> For loads with over 20k rows/s and people pushing changes, you may need a full-time person to handle the cluster and take a look at the crazy queries people are going to write.
I think this was a benefit of DBA culture in previous eras. Not that the DBAs were specifically necessary to write good queries (often they'd need to work with application teams to guide them towards schemas/behavior that worked well) or to maintain the database (managed DB offerings obsolete a lot of this work), but because they functioned as gatekeepers and rate-limiters of what queries and schemas could exist.
In that mode, DBAs functioned a bit like a human/process version of a thin microservice wrapping database access functionality. A big benefit was that the rate of change of queries/schema changes/access patterns was controlled and had a higher probability of being reviewed and thought about by humans before it went live. This also resulted in an increased end-database-user culture of trying to make existing schemas/query patterns work before jumping straight to bespoke access patterns. That culture's not what you want as e.g. a startup or pro-rapid-big-refactors shop, but it is what you want when your DB reliability needs or query rate/dataset size are high.
I don't think it's a given that a gatekeeper team is worth the overhead and cost; that's situational. I do think that the code version of that team (aforementioned microservice that wraps DB accesses/schema changes and nothing else) is usually not worth the cost. In my experience, that pretty much always reduces reliability and free performance gains that come from using direct DB clients from user code.
Most startups can just scale your traditional separation of compute & storage here though. You’d be shocked how well duckdb against s3 scales for 99.9% of use cases
It's one thing to be able to scale database compute/storage; it's another thing to be able to partition it. It's extremely common for bad queries/access patterns to cause noisy-neighbor effects on other simultaneous accesses to the database, to the extreme of knocking the whole database over with timeouts/OOMs/etc.
Scaling out DB compute can only help with that to a (expensive) point; eventually, you end up wanting to either prevent the bad queries from being added to the system (DBA culture) or ensure that the bad query runs on database infrastructure that doesn't affect other queries. That's why partitioning DB compute (and storage: noisy-neighbor effects from a bad query at the storage layer don't require storage to be running e.g. a BookKeeper or whatever on a server; they can manifest as hot S3 keys or cloud object/block store rate limiting) is a necessary capability if your plan for dealing with a culture of "anyone can add any access pattern they want" is to scale the DB.
Iceberg & Delta Lake do partition the data. You can add as many DuckDB servers processing them as you want, each app can have as many as you need. You pay Amazon for all the I/O and let them worry about it scaling it.
I'm not saying there is no point of contention - depending upon your write patterns you have practical limits, but for reading data that is already stored you pretty much just pay more to do more and the rate of that cost is close to linear.
I'm sure there is a scale at which that pattern breaks, and I'm sure 99.9% of people reading this will never operate a system at that scale.
Iceberg addresses the hot s3 key prefixes. It places entropy in the prefix path per object so reads/writes are randomized over the bucket key prefix space.
It’s much less of an issue. Also, 99%+ of Aws customers likely don’t have enough QPS to hit the per prefix limits anyway
Instead we're... listen to this... we're going to take a software developer right. Just a normal developer right. We're going to make them be the database expert right. And the cloud expert. And we're going to put them on call. We're going to have them debug linux logs, and optimize our AWS costs. They'll be there for client escalation work. And big sales calls. From time to time we'll even have them do front end work.
Eerily accurate how it works these days, I wish you weren't correct. I met a DBA wizard (he looked like the creator of c++) at a banking IT dept and this guy intuitively sensed what you needed and how its done.
And don't forget contract management with the supplier, L1-L2-L3 support, all in one, and integrating as the supplier is useless and your contract is shit.
Oh and you will do also business analysis with the business as claude is too complex for them (read any version of the nocode initiative).
I think a DBA/ops/infrastructure person as an imposed bottleneck is a useful capability in some environments.
But I won't follow you as far as "expecting developers to have expertise in how and where their software runs is unreasonable".
