> Instead of one row per item with a quantity column, we use one row per sellable unit. An item with 10 units has 10 rows.
> But one row per unit for all inventory would break down at scale—an item with 50,000 units across 10 locations would mean 500,000 rows, and the reserve query would slow as it scans through them. Instead, we maintain a bounded pool of available rows, capped at 1,000 per item/location combination. Reservations consume rows from this pool; a replenishment process refills it from the inventory ledger.
Shouldn't I feel uncomfortable with such approach? It seems to create a backoff (pool) for lowering the chance of having a synchronization issue.
Comes down to type of items, when you have physical inventory the number is limited so more manageable and interestingly enough the problem only applies to physical inventory.
You are just spending some more disk space to avoid synchronization issues. Denormalization for performance is a really common pattern, just that people do not start with it in the first place itself
Can you explain how that works? With the row-per-item I can see how you’d use locking primitives etc easily to deal with multiple concurrent shopping carts claiming available inventory.. but how does your solution solve contention? There’d need to be some “number of items in inventory” row, wouldn’t there be contention on that?
The point of one row per item is that thousands of concurrent shoppers don’t need to block each other as they can each claim as many free rows as they need for themselves?
1. Deduct the reservation from the inventory when the user starts to order, but in the same txn also maintain a separate row for the in progress order flow.
2. If the order flow is aborted or times out have a background process that returns these to the inventory.
That seems simpler than this approach and involves no locking. Though their presented approach is also reasonable, there must be some reason not to choose a simpler flow. It is not that difficult to have a gc service that scales, but may be they didn't want to separate that.
I was investigating Durable Objects (DO) and had Fable walk me through where in my app they might be appropriate. One place had a dependency with billing (where I use a transaction now) and the proposed re-work to allow for concurrent editing with DO looked very much like this, reservations with idempotency keys. And if you add hierarchical allotments then it scales pretty well.
I disagree with the other posters about the bg process, if you have any bg processing already you should be able to handle the few edge cases without too much trouble.
My understanding is: your proposal is not very different from what Shopify is doing except they are tracking 'reserved units' (one per row) and you are proposing tracking 'orders' as the temporary state to then reconcile back with inventory quantities.
Yes, at a high level. It doesn't rely on skip locked, which is not cheap at DB level. DB has to still typically run query and keep going until it finds an unlocked item. Deducting and checking inventory counts are simpler ops inside the DB.
This seems like what triggers are for and how we do similar type things. Update trigger on order does select for update on the inventory and increases/decreases it as appropriate.
I don't think you really need that even. An indexed lookup is fast and you don't need to store a computed quantity generally.
Two concurrent deductions of inventory do contend but only during the actual DB update. That is just normal DB locking for SQL isolation levels. The blog refers to explicit locking by the app, which is where skip locked comes in.
This is absolutely fascinating. I enjoy real life stories like this. I went to a Node meetup in 2013 when Target had just switched to Node from PHP and it was a similar experience to see their metrics and hear their strategy.
not the best design to have 1000 rows for each shop*SKU combination. If a candidate proposed this solution during Shopify's System Design interview, i doubt he would be vetted for Senior+ position.
Instead of having 1000 rows per shop*SKU, why not just have one row per shopping cart*SKU?
That way a single row would represent a single cart, and will hold info of multiple items of the same SKU.
No need a cludge with 1000 rows limit and replenishment process. Instead of dealing with N rows, you always deal with a single row.
per my reading of the article, the protection is only needed for a few seconds, while payment is being processed by the payment system.
so the row is inserted when Payment is initiated, and row is deleted when Payment succeeds
What is oversell protection?
Reserve: When payment starts, we mark items as reserved (a short hold, e.g. several minutes).
Claim: When payment succeeds, we permanently deduct quantity from the inventory ledger (source of truth).
but that system could be easily improved to reserve item when user Adds item to a cart, to prevent scenario when user adds item to a cart, goes through checkout, and after initiating payment gets "soldout error":
1. Let user add item to a cart by default (happy path)
2. Initiate async check in the background for SKU and quantity
2a. The check sums up rows for all SKUs and compares to Inventory table (very cheap check since its done to only active shopping carts)
3. After few seconds the check comes back, and we let user know that item is soldout, before/the moment user goes to Checkout.
