, Stack Overflow prove this)
Because Shopify, Github and Stack Overflow have 4 engineers each as well.
It kind of seems real because it reads like the it's written by the kind of person that would make high level arch decisions without even understanding what the f they are doing.
Based on my experience microservices do introduce additional fixed costs compared to monoliths (and these costs can be too expensive for small teams), so everything you've quoted makes complete sense.
https://en.wikipedia.org/wiki/Stack_Overflow
It is NOW that Stack Overflow may have more than one engineer working on it.
> Microservices didn’t scale our startup. They killed it.
...and then at the end,
> We lost 6 months. We lost some good engineers. We burned through money we didn’t have. But we survived.
...So did microservices kill the startup or not?
If you are small and not have scaling problems, it is highly unlikely that you see a real difference between monolith or microservice except on the margin.
But lots of things look off in the article: Billing needed to ... Create the order
What? Billing service is the one creating the orders instead of the opposite?
Monday: A cascading failure took down the entire platform for 4 hours. One service had a memory leak, which caused it to slow down, which caused other services to time out, which caused retries, which brought everything down. In the monolith days, we would’ve just restarted one app. Now we had to debug a distributed system failure.
Finally it claims that Monday their service started to be slow, and already Wednesday the customer threatened to leave them because of the service to be slower. Doesn't look like to be a customer very hooked or needing your service if only after 2 days of issues they already want to leave.
Also, something totally suspicious is that, even if small or moderate size of company you could still have people push some architecture that they prefer, no company with a short few months cash runaway will decide to do a big refactor of the whole architecture if everything was good on the first place and no problem encountered. What will happen in theory is that you will start to face a wall, degrading performances with scale of something like that and then decide that you will have to do something, a rework. And then there will be the debate and decision about monolith, microservice, whatever else...
Was it something in the payment space?
The mistake here was having an architect full stop. The team is too small, a good tech lead can manage to plan a service with 50k MAU (and way beyond) without an architect. The problem with some companies that get millions in seed funding is that they need to spend the money and they do so by adding roles that shouldn't exist at that stage.
Another favourite antipattern: making devops a bottleneck. Don’t over-engineer production, don’t buy abstraction you can’t afford, and educate your colleagues to lower the bus factor.
Dedicated devops that aren’t co-founders are notorious for cv optimizing: working with cool, but time-consuming stuff they don’t yet master, at the cost of delivery-time risk.
But that starts to fall down too any time too many people are talking about software they aren’t responsible for deploying or fixing.
> ...you should be wary of spending too much time on things that customers don’t see
It’s when the invisible stuff becomes a chore, and blocks or slows down releasing value, such as worrying about micro services.
> And ironically? Now that we’re back on a monolith and shipping fast again, we’ve started growing again. Customers are happier. The team is happier.
So Microservices did not kill your startup?
And why did you stop instances of your monolith before the Microservices version was mature and ready???
That's because Medium is a bunch of APIs and (micro) services, not a monolith like it should be.
Heck, it could be plain static HTML because it's just text for crying out loud!
Instead, it uses a GraphQL query through JSON to obtain the text of the article... that it already sent me in HTML.
Total page weight of 17 MB, of which 6.7 MB is some sort of non-media ("text") document or script.
This is user-hostile architecture astronaut madness, and is so totally normal in the modern internet that nobody even bats and eye when text takes appreciable amounts of time to render on a 6 GHz multi-core computer with 1 Gbps fibre Internet connectivity.
Your customers hate this. Your architects love it because it keeps them employed.
You don’t have a market fit and you’re running your dev team like a larger company? What are the odds? Pretty high actually.
- don't blindly jump into a new architecture because it's cool
- choose wisely the size of your services. It's not binary, and often it makes sense to group responsibilities into larger services.
- microservices have some benefits, moduliths (though not mentioned in the article) and monoliths have theirs. They all also have their set of disadvantages.
- etc
But anyway, the key lesson (which does not seem like a conclusion the author made) is:
Don't put a halt to your product/business development to do technician only work.
I.e if you can't make a technical change while still shipping customer value, that change may not be worth it.
There are of course exceptions, but in general you can manage technical debt, archtectural work, bug fixing, performance improvements, dx improvements, etc, while still shipping new features.
Microservices solve a logistical problem. Rob wants to push code every two days. Steve wants to push every three. Thom deals with business who wants to release at whim and preferably within a few hours. Their commissions and bonuses are not reduced by how much chaos they case the engineering team. It’s an open feedback loop.
As you add more employees they start tripping over each other on the differences between trunk and deployed. Thats when splitting into multiple services starts to look attractive. Unfortunately they create their own weather and so if you can use process to delay this point you’re gonna be better off.
Everyone eventually merges code they aren’t 100% sure about. Some people do it all the time. However microservices magnify this because it’s difficult to test changes that cross service boundaries. You think you have it right but unless you can fit the entire system onto one machine, you can’t know. And distributed systems usually don’t concern themselves with whether the whole thing will fit onto a dev laptop.
So then you have code in preprod you are pretty sure will work but aren’t completely sure. Stack enough “pretty sure”s over time and as team sizes grow and you’re gonna have incidents on the regular. Separate deployment reduces the blast radius, but doesn’t eliminate it. Feature toggles reduce it more than an order of magnitude, but that still takes you from problems every week to a couple a year. Which in high SLA environments still makes people cranky.