Like, yeah, it sucks that added DevOps responsibilities etc. don't come with adjusted compensation/time allocation expectations. I'm with you there.
But it's simultaneously true that a ton of "just regular developer" people are significant liabilities because they don't understand anything about the environment where their software runs. That liability manifests operationally (if someone's just running integration tests on Windows for their Java business logic changes and don't have any familiarity with e.g. the Linux, container, or cloud environments where their code runs, they're going to be useless when their code breaks in production and operations staff needs context), and it also makes them less effective when writing code--this culture of "developers should just live in business logic and not have to context-switch or fill their brains with other levels of the stack" is what leads to full table scans, lack of awareness of memory use, N+1 query hell, looping microservice dependencies, misunderstanding of what HTTP fields are set on requests that are mutated by load balancers, mistaken assumptions about how many instances of code can run and what concurrency/thread/coroutine behaviors are present, and so on. Those are very common problems, and it's incumbent on developers in every specialty to gain familiarity with how and where their code runs in order to write and maintain that code effectively.
If your code runs on Linux in Kubernetes, all of your developers should know how to read Linux system logs, check database sessions/queries issued by parts of the application, ls/grep/cat/strace/ps their way around, interpret k8s/application dashboards, check application logs both in log storage and as they're emitted from a process, exec into a container, restart pods, check deployment liveness, etc. Even if they don't have permission to do those things in production.
That was true in 2005 when they deployed their code to IIS on Windows Server/MSSQL, too--just with different operational specifics.
That's a low bar that's often unmet, and all sorts of teams suffer from that failure. Those skills can be trained, kept up to date, and hired for; I don't think there's a great excuse for not expecting them.
> But I won't follow you as far as "expecting developers to have expertise in how and where their software runs is unreasonable".
In my professional experience, management tends to treat this as a dichotomy rather than a range. And its lead to situations where we're choosing to route every decision through a small number of 'experts' and expecting that to scale, or have people make very bad decisions about things like indices, data layout, etc.
I align with you on this. If you are using some sort of system, there is a sort-of minimum bar of expertise one should acquire. When using analytical platforms, having a basic understanding of how data is stored and the levers it gives you to optimize data processing should be standard.
Some examples:
MySQL - I would expect developers to be able to come up with a list of columns to index and why, understand why UUIDv4 is a bad choice for primary keys, know how to write performant SQL by hand and understand when their ORM is going to do the wrong thing, etc.
Redshift - Understand dist and sort keys and have some mental models on how data shuffles and is pruned based on each choice.
Iceberg + Spark/Trino - Understand that data tends to be stored/loaded as batch files. Understand how deletes work. Understand partitions and sort as options. Understand generally how bloom filters work and when you should decide to place a bloom filter. Understand why UUIDv7 is better than UUIDv4 when its part of a partition key.
Apologies my comment was very tongue and cheek. I understand what you're getting at with regards to dedicated DB style teams being a potential bottleneck and not effective for every organization, as for example most orgs get by fine with a simple RDS instance with minimal management and optimization.
What I'm calling out practically is a de-specialization in roles and the trade-offs it imposes.
For example, the "jack of all trades master of one" that was already propagating under the guise of "Full stack engineer" or "DevOps" is further exasperated when those consolidated roles are also delegating critical knowledge acquisition to LLMs.
Further, as we consolidate roles and delegate knowledge acquisition to llms, we're also paying people less, giving less job security, lowering the general quality of life not just for developers, but for operations staff and users, and generally promoting total consolidation within big tech.
I usually see full-stack and DevOps roles as being a creation of the startup and early growth phases of a company. In those phases, the market and product definition is changing rapidly. Whatever you lose in reduced technical depth of these engineers, you gain in reduced communication costs, faster ability to make changes, and additional context and ability to spot efficiencies because they understand the full stack. And these companies are usually default-dead anyway, so if they screw things up and cause an outage, it's no worse than if they build the wrong product and have no users.