Ok, but before inserting you must ensure that inventory is not depleted, which means you need to know the count and you need to lock the row. So you still have contention on that item. Them having a 1k buffer allows not to take a lock on a single row every time, and only do it when buffer is empty
there is no need to lock the row, since you a dealing with a shopping cart, not individual item piece. when you run aggregate functions, lock is no needed, it is actually better to run it with SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED; for aggregation
the check for oversold items is extremely cheap:
with carts as (
select sku, sum(quantity) as reserved
from active_carts
group by sku
),
with warehouse as (
select sku, available_units
from inventory
group by sku
)
select * from carts
inner join warehouse using (sku)
where carts.reserved >= warehouse.available_units
assuming there are indexes on sku field in both, results in efficient single pass and agg over 2 tables
Makes sense... if you are counting something in MySQL and now your counter is in Redis that's already strange
But I guess the point is that even in the MySQL scenario the 'reserved_quantities' is almost like a temporary table so either way is not the 'Real' inventory
Who knows. I wish ant harshly penalized speaking litotically because it’s essentially reward hacking as it can often be read multiple ways.
It’s also annoying as a human because Claude et al rate their own writing very highly, putting human<>LLM interactions at a disadvantage to human->LLM<>LLM interactions.
It's not hard to parse, but it's a dense pair of sentences that say nothing. It just pads the length of the article and gives readers mental fatigue trying to read between the lines to figure out what the point is.
I actually don't think this article was LLM generated but these two sentences suck. I think they were moved from another part of the article without being modified.
First, "the hardest lesson". What lesson? It is out of context. Nobody was talking about lessons before this.
Second, "the bottleneck wasn't what we were measuring and observing". Of course the bottleneck itself wasn't that. They couldn't discover what the bottleneck was using the information in their measurements and observations.
It is a clunky and frankly incorrect passage in an otherwise well written article.
Shopify’s founder and their coo both fund far-right extremism, and its founder thinks only rich people should be able to vote. But anyway, they switched databases.
Yeah it's an awful place to work unless you're a far-right bro. My old director used to use slurs and vape in the office. The founder hires pro gamers with no technical expertise because he thinks they're cool.
Because "slurs" is vague and covers a wide variety of utterances that run the gamut from unprofessional to unemployable and I had a tingle of Spidey sense that OP might have chosen the vague phrasing specifically to inflate the sins of the nameless Shopify director in the mind of the reader.
They were really so proud of that AI image that they just had to tack it on at the end? Did nothing but make the blog post feel like cheap mass produced slop
so this is interesting to me, im in retail i work closely with platforms ive used shopify ive used magento ive used smaller players ive helped implement various pieces of all of them.
and i was excited to get some insight, then i realized that this whole thing was written by AI and im going to guess the idea and implementation were probably very AI driven.
> The solution: SKIP LOCKED
> Core idea: one row per unit, bounded by design
cool, thanks claude.
Now I'm wondering what the engineering culture is even like at shopify.
Here's the thing. I like databases, I think there's a lot of shit in this space that went and smoked a shit ton their own good stuff to come up with these pure event driven designs that lock you into event workflows with no isolation and remove the ability to do broader bulk-functions.. and then do something even stupider and say "all you need for the interface is graphql" and such service/platform doesn't give you any other way to reconcile or do reporting for your org you have to warehouse from graphql.. this is crap. So seeing a headline where shopify says they want to kinda get behind a unified database strat behind the scenes even if it's not necessarily customer facing, like that's good imo. SQL is many decades of relational algebra that makes insane computations acrossed vast sets of data pure magic and one of the best query dml interfaces of all time.
..however i dont even agree with the claim their making here that redis isnt the tech for a reservation system. redis when used correctly feels like an insanely awesome way to do a reservation system, i lurv redis for stuff like that.
I'm just gonna go forward with the assumption that current and future shopify updates are pure vibeslop. I already hate their data interfaces, but compared to other saas offerings i appreciate that they do have bulk-features.
I found Shopify’s post very easy to read, and learned about some features of MySQL. On the other hand, I didn’t get any value from reading your comment. You seem to have a bunch of opinions about how things should be done, but haven’t given any details about how you came to these conclusions.