Those grey loading placeholders for text are called skeleton loaders BTW, polyfills are libraries used to support newer browser APIs in older browsers and not something you can exactly see on a website (without checking the devtools)
You know a team has lost the architectural plot when their answer for all performance problems is more caching. And once you add caching it’s hard to sell any other sort of improvements because the caching poisons the perf analysis.
Their solution took forever because the system was less deterministic than we even knew. They were starting to wrap it up when I went on a tear cleaning up low level code that was nickel and diming us. By the time they launched they were looking at achieving half of the response time improvement they were looking for, in twice the time they estimated to do so. And they cheated. They making two requests about 10% of the time, which made the p50 time into a lie, because two smaller requests pull down the average but not the cost per page load. But I scooped them and made the slow path faster, undercutting another 25% of their perf improvements.
I ended up doing more to improve the Little’s Law situation in three months of working on it half time than they did in two man years. And still nothing changed. They are now owned by a competitor. That I believe shut down almost all of their services.
I'm not sure why Medium does the weird blanking thing but my guess is that it's because it's deciding whether to let you read the article or instead put up a paywall. There are a lot of SPA sites out there, many of which aren't particularly economical with frontend resources, and they generally don't do that unless they're trying to enforce some kind of paywall or similar.
Sure, but a significant motivation for using GraphQL is to stitch together a bunch of microservices into a cohesive API for the front end.
My comment about Medium using microservices was just an informed guess, but a good one. They started migrating from a monolith to microservices back in 2018: https://medium.engineering/microservice-architecture-at-medi...
Is it a coincidence that that's around the time frame that I noticed the Medium web site becoming slower than it used to be?
The first one says we shouldn't follow Netflix's example because it is a massive company with an enormous team. The second one says we should follow the example of these companies instead, while ignoring that they are also a huge company with a massive team.
So the criticism/joke stems from the logical inconsistency between the two. The fact that you stopped with microservices, using a rant about Netflix, while at the same time lauding monoliths, using companies of similar scale as examples, highlights your lack of understanding of using team scale as a reason to pursue either alternative. Dealing with such a person in management is common where they often contradict their own reasoning and pick whatever they fancy at that time. You cannot argue logically when the system changes are not based on objective standards but subjective standards, where you can be wrong for one thing but they can be right for the same thing.
That's why it seems like the person making the decisions is lost in terms of the choices they're making.
- We know that full rewrites are expensive & can kill growing companies, so it's best to start with an architecture that you can keep as you scale
- Common argument for microservices: They scale best, look at Netflix etc.
- Counter argument 1: Netflix has a large team, and microservices add fixed complexity that can kill small teams
- Counter argument 2: Monoliths can also scale (see examples)
That was my initial understanding as someone who has had these discussions before. I don't think I'm adding any arguments, my first point is pretty much universally accepted and known. The author is just assuming a certain level of industry knowledge.
In fact, I don’t think “scale” is ever part of the pitch of micro services. Independent scaling maybe if you have some particular hot spot. But the real pitch for micro services is and always has been about isolation. Isolating failure domains, teams and change management. That’s been the story since the Bezos letter and if the leadership didn’t understand that it’s a leadership skill issue. Not an architectural problem.
So this is a story about bad technical leadership, not a particular architecture. And if anything the initial pitch by the architect is the most technically valid leadership in the story (as poor as it is). They failed to understand the problem space but at least they identified what problem the architecture would solve. The rest of engineering leadership did the classic pointy haired boss thing of not listening and hearing what they wanted. They paid for it.
Or they're just bad at communicating and likely decision-making as well. I would say you're giving the author too much credit to be honest but I get your point. It's a poorly-written article in general imo.
Consider moving to micro-services only AFTER reasonable algorithms on commodity bare metal show real capacity limits. There's still higher spec bare metal to carry said designs through a refactor / expansion based on where the performance bottlenecks are. Even absent literal micro-services there's still partitioning / sharding which can spread out many of the pain points.
Netflix also didn't appear into existence with an army of engineers. It also scaled from a few engineers to what it is today. Which means you can scale using both depending on your setup. That cannot be the reason why you pick one arch over another. The reason has to do with your own setup, your company, and the application's specific context, all of which the author is missing.
Broadly, it feels like decision making without context/understand the why behind the decisions.
The specific comment has nothing to do with how Github or Stack Overflow scaled etc.
In my daily job I see an effort to bring distributed system into a more monolith one, because it is just easier to debug.
I have explained it enough, not sure what you're missing at this point.
> not sure what you're missing at this point.
Everyone, including me, are making some assumptions when we are writing something. When it is technical, it is easy to verify - obtain source code, run it, check results. When it is somewhat managerial, it is much harder to verify.
For example, original post emphasizing Stack Overflow being scaling monolith (pun intended) may refer to the point of time when SO were run by basically one man, yet scaling.
You dismiss it, that's okay. You do not answer my or OP points, that's okay too.
Our readers are smart enough to judge OP (and ours) points on their merits.
If we go solely by that criterion, the upvotes on the original comment do prove that people are smart enough to judge the points on their merits. It was good talking to you and to a much wider audience through you.