I've never seen a DevOps role in Big Tech (I've ping-ponged back and forth between the startup and FANG worlds a few times), and usually a full-stack engineer is someone very senior who designed a lot of the product when it was young and stuck with it for a decade. Instead there's usually a clear SRE/SWE split, and hyper-specialization, enough that someone might actually become the expert on one particular screen of a mobile app.
If you're frustrated with the de-specialization of roles, it's entirely possible you're working at companies at the wrong stage of their lifecycle. Move to a big company and the struggle is to avoid becoming pigeon-holed into one very specific task (and then laid off when the structure of the product changes so that task is no longer necessary). Such a career change can be challenging at first because your career can get specialized to "working for startups and early growth companies", and then big companies won't even look at your resume. But if you try enough places and say in the interview that you're looking for a place where you can really specialize and develop expertise, you can often find a manager desperate enough to take a chance.
LLMs are another issue. I'm worried that they're going to make the economy very unstable because the whole reason a firm exists is to capture institutional knowledge and the LLM captures that institutional knowledge in the model, where a.) it's not specific to a differentiated business and b.) it's subject to model-collapse issues. Could easily see the whole technological world collapse through lack of expertise, but if that happens it's a bigger problem that you can only fix by developing the expertise yourself.
What's worked really well for us is not having the DBAs as gate keepers, but rather having them as guardians and experts for devs to rely on.
Our DBAs are the first line of defense when a DB starts to struggle. They crack the database open, see what's going on, and troubleshoot and sometimes fix the problems. If it's just a simple stored proc change or table change, our DBAs are empowered to go in and fix them. If there's a code change needed, they work with the devs to fix the code.
But by and large the devs still write most of the queries and tables. We as devs get DB training, but we also have a very open door policy with the DBAs so the devs can lean on their expertise and ask advice before sending stuff out. Our devs get a much better sense of "This seems like it could be expensive, let's talk to the DBAs and see what they'd suggest".
It's worked great.
I think any company would benefit from this sort of subject matter expert employment.
No apology needed; I understand what you're getting at, and I broadly agree. It's a spectrum between "extremely easy-to-hire people that operate in such a narrow niche that they're an operational liability with limited capabilities" and "expect everyone to be an expert at every level of the stack". The right point on that spectrum is different depending on context, but I do think that a majority of software shops would be better served by moving their required skillset more towards the generalist end of that spectrum, because the default is often far too niche (driven by poor tradeoffs and short-termism in service of growth/hiring, usually).
I wanna re-emphasize that this is not a new problem. It's not because of DevOps culture or cloud complexity or scale or whatever. Very limited-specialty people were always operational liabilities and had limited positive impact on feature delivery once you accounted for the help they needed to do anything that spanned multiple levels of the stack. There are just more engineers working on more systems with tighter timeline expectations now, so it seems like the complexity incumbent on the engineering role went up in general. It didn't (it went up in some situations and down in some situations), we just started noticing operational pain more often.
I definitely do agree that there's widespread ignorance of the velocity and difficulty-of-work tradeoffs that arise from requiring a wider range of specialties from engineers, and a similarly widespread failure to adjust compensation and timeline expectations accordingly.
Why are you complaining? You should be grateful of the learning opportunity not everyone gets to have. Sure, we pay you peanuts for it. But, it's ultimately for your own good. Don't mind my yacht and Ferrari, though.
It's defensive language for sure, i don't know how much it adds of protection in reality, but i nonethelesss sympathize with the author if he feels the need to protect himself that way or signaling the risk he takes.
As an aside, I was stuck when turning on the Fulham v Crystal Palace game last week to find that Fulham have ClickHouse on their shirts this year, and Palace have Temporal AI. Talk about my worlds colliding.
I'm aware; unfortunately that then leads to "must have a vendor approval" and a lot more process. If it were RDS, then it'd just be provision and done.
Do any two teams actually operate it in anything like the same way though?
