I have mixed feelings reading this article. There was no requirement around code refactoring, however multiple solutions were axed because of “how much of our application I was going to need to refactor”. Maybe have that as a hard requirement then and don’t invest time learning tech that doesn’t meet this requirement? Also, was that a joke about the divorce? Either way bad taste and red flag on the personality side for me.
Building the world’s fastest website analytics (2021)
11–20 of 40 posts
Re: Building the world’s fastest website analytics (2021)
#12At some point, isn't it better to build your own database? And choose the trade offs that fit your project. Or pay/hire someone to do it.
Definitely not. No regrets after making the move. March next year will be the two year mark.
I'm curious how you (and your team I presume) came to decide you need to build your own database?
Re: Building the world’s fastest website analytics (2021)
#13At some point, isn't it better to build your own database? And choose the trade offs that fit your project. Or pay/hire someone to do it.
These people don't have the inhouse skills to maintain a simple SQL server, let alone optimize their schema (MySQL being too slow to sum/group 11m records in a timely manner? Please.)
Attempting to make your own db while being ignorant of competing products would be an absolute disaster
Re: Building the world’s fastest website analytics (2021)
#14While reading the article, I found that they are rediscovering the tricks, already well-known for ClickHouse:
> The biggest mistake I made was that I kept the UPDATEs in our code (we update the previous pageview with the duration, remove the bounce, etc.). In a few weeks, I'll be moving to 100% append-only by utilizing negative numbers.
> For example, if you want to set bounce_rate to 0%, you would write a 1 for bounce_rate on the first pageview and then insert a duplicate with -1 for bounce_rate. And for the duplicate row, you'd have nothing set for pageviews, visits and uniques, so it would all group nicely.
This is nearly a copy-paste from ClickHouse documentation.
> We shard on UUID, and then we set SiteId as the sort key. We do this because we want to utilize something called "local joins" in SingleStore. Long story short, events can be joined with event_properties (allowing you to have thousands of dynamic properties per event you track), and it's fast.
This statement is slightly disappointing as well. No magic, no advantage to ClickHouse.
The outcome can be read as follows: "use SingleStore exactly as you'd use ClickHouse and maybe it will work alright".
SingleStore advantages:
- good compatibility with MySQL dialect;
- offers good UPDATE/DELETE for fresh data (invalidated by this article, as they dropped the usage of UPDATE for better performance);
- good support for JOINs (invalidated by this article, as they ended up using local JOINs);
SingleStore disadvantages:
- not open-source, worse brand recognition, the old brand (MemSQL) did not work well too;
- performance claims don't validate;
- older company and many initial developers no longer work there;
Fathom sounds like a toy use case as the data amount is too low. ClickHouse is almost universally selected for similar and larger use cases, with hundreds of billions of events each day.
PS. Nevertheless, the post is hilarious:
> I was suffering from low energy in the two weeks leading up to this migration, and I was feeling awful throughout migration week, especially on migration day. On Sunday 14th March 2021, two days after we had finished the migration, my wife showed me an already-half-used bag of coffee beans I had been drinking through most of this migration, and the bag said "DECAF." Divorce proceedings are underway.
Re: Building the world’s fastest website analytics (2021)
#15I have mixed feelings reading this article. There was no requirement around code refactoring, however multiple solutions were axed because of “how much of our application I was going to need to refactor”. Maybe have that as a hard requirement then and don’t invest time learning tech that doesn’t meet this requirement? Also, was that a joke about the divorce? Either way bad taste and red flag on the personality side f…
My wife thought it was funny ;)
Re: Building the world’s fastest website analytics (2021)
#16They would have saved a whole lot of time just using Clickhouse which is a US company and has completely severed ties with Russia [1]. Altinity has offered managed clickhouse since 2020 [2], and now Clickhouse Cloud does too [3]
[1] https://clickhouse.com/blog/we-stand-with-ukraine
Re: Building the world’s fastest website analytics (2021)
#17While reading the article, I found that they are rediscovering the tricks, already well-known for ClickHouse: > The biggest mistake I made was that I kept the UPDATEs in our code (we update the previous pageview with the duration, remove the bounce, etc.). In a few weeks, I'll be moving to 100% append-only by utilizing negative numbers. > For example, if you want to set bounce_rate to 0%, you would write a 1 for boun…
Clickhouse is great a single table analytical queries (Group-by + aggregates + filters over a single table). It can match any top tier columnstore database at this, maybe even better at smaller scales, and if that is all you need it will do the job well
Beyond this, it has a naive query optimizer and limited ability to run distributed joins. This is why you won't find well known analytical benchmark results like TPC-H and TPC-DS for clickhouse vs other SQL data warehouses.