You should be able to do these increments/decrements in a database at the rate you can write WAL to the disk. But the problem is in a lot of these databases the transaction will hold locks until the WAL hits the disk which causes a massive serialisation problem when you have lots of writes to the same row.
For example if it takes 20ms to write a batch to the WAL then if you do 5 updates to the same row then that is a minimum of 100ms. But without waiting on locks if you can batch all the WAL writes together then this could be just 20ms.
I don’t think holding locks while waiting for WAL is strictly necessary. There is definitely some anomalies that can happen if you don’t wait for WAL to be durable because transactions that don’t write WAL can observe non-durable writes in some situations. So for example conditional updates that don’t perform work. But I assume this can be fixed by making these wait on the commit for dependent transactions to become durable if they are empty. There is also the problem of failing writes that reveal information about non-durable writes which is more tricky. For example you try to insert into a unique index and it fails, but the duplicate was due to a non-durable write that is lost.
Pure reads should be fine when using MVCC because you just show the latest durable version of the DB. I know some other replication systems will run all transactions including reads through the WAL/replicated log in order to not have anomalies.
They do heavily use AI, but you haven’t refuted their point that if inventory is in SQL, storing reservation in a second storage system increases complexity.
> Instead of one row per item with a quantity column, we use one row per sellable unit. An item with 10 units has 10 rows.
> But one row per unit for all inventory would break down at scale—an item with 50,000 units across 10 locations would mean 500,000 rows, and the reserve query would slow as it scans through them. Instead, we maintain a bounded pool of available rows, capped at 1,000 per item/location combination. Reservations consume rows from this pool; a replenishment process refills it from the inventory ledger.
Shouldn't I feel uncomfortable with such approach? It seems to create a backoff (pool) for lowering the chance of having a synchronization issue.
Comes down to type of items, when you have physical inventory the number is limited so more manageable and interestingly enough the problem only applies to physical inventory.
You are just spending some more disk space to avoid synchronization issues. Denormalization for performance is a really common pattern, just that people do not start with it in the first place itself
you should, their design is not the best. There is middle ground between "one row per SKU" and "1000 rows per SKU".
Its called one row per shopping cart*SKU combo.
if two people order 100 and 500 items of the same SKU, respectively, the table should have only two rows: for order1 and order2. Not 600 rows.
Can you explain how that works? With the row-per-item I can see how you’d use locking primitives etc easily to deal with multiple concurrent shopping carts claiming available inventory.. but how does your solution solve contention? There’d need to be some “number of items in inventory” row, wouldn’t there be contention on that?
The point of one row per item is that thousands of concurrent shoppers don’t need to block each other as they can each claim as many free rows as they need for themselves?
Depends on the scale. Most companies don't approach the scale where this matters.
I guess it depends on how the replenishment process works. Unless you're ordering over 1000 of an item, I doubt it would be a problem.
replenishment is an unnecessary cludge that only exists due to poor design. an "algorithmical smell" if you wish
It seems there could be a simpler solution.
1. Deduct the reservation from the inventory when the user starts to order, but in the same txn also maintain a separate row for the in progress order flow. 2. If the order flow is aborted or times out have a background process that returns these to the inventory.
That seems simpler than this approach and involves no locking. Though their presented approach is also reasonable, there must be some reason not to choose a simpler flow. It is not that difficult to have a gc service that scales, but may be they didn't want to separate that.
I was investigating Durable Objects (DO) and had Fable walk me through where in my app they might be appropriate. One place had a dependency with billing (where I use a transaction now) and the proposed re-work to allow for concurrent editing with DO looked very much like this, reservations with idempotency keys. And if you add hierarchical allotments then it scales pretty well.
I disagree with the other posters about the bg process, if you have any bg processing already you should be able to handle the few edge cases without too much trouble.
My understanding is: your proposal is not very different from what Shopify is doing except they are tracking 'reserved units' (one per row) and you are proposing tracking 'orders' as the temporary state to then reconcile back with inventory quantities.
Yes, at a high level. It doesn't rely on skip locked, which is not cheap at DB level. DB has to still typically run query and keep going until it finds an unlocked item. Deducting and checking inventory counts are simpler ops inside the DB.
This seems like what triggers are for and how we do similar type things. Update trigger on order does select for update on the inventory and increases/decreases it as appropriate.
I don't think you really need that even. An indexed lookup is fast and you don't need to store a computed quantity generally.