What I saw of it, especially some years ago, was it was highly particular, and everyone had their own odd habits built around running it, ingestion, querying, everything, to the point I suspect there are a non trivial number of companies using it where it is actually the core operational expertise of the company, despite them all appearing to be in totally different domains.
Using BYOC management reduces the costs significantly. The big cost in analytic SaaS offerings is generally compute, which vendors mark up significantly. (They keep margins low on storage.)
Disclosure: My company Altinity offers BYOC management of ClickHouse.
Where does ClickHouse fit between ElasticSearch, Pinot, TrinoDB, or just plain Spark? These are very different tools but I’m curious to know if any one has already compared them and can share some thoughts regarding their maintainability, QPS, latency, etc...
The relationship depends on the use case and is not linear. For web analytics and log management ClickHouse is a great replacement for ElasticSearch, for example.
I would put it closest to Pinot though pinot makes different architectural choices about storage as a difference that stands out to me. Then i’d rank trino as a second closest in only some ways. Spark the furthest for sure.
I would like to see some sources on your claim about it being Russian. It is incorporated in the US with most developers in Amsterdam.
After the invasion of Ukraine they stayed silent for a while but that was because they needed to allow there developers to get out of Russia. Many of them are Ukrainians including the CTO and founder. As soon as it was safe for there team they took a very firm stance against Russia with Ukrainian flags on the website and written statement from the team.
I am not aware of any Russian influence currently.
When replaying customer queries against the next version, how do you compare results for queries using now() or approximate aggregates? Curious how you separate expected differences from actual regressions.
> A quick note about HTTP: ClickHouse® offers a TCP connector with a native protocol, but we don't use it. It does not offer many advantages for the type of application we build
This needs more elaboration. One of the major goals of running a ClickHouse cluster is to provide low-latency queries; a persistent TCP connection removes the need to re-establish a new connection for each query and thus reduces overall latency in line with CH goals. So I really didn't understand this.
> ClickHouse® open source faces a significant challenge: limited support for cloud storage. Modern OLAP databases and data systems should leverage cloud storage for cost efficiency and independent scaling of compute and storage resources. Snowflake established this standard over a decade ago, and ClickHouse® (open source) lags behind
ClickHouse writing to NVMEs is exactly how they provide their latency and performance advtanges. Writing and reading to object buckets is fundamentally slower with multiple network hops to reach what is, in this architectural context, a storage server for your storage server. If you really need far more storage, and are willing to sacrifice query latency to get it... why not architect for one of the OLAP databases, like Snowflake, where that was part of their architecture from day one?
> Because you are testing your analytics queries, right?
No? Half the point of an OLAP database is to let users write their own queries. If we knew the queries ahead of time, we probably wouldn't need an OLAP database, and instead use a less-flexible streaming architecture storing intermediate calculations so as not to need to pay for petabyte-scale storage. The expected value from paying for all of that storage is to support not knowing which queries will be written by users.
> Every single company handling ClickHouse® struggles with ingestion... Backpressure mechanism: Some people put Kafka before ClickHouse®. This does the job
The whole trade-off that you make with column-store databases like ClickHouse (instead of row-store databases like Postgres) is that inserts are slow for column stores (whereas they are fast for row stores). Inserts happen slowly, asynchronously, in the background. It is the price you pay for fast analytics queries. This is why OLAP databases have a latency lag and do not show real-time results. This is why stores like Kafka are usually a good fit, you let Kafka hold onto new data until batch insertions can catch up. If you do need real-time queries, you don't write to an OLAP directly; you write to a stateful frontend that answers the query itself, then streams out historical data from the OLAP that was successfully written there. And the first thing you do in a "I want to have my cake, and eat it too, and yes I'm willing to pay for the privilege" architecture like that is... to keep the persistent TCP connections, because that's really low-hanging fruit.