SinglestoreDB has been doing real time analytics much longer then clickhouse and has a much more mature feature set at this point[1][2]. Our go to market is more big enterprise driven so we are definitely much less well known among developers. Revenue share-wise I suspect we are the leader in real time analytics (see disclaimer below - I'm biased but our revenue is reaching thresholds for IPO readiness).
Also, if you follow Jacks later posts you'll see he replaces DynamoDB and Redis with SinglestoreDB as well. Now were getting into places Clickhouse doesn't tread at all...
Disclaimer: I'm one of the cofounders of MemSQL/SingleStoreDB (still working away on making it better all these years later...).
[1] https://www.singlestore.com/blog/the-technical-capabilities-...
Re: Building the world’s fastest website analytics (2021)
#18While reading the article, I found that they are rediscovering the tricks, already well-known for ClickHouse: > The biggest mistake I made was that I kept the UPDATEs in our code (we update the previous pageview with the duration, remove the bounce, etc.). In a few weeks, I'll be moving to 100% append-only by utilizing negative numbers. > For example, if you want to set bounce_rate to 0%, you would write a 1 for boun…
Clickhouse is great a single table analytical queries (Group-by + aggregates + filters over a single table). It can match any top tier columnstore database at this, maybe even better at smaller scales, and if that is all you need it will do the job well Beyond this, it has a naive query optimizer and limited ability to run distributed joins. This is why you won't find well known analytical benchmark results like TPC-…
Can you provide SingleStore result for TPC-H and TPC-DS? I can't find it. Why?
[1] https://www.tpc.org/tpcds/results/tpcds_results5.asp?orderby...
[2] https://www.tpc.org/tpch/results/tpch_results5.asp?orderby=d...
Disclaimer: I work for ClickHouse
Re: Building the world’s fastest website analytics (2021)
#19They would have saved a whole lot of time just using Clickhouse which is a US company and has completely severed ties with Russia [1]. Altinity has offered managed clickhouse since 2020 [2], and now Clickhouse Cloud does too [3] [1] https://clickhouse.com/blog/we-stand-with-ukraine [2] https://altinity.com/cloud-database/ [3] https://clickhouse.com/cloud
If you're really confident that your SaaS application will only have that single workload (single table group-by) into perpetuity, it's a workable choice. But, that's a big bet against the reality of the diverse workloads required today in scale-up SaaS applications. Sure, you can stitch together specialized stores to make your own accidental distributed database for your app, but the simpler, more straight-forward and lower-latency alternative available. Yes, I'm a SingleStore employee, but it is worth considering the argument on its merit. It's also worth trying for free in production. The free tier for production is not time-bound, just capacity-bound.
Re: Building the world’s fastest website analytics (2021)
#20Earlier quoted context omitted.
Clickhouse is great a single table analytical queries (Group-by + aggregates + filters over a single table). It can match any top tier columnstore database at this, maybe even better at smaller scales, and if that is all you need it will do the job well Beyond this, it has a naive query optimizer and limited ability to run distributed joins. This is why you won't find well known analytical benchmark results like TPC-…
Can you provide SingleStore result for TPC-H and TPC-DS? I can't find it. Why? [1] https://www.tpc.org/tpcds/results/tpcds_results5.asp?orderby... [2] https://www.tpc.org/tpch/results/tpch_results5.asp?orderby=d... Disclaimer: I work for ClickHouse
Almost no one does "official" TPC results these days (maybe other then Oracle and some of the Chinese vendors).
Most other cloud DWs have public TPC-H or TPC-DS results that are easily googlable. Clickhouse is missing for a reason...
- https://research.gigaom.com/report/data-warehouse-cloud-benchmark/
- https://www.databricks.com/blog/2021/11/02/databricks-sets-official-data-warehousing-performance-record.html
- https://celerdata.com/blog/starrocks-queries-outperform-clickhouse-apache-druid-and-trino
- https://aws.amazon.com/blogs/big-data/amazon-redshift-continues-its-price-performance-leadership/