Can you clarify why this involves no locking? There can still be 2 actors fighting for the same row.
Two concurrent deductions of inventory do contend but only during the actual DB update. That is just normal DB locking for SQL isolation levels. The blog refers to explicit locking by the app, which is where skip locked comes in.
Yes, the point is to spread contention across multiple rows. They also mention this in the beginning of the article
The moment you added a background process you just replaced the complexity.
1. Backgrounds process can back up
2. They need context of the user and need to switch context per user
3. What if they fail, you create some DLQ or another process to handle the failure
4. Who looks on those failure and how do they act
TLDR; there is always a cost
The design in the shoppify post already had a background process for the item replenishment.
now you have two problems. what happens when your reservation system backs up?
This is absolutely fascinating. I enjoy real life stories like this. I went to a Node meetup in 2013 when Target had just switched to Node from PHP and it was a similar experience to see their metrics and hear their strategy.
not the best design to have 1000 rows for each shop*SKU combination. If a candidate proposed this solution during Shopify's System Design interview, i doubt he would be vetted for Senior+ position.
Instead of having 1000 rows per shop*SKU, why not just have one row per shopping cart*SKU?
That way a single row would represent a single cart, and will hold info of multiple items of the same SKU.
No need a cludge with 1000 rows limit and replenishment process. Instead of dealing with N rows, you always deal with a single row.
> Instead of having 1000 rows per shopSKU, why not just have one row per shopping cartSKU?
At what point that row is inserted?
per my reading of the article, the protection is only needed for a few seconds, while payment is being processed by the payment system.
so the row is inserted when Payment is initiated, and row is deleted when Payment succeeds
but that system could be easily improved to reserve item when user Adds item to a cart, to prevent scenario when user adds item to a cart, goes through checkout, and after initiating payment gets "soldout error":Ok, but before inserting you must ensure that inventory is not depleted, which means you need to know the count and you need to lock the row. So you still have contention on that item. Them having a 1k buffer allows not to take a lock on a single row every time, and only do it when buffer is empty
there is no need to lock the row, since you a dealing with a shopping cart, not individual item piece. when you run aggregate functions, lock is no needed, it is actually better to run it with SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED; for aggregation
the check for oversold items is extremely cheap:
assuming there are indexes on sku field in both, results in efficient single pass and agg over 2 tablesMakes sense... if you are counting something in MySQL and now your counter is in Redis that's already strange
But I guess the point is that even in the MySQL scenario the 'reserved_quantities' is almost like a temporary table so either way is not the 'Real' inventory
"But the hardest lesson wasn't about database design. It was discovering that the real bottleneck wasn’t what we were observing and measuring."
But was it load bearing?
Even better.
It's web-scale.
load = bearing
gun = smoking
insight = key
gap = closed
summary = executived
belt = suspended
boot = strapped
It's honestly weird Claude converges on this language because it's incredibly wordy and hard to parse.
One would think semantic density would win out in training.
Who knows. I wish ant harshly penalized speaking litotically because it’s essentially reward hacking as it can often be read multiple ways.
It’s also annoying as a human because Claude et al rate their own writing very highly, putting human<>LLM interactions at a disadvantage to human->LLM<>LLM interactions.
> it's ... hard to parse.
No, it's not. Honestly, the more I see these complaints, I honestly just think people are bad writers and shit readers.
Simply put, if that's hard to parse for you, and you are a native-English speaker, pick up a book for once in your life.
It's not hard to parse, but it's a dense pair of sentences that say nothing. It just pads the length of the article and gives readers mental fatigue trying to read between the lines to figure out what the point is.
I actually don't think this article was LLM generated but these two sentences suck. I think they were moved from another part of the article without being modified.
First, "the hardest lesson". What lesson? It is out of context. Nobody was talking about lessons before this.
Second, "the bottleneck wasn't what we were measuring and observing". Of course the bottleneck itself wasn't that. They couldn't discover what the bottleneck was using the information in their measurements and observations.
It is a clunky and frankly incorrect passage in an otherwise well written article.
It’s fascinating that in order to do this, they had to remove 50% of reads and 33% of transactions from the main DB.
Shopify’s founder and their coo both fund far-right extremism, and its founder thinks only rich people should be able to vote. But anyway, they switched databases.
https://www.techwontsave.us/episode/340_shopifys_leaders_are...