I think this was a benefit of DBA culture in previous eras. Not that the DBAs were specifically necessary to write good queries (often they'd need to work with application teams to guide them towards schemas/behavior that worked well) or to maintain the database (managed DB offerings obsolete a lot of this work), but because they functioned as gatekeepers and rate-limiters of what queries and schemas could exist.
In that mode, DBAs functioned a bit like a human/process version of a thin microservice wrapping database access functionality. A big benefit was that the rate of change of queries/schema changes/access patterns was controlled and had a higher probability of being reviewed and thought about by humans before it went live. This also resulted in an increased end-database-user culture of trying to make existing schemas/query patterns work before jumping straight to bespoke access patterns. That culture's not what you want as e.g. a startup or pro-rapid-big-refactors shop, but it is what you want when your DB reliability needs or query rate/dataset size are high.
I don't think it's a given that a gatekeeper team is worth the overhead and cost; that's situational. I do think that the code version of that team (aforementioned microservice that wraps DB accesses/schema changes and nothing else) is usually not worth the cost. In my experience, that pretty much always reduces reliability and free performance gains that come from using direct DB clients from user code.
Scaling out DB compute can only help with that to a (expensive) point; eventually, you end up wanting to either prevent the bad queries from being added to the system (DBA culture) or ensure that the bad query runs on database infrastructure that doesn't affect other queries. That's why partitioning DB compute (and storage: noisy-neighbor effects from a bad query at the storage layer don't require storage to be running e.g. a BookKeeper or whatever on a server; they can manifest as hot S3 keys or cloud object/block store rate limiting) is a necessary capability if your plan for dealing with a culture of "anyone can add any access pattern they want" is to scale the DB.
I'm not saying there is no point of contention - depending upon your write patterns you have practical limits, but for reading data that is already stored you pretty much just pay more to do more and the rate of that cost is close to linear.
I'm sure there is a scale at which that pattern breaks, and I'm sure 99.9% of people reading this will never operate a system at that scale.
It’s much less of an issue. Also, 99%+ of Aws customers likely don’t have enough QPS to hit the per prefix limits anyway
And get this. We pay them the exact same.
Oh and you will do also business analysis with the business as claude is too complex for them (read any version of the nocode initiative).
But I won't follow you as far as "expecting developers to have expertise in how and where their software runs is unreasonable".
Like, yeah, it sucks that added DevOps responsibilities etc. don't come with adjusted compensation/time allocation expectations. I'm with you there.
But it's simultaneously true that a ton of "just regular developer" people are significant liabilities because they don't understand anything about the environment where their software runs. That liability manifests operationally (if someone's just running integration tests on Windows for their Java business logic changes and don't have any familiarity with e.g. the Linux, container, or cloud environments where their code runs, they're going to be useless when their code breaks in production and operations staff needs context), and it also makes them less effective when writing code--this culture of "developers should just live in business logic and not have to context-switch or fill their brains with other levels of the stack" is what leads to full table scans, lack of awareness of memory use, N+1 query hell, looping microservice dependencies, misunderstanding of what HTTP fields are set on requests that are mutated by load balancers, mistaken assumptions about how many instances of code can run and what concurrency/thread/coroutine behaviors are present, and so on. Those are very common problems, and it's incumbent on developers in every specialty to gain familiarity with how and where their code runs in order to write and maintain that code effectively.
If your code runs on Linux in Kubernetes, all of your developers should know how to read Linux system logs, check database sessions/queries issued by parts of the application, ls/grep/cat/strace/ps their way around, interpret k8s/application dashboards, check application logs both in log storage and as they're emitted from a process, exec into a container, restart pods, check deployment liveness, etc. Even if they don't have permission to do those things in production.
That was true in 2005 when they deployed their code to IIS on Windows Server/MSSQL, too--just with different operational specifics.
That's a low bar that's often unmet, and all sorts of teams suffer from that failure. Those skills can be trained, kept up to date, and hired for; I don't think there's a great excuse for not expecting them.
