Run affordable open source ClickHouse® anywhere with #AltinityCloud, Expert Support, BYOC, & real-time data lake integration to scale queries on cheap storage.
Berkeley, CA
Based in India
- Tweets 2,445
- Following 366
- Followers 1,731
- Likes 1,006
📣 Raw EXPLAIN text can be a lot to parse.
The #AltinitySQLBrowser turns #ClickHouse® EXPLAIN PIPELINE output into an interactive processor graph, so parallelism and Resize points are easier to spot. Pan, zoom, or open it fullscreen.
💡 Full docs: hubs.la/Q04ypVrf0
13
Slow #ClickHouse® query? Start with the built-in tools. 💪🏽
▪ query_log: resources + runtime
▪ processors_profile_log: time per stage
▪ query_metric_log: live stats
▪ EXPLAIN + the chdig CLI
💡 Get the full breakdown: hubs.la/Q04xYzy60
1
103
Scaling #ClickHouse® storage without scaling compute? Meet CAS. 💥
Content-Addressed Storage (#CAS) adds shared object storage to existing MergeTree model, so storage can scale independently without replacing #MergeTree with a new engine.
CAS explained: hubs.la/Q04y7dTR0
2
132
0.65s vs. 1.42s on ~126M rows. Same #ClickHouse® table, same filter. The only change was the query’s ORDER BY. ⚡
If a query is sorting more than expected, check whether its ORDER BY matches the table’s: kb.altinity.com/engines/merg…
1
2
133
🎡 Hey Londoners, we'll be at the Data Engineering London #meetup tomorrow for an evening on Storage + #AI.
Expect deep dives on Hot SSDs to #lakehouses, data movement, metadata performance, and more.
📍 London Art Bar | 6 pm
🍻 Food + drinks on us
🔗 hubs.la/Q04xY3Lt0
1
153
Your analytical database is columnar & your dataframe library is too. But the path between them often isn't. That can mean extra copies & transposes to move query results. 😓
See how #ApacheArrow & #ADBC can eliminate that in this Unevenly Distributed ep: hubs.la/Q04xRPlG0
80
🇵🇱 Warsaw - thank you for making it an evening to remember.
Huge thanks to #OSACommunity for bringing everyone together, @dbcicero & other speakers for sharing the stage, & Josh Lee for hosting. Shoutout to everyone from the open source community who made the meetup so special.
2
92
Disk, memory, CPU, network. When #ClickHouse® has a problem, which one do you check first? 🤔
Our Sept. admin #training gets into ClickHouse internals so you can diagnose issues instead of guessing.
4 live sessions starting tomorrow (limited seats left): hubs.la/Q04x0mqb0
4
111
If older #ClickHouse® data is rarely queried, it may not need to stay on your fastest storage tier. 😎
#TTLs & tiered storage can keep recent data on fast block storage, move older data to S3, remove expired data automatically, & save storage cost by up to 30% (shown below). 👇🏻
1
96
A large export fails halfway through. So, restart from zero? 😬
While moving from #ClickHouse® ingestion to #ProjectAntalya, MindWise built o_rabbit so failed tasks retry independently & exports run in parallel.
⚡500GB export: 86 → 40 min on 4 workers: hubs.la/Q04wQWlJ0
1
84
A production incident is a bad time to learn how your #ClickHouse® cluster actually works. 🫠
Go deeper on storage, replication, resources, debugging, upgrades, monitoring, Keeper, & backups in 4 live Altinity #Administrator #Training sessions.
🚀 9/15: hubs.la/Q04x1Ny10
1
95
Warsaw, we’re on tomorrow! 🇵🇱🕺🏻
We're joining #OSACommunity for an evening on building #AI w/ live data, low latency, & real-world scale.
Expect streaming, RAG, #observability, databases, open-source infra discussion.
📍 Workin, Senatorska 2
🕕 6 PM
🔜 hubs.ly/Q04w_Zlv0
2
73
More partitions ≠ faster #ClickHouse®. 👎🏻
Too-granular partitioning means more parts, more files, more background work, & potentially more remote-storage requests.
Use PARTITION BY for the job it’s meant to do, not as a generic faster-SELECT switch.
👉🏻 hubs.la/Q04vVpxz0
1
3
121
“Have you guys called Altinity yet?” - Avi Press, CEO of @scarf_oss
As the data grew 200–300x, Scarf needed to bring #ClickHouse® costs under control. With Altinity’s help, they moved to #BYOC, optimized the schema & queries, & brought #TCO down by 60%.
hubs.la/Q04wQWnq0
3
187
Optimize the #SQL or add more hardware? 🤷🏻♀️
For ClickHouse®, optimize first. If traffic forces you to scale, scale. Then come back to the workload and right-size again.
Scale for demand, not inefficiency.
💰 See what actually lowers #ClickHouse costs: hubs.la/Q04vT_Wl0
1
93
Your slowest #ClickHouse® query may be the wrong one to optimize first.