Yeah it's an awful place to work unless you're a far-right bro. My old director used to use slurs and vape in the office. The founder hires pro gamers with no technical expertise because he thinks they're cool.
Don't get shy now. Which slurs?
Why are you trying to get someone to repeat slurs?
Because "slurs" is vague and covers a wide variety of utterances that run the gamut from unprofessional to unemployable and I had a tingle of Spidey sense that OP might have chosen the vague phrasing specifically to inflate the sins of the nameless Shopify director in the mind of the reader.
Which slurs are ok, and which aren't?
Are you implying that vaping is far right?
I think that was part of the “bro” bit, not the far right bit
Sounded to me like they were saying people vape in the office, which would make a bad work environment on top of the far right bros.
The lengths companies will go to avoid running different pieces of software...
It can be easier and cheaper to solve problems via technology changes than operations and people
Now you only need MySQL expertise and maintenance rather than Redis and MySQL
They were really so proud of that AI image that they just had to tack it on at the end? Did nothing but make the blog post feel like cheap mass produced slop
This is Shopify, the leadership is full steam ahead on AI in a big way and they review employee performance based on AI usage.
The blog probably was.shopify was pretty early and publicly all in on using AI for everything
so this is interesting to me, im in retail i work closely with platforms ive used shopify ive used magento ive used smaller players ive helped implement various pieces of all of them.
and i was excited to get some insight, then i realized that this whole thing was written by AI and im going to guess the idea and implementation were probably very AI driven.
> The solution: SKIP LOCKED > Core idea: one row per unit, bounded by design
cool, thanks claude.
Now I'm wondering what the engineering culture is even like at shopify.
Here's the thing. I like databases, I think there's a lot of shit in this space that went and smoked a shit ton their own good stuff to come up with these pure event driven designs that lock you into event workflows with no isolation and remove the ability to do broader bulk-functions.. and then do something even stupider and say "all you need for the interface is graphql" and such service/platform doesn't give you any other way to reconcile or do reporting for your org you have to warehouse from graphql.. this is crap. So seeing a headline where shopify says they want to kinda get behind a unified database strat behind the scenes even if it's not necessarily customer facing, like that's good imo. SQL is many decades of relational algebra that makes insane computations acrossed vast sets of data pure magic and one of the best query dml interfaces of all time.
..however i dont even agree with the claim their making here that redis isnt the tech for a reservation system. redis when used correctly feels like an insanely awesome way to do a reservation system, i lurv redis for stuff like that.
I'm just gonna go forward with the assumption that current and future shopify updates are pure vibeslop. I already hate their data interfaces, but compared to other saas offerings i appreciate that they do have bulk-features.
I found Shopify’s post very easy to read, and learned about some features of MySQL. On the other hand, I didn’t get any value from reading your comment. You seem to have a bunch of opinions about how things should be done, but haven’t given any details about how you came to these conclusions.
You should be able to do these increments/decrements in a database at the rate you can write WAL to the disk. But the problem is in a lot of these databases the transaction will hold locks until the WAL hits the disk which causes a massive serialisation problem when you have lots of writes to the same row.
For example if it takes 20ms to write a batch to the WAL then if you do 5 updates to the same row then that is a minimum of 100ms. But without waiting on locks if you can batch all the WAL writes together then this could be just 20ms.
I don’t think holding locks while waiting for WAL is strictly necessary. There is definitely some anomalies that can happen if you don’t wait for WAL to be durable because transactions that don’t write WAL can observe non-durable writes in some situations. So for example conditional updates that don’t perform work. But I assume this can be fixed by making these wait on the commit for dependent transactions to become durable if they are empty. There is also the problem of failing writes that reveal information about non-durable writes which is more tricky. For example you try to insert into a unique index and it fails, but the duplicate was due to a non-durable write that is lost.
Pure reads should be fine when using MVCC because you just show the latest durable version of the DB. I know some other replication systems will run all transactions including reads through the WAL/replicated log in order to not have anomalies.
They do heavily use AI, but you haven’t refuted their point that if inventory is in SQL, storing reservation in a second storage system increases complexity.
outside the slop, i liked this post that was linked on innodb locking: https://jahfer.com/posts/innodb-locks/