In my professional experience, management tends to treat this as a dichotomy rather than a range. And its lead to situations where we're choosing to route every decision through a small number of 'experts' and expecting that to scale, or have people make very bad decisions about things like indices, data layout, etc.
I align with you on this. If you are using some sort of system, there is a sort-of minimum bar of expertise one should acquire. When using analytical platforms, having a basic understanding of how data is stored and the levers it gives you to optimize data processing should be standard.
Some examples:
MySQL - I would expect developers to be able to come up with a list of columns to index and why, understand why UUIDv4 is a bad choice for primary keys, know how to write performant SQL by hand and understand when their ORM is going to do the wrong thing, etc.
Redshift - Understand dist and sort keys and have some mental models on how data shuffles and is pruned based on each choice.
Iceberg + Spark/Trino - Understand that data tends to be stored/loaded as batch files. Understand how deletes work. Understand partitions and sort as options. Understand generally how bloom filters work and when you should decide to place a bloom filter. Understand why UUIDv7 is better than UUIDv4 when its part of a partition key.
What I'm calling out practically is a de-specialization in roles and the trade-offs it imposes.
For example, the "jack of all trades master of one" that was already propagating under the guise of "Full stack engineer" or "DevOps" is further exasperated when those consolidated roles are also delegating critical knowledge acquisition to LLMs.
Further, as we consolidate roles and delegate knowledge acquisition to llms, we're also paying people less, giving less job security, lowering the general quality of life not just for developers, but for operations staff and users, and generally promoting total consolidation within big tech.
I've never seen a DevOps role in Big Tech (I've ping-ponged back and forth between the startup and FANG worlds a few times), and usually a full-stack engineer is someone very senior who designed a lot of the product when it was young and stuck with it for a decade. Instead there's usually a clear SRE/SWE split, and hyper-specialization, enough that someone might actually become the expert on one particular screen of a mobile app.
If you're frustrated with the de-specialization of roles, it's entirely possible you're working at companies at the wrong stage of their lifecycle. Move to a big company and the struggle is to avoid becoming pigeon-holed into one very specific task (and then laid off when the structure of the product changes so that task is no longer necessary). Such a career change can be challenging at first because your career can get specialized to "working for startups and early growth companies", and then big companies won't even look at your resume. But if you try enough places and say in the interview that you're looking for a place where you can really specialize and develop expertise, you can often find a manager desperate enough to take a chance.
LLMs are another issue. I'm worried that they're going to make the economy very unstable because the whole reason a firm exists is to capture institutional knowledge and the LLM captures that institutional knowledge in the model, where a.) it's not specific to a differentiated business and b.) it's subject to model-collapse issues. Could easily see the whole technological world collapse through lack of expertise, but if that happens it's a bigger problem that you can only fix by developing the expertise yourself.
Our DBAs are the first line of defense when a DB starts to struggle. They crack the database open, see what's going on, and troubleshoot and sometimes fix the problems. If it's just a simple stored proc change or table change, our DBAs are empowered to go in and fix them. If there's a code change needed, they work with the devs to fix the code.
But by and large the devs still write most of the queries and tables. We as devs get DB training, but we also have a very open door policy with the DBAs so the devs can lean on their expertise and ask advice before sending stuff out. Our devs get a much better sense of "This seems like it could be expensive, let's talk to the DBAs and see what they'd suggest".
It's worked great.
I think any company would benefit from this sort of subject matter expert employment.
I wanna re-emphasize that this is not a new problem. It's not because of DevOps culture or cloud complexity or scale or whatever. Very limited-specialty people were always operational liabilities and had limited positive impact on feature delivery once you accounted for the help they needed to do anything that spanned multiple levels of the stack. There are just more engineers working on more systems with tighter timeline expectations now, so it seems like the complexity incumbent on the engineering role went up in general. It didn't (it went up in some situations and down in some situations), we just started noticing operational pain more often.