A fast query running thousands of times can drive more total load than one obvious slow query. system.query_log + normalized_query_hash can help surface the patterns actually worth your optimization time. 👇🏻
5
162
#BYOC sounds great until the obvious question comes up: are you getting more control, or just taking on more ops? 🤔
Tomorrow in SF, we dig into that with #ClickHouse® as the example: where ops end, customer control begins, and when BYOC works well.
🎟️ hubs.la/Q04tQwt70
1
2
143
"The #AI agent dropped a table in production" is a sentence nobody wants to say in a postmortem. 😬
The fix? Put the boundary where the agent can't talk its way around, using #ClickHouse® RBAC, OAuth, and MCP server.
⚙️ See how to fit all three together: hubs.la/Q04vRNtw0
1
85
Added a 14-day #TTL to system.query_log but older data is still there? Check the partitioning. 🕵🏻♂️
With monthly partitions, #ClickHouse® may need to rewrite parts or wait for the whole partition to expire.
Daily partitions make cleanup much simpler:
hubs.la/Q04tZ4pp0
1
237
The best duplicate in #ClickHouse® is the one you never write. 😅
If you can dedup before ingest, great. If not, #ReplacingMergeTree handles old versions as parts merge, with FINAL for a deduplicated view before merges complete.
📺 Watch from 0:57: hubs.la/Q04tYPNv0 📺
1
119
#BYOC isn’t just about data residency. It gives you control over infra choices that affect cost & performance: VM/storage types, your cloud discounts, compute close to the data, while still handing off DB ops.
💪🏻 Read why that matters for #ClickHouse®: hubs.la/Q04tYRx30
1
123
One unrelated read could flush everyone else’s write batch in #ClickHouse® Keeper. 💡
26.6 reworked batching/pipelining & fixed NuRaft follower appends. Keeper is now ~2x faster overall.
Still on #ZooKeeper because of old benchmarks? Retest 26.6: hubs.la/Q04tYwnj0
1
2
131
SaaS is convenient until workloads get expensive, data needs to stay put, or security wants tighter control. 😵💫
SF #meetup on 9/2: Altinity, @last9io, @nirvanalabsai, @tensor9ine discuss #BYOC, IOPS costs, and running software in customer clouds.
🎟️ hubs.la/Q04tQd8q0
2
5
208
#AI makes it easier to open issues & PRs. It doesn't give maintainers more hrs to triage them. 🧹
Evan Rusackas (PMC member of #ApacheSuperset) went from ~800 open issues to ~200 in a couple of months by splitting the backlog into narrow jobs.
🎧 Ep 9: hubs.la/Q04tk_Df0
1
87
📣 Happening Tomorrow → bring your #ClickHouse® cost questions to Altinity #OfficeHours.
We’ll cover where to start, how to find your most expensive queries, and which ones to optimize first.
🕗 Aug 20, 8am PT
🚀 hubs.la/Q04szlQj0
1
87
🚨 Happening Tomorrow: 6 best practices to save big money on your #ClickHouse® bill.
We'll cover how to reduce compute, storage, & networking costs, why #opensource matters, & how #BYOC & cost-efficient hardware providers change the hosting math.
💪🏻 hubs.la/Q04tqDJT0
2
112
Replication queue stuck in #ClickHouse®? ♻️
Before you hit “delete GET_PARTS,” check why the queue is stuck first. Task type matters, and the fix may be much simpler.
💪🏻 Full troubleshooting instructions: kb.altinity.com/altinity-kb-…
3
126
Your #ClickHouse® bill isn't only a query problem. 🧐
Tomorrow, Altinity CEO @dbcicero will look beyond tuning & discuss how compute, storage, networking, hosting, #BYOC & alternative hardware, & using open source affect your bills. 💸
Aug 19, 8am PT: hubs.la/Q04szxVs0
1
1
870
💰 Your #ClickHouse® bill is climbing. Where do you look first?
This month’s #OfficeHours focuses on cost optimization: how to find expensive queries and decide what to optimize first.
Bring cost questions. We'll answer them live.
📆 Aug 20, 8am PT
🔗 hubs.la/Q04szjqw0
1
4
237
🧐 #ClickHouse® cost optimization doesn't start with query tuning. It starts with where & how you host.
Altinity CEO @dbcicero will cover 6 practices across compute, storage, networking, open source, #BYOC, & alternative hardware options.
Aug 19, 8am PT
hubs.la/Q04sc_-Y0
458
Can a single #BYOC node stay responsive across 3B rows? Here are the results. 📊
Single-node #ClickHouse®, managed by Altinity, on @nirvanalabsai. SSB SF=500: 34 ms selective queries, ~70M rows/sec full scans from disk.