I definitely do agree that there's widespread ignorance of the velocity and difficulty-of-work tradeoffs that arise from requiring a wider range of specialties from engineers, and a similarly widespread failure to adjust compensation and timeline expectations accordingly.
Why are you complaining? You should be grateful of the learning opportunity not everyone gets to have. Sure, we pay you peanuts for it. But, it's ultimately for your own good. Don't mind my yacht and Ferrari, though.
Maybe I'm old, and I am, but I just can't get past this point with such annoying writing. Like if you actually spoke like this people would hate you.
But seriously Clickhouse does love disk space.
What I saw of it, especially some years ago, was it was highly particular, and everyone had their own odd habits built around running it, ingestion, querying, everything, to the point I suspect there are a non trivial number of companies using it where it is actually the core operational expertise of the company, despite them all appearing to be in totally different domains.
But given the majority of use-cases of CH, AWS can be quite expensive.
Disclosure: My company Altinity offers BYOC management of ClickHouse.
Very true. Reading about "too many parts" gave me flashbacks.
(previously owned ingestion into CH at PostHog, no longer)
:oof-1: CH needs more disk to keep up with merges.
We ran CH way too lean in those days.
(note: the first part was originally published April 2025 according to the date tooltip)
It is very easy to ingest data into CH. We connected exchange topics and it just worked with zero code.
But the thing with CH is that it is pretty much a Russian product so you should not use it for production anymore.
After the invasion of Ukraine they stayed silent for a while but that was because they needed to allow there developers to get out of Russia. Many of them are Ukrainians including the CTO and founder. As soon as it was safe for there team they took a very firm stance against Russia with Ukrainian flags on the website and written statement from the team.
I am not aware of any Russian influence currently.
> A quick note about HTTP: ClickHouse® offers a TCP connector with a native protocol, but we don't use it. It does not offer many advantages for the type of application we build
This needs more elaboration. One of the major goals of running a ClickHouse cluster is to provide low-latency queries; a persistent TCP connection removes the need to re-establish a new connection for each query and thus reduces overall latency in line with CH goals. So I really didn't understand this.
> ClickHouse® open source faces a significant challenge: limited support for cloud storage. Modern OLAP databases and data systems should leverage cloud storage for cost efficiency and independent scaling of compute and storage resources. Snowflake established this standard over a decade ago, and ClickHouse® (open source) lags behind
ClickHouse writing to NVMEs is exactly how they provide their latency and performance advtanges. Writing and reading to object buckets is fundamentally slower with multiple network hops to reach what is, in this architectural context, a storage server for your storage server. If you really need far more storage, and are willing to sacrifice query latency to get it... why not architect for one of the OLAP databases, like Snowflake, where that was part of their architecture from day one?
> Because you are testing your analytics queries, right?
No? Half the point of an OLAP database is to let users write their own queries. If we knew the queries ahead of time, we probably wouldn't need an OLAP database, and instead use a less-flexible streaming architecture storing intermediate calculations so as not to need to pay for petabyte-scale storage. The expected value from paying for all of that storage is to support not knowing which queries will be written by users.
> Every single company handling ClickHouse® struggles with ingestion... Backpressure mechanism: Some people put Kafka before ClickHouse®. This does the job
The whole trade-off that you make with column-store databases like ClickHouse (instead of row-store databases like Postgres) is that inserts are slow for column stores (whereas they are fast for row stores). Inserts happen slowly, asynchronously, in the background. It is the price you pay for fast analytics queries. This is why OLAP databases have a latency lag and do not show real-time results. This is why stores like Kafka are usually a good fit, you let Kafka hold onto new data until batch insertions can catch up. If you do need real-time queries, you don't write to an OLAP directly; you write to a stateful frontend that answers the query itself, then streams out historical data from the OLAP that was successfully written there. And the first thing you do in a "I want to have my cake, and eat it too, and yes I'm willing to pay for the privilege" architecture like that is... to keep the persistent TCP connections, because that's really low-hanging fruit.