Full methodology included: hubs.la/Q04s7Fv30
2
123
Google Maps, but for your #ClickHouse® #schema. 🗺️
The Altinity #SQL Browser gives you an interactive view of your tables, MV, dictionaries, & Distributed tables. Drag a database in & see what writes to what, w/o chaining endless SHOW CREATE calls.
🔗 hubs.la/Q04rR7wL0
1
92
"Won't cross-AZ replication wreck our bill?" 💸 Good news: #ClickHouse® replicates compressed parts, so ~10% of raw ingest crosses AZ. Merges are local. The real cost trap is Keeper cross-AZ traffic at ~$0.02/GB. Keep Keeper local.
💡 Full breakdown: hubs.la/Q04rG4Q30
1
91
One change to a materialized view cut #ClickHouse® ingestion lag by more than half. No infra changes. No query rewrites. ⚡
The fix: repoint the MV from a Distributed table to the local replica table. 🚀
We walk you through the full diagnosis: hubs.la/Q04q-8s60
156
🎉 So excited to introduce the Altinity SQL Browser—a rich #SQL console for ClickHouse® with charts, EXPLAIN pipeline graphs, autocomplete, & a saved-query library.
Deployed as a single HTML file.
🚀 Take it for a spin: hubs.la/Q04rjhMc0
📝 Docs: hubs.la/Q04rjhYd0
4
145
📣 We're partnering with @last9io because #observability costs shouldn't grow faster than the systems they monitor.
Last9 runs unified telemetry on #ClickHouse® and the full stack lives in your environment while we handle ClickHouse ops 24/7.
Learn more: hubs.la/Q04r8FyY0
2
4
216
Removing a #ClickHouse® replica? SYSTEM DROP REPLICA has sharp edges. ✂️
Run it from the wrong node and it fails silently. It only clears Keeper metadata, not data on disk.
Confirm removal in system.replicas before you touch config. Full tutorial: hubs.la/Q04q-d9L0
102
~10x cold read speedup vs AWS gp3. Database and workload on the same rack. $0 internal egress. ~70M rows/sec sustained from disk. That's #ClickHouse® on @nirvanalabsai, the new BYOC option for #AltinityCloud. Benchmarks & repos are public. 💪🏽
Learn more: hubs.la/Q04qTYZN0
4
12
736
👇🏻 This interactive demo is powered by #ClickHouse® API Endpoints in #AltinityCloud. Every chart, filter, interaction retrieves data through REST endpoints. An analytics app written w/o a single line of SQL.
Try it: hubs.la/Q04p7byx0
Read the blog: hubs.la/Q04p7ftW0
120
Tomorrow: @last9io ingests 480M log lines/min into #ClickHouse®. On the same shared layer as metrics & traces. 💪🏻
Getting there meant hitting every failure mode. Josh Lee (Altinity) + @prathamesh2_ (Last9) on what broke at scale & how to model around it.
hubs.la/Q04pB_vJ0
2
1
195
July's Altinity #OfficeHours topic: Projections & Materialized Views in #ClickHouse®.
If you've got questions about when to use which, how they interact, or whether you're overcomplicating things, this is it!
📆 July 28, 8-9 AM PT
💬 Ask your Qs live: hubs.la/Q04q4zJg0
116
The "too many parts" error in #ClickHouse® usually comes from too many small inserts.
The fix is to help it make big parts:
1️⃣ stream big files via HTTP
2️⃣ turn on async inserts
3️⃣ use #Kafka engine + #MaterializedViews
Altinity CEO @dbcicero explains: hubs.la/Q04pSvZK0
1
187
🚨 If you're building apps on ClickHouse®, you shouldn't have to write SQL.
With #ClickHouse API Endpoints in #AltinityCloud, you call a REST API while ClickHouse handles the SQL behind the scenes.
Focus on building features, not writing queries.
👉 hubs.la/Q04p7GYG0
79
"It works fine in a demo." That sentence has preceded a lot of #ClickHouse® schema pain at scale. 😅 @last9io hit the failure modes at 480M log lines/min. Altinity's Josh Lee + Last9's @prathamesh2_ will show what broke & how to model around it on 7/23.
hubs.la/Q04pBXhV0
3
2
193
🛡️ #AltinityStableBuilds go through 5 stages before certification: build, upstream tests, regression tests, security, then release, not just the newest #ClickHouse® tag.
Open source, upstream-compatible, #FIPS-compatible, & 3-years support.
Try the demo: hubs.la/Q04nSTMn0
2
98
⚡ #ProjectAntalya lets one SQL statement query hot data in #ClickHouse® MergeTree & cold data in #Iceberg together.
What makes it fast - stateless swarm clusters that stay sub-second as compute scales up/down. It's 100% open source too.
Try the demo: hubs.la/Q04nTf8-0
1
98
Your app developers need access to your data. Not your database.
With #ClickHouse® API Endpoints in #AltinityCloud, your developers call a REST API w/o needing to know your schema or write SQL.
Live dashboard: hubs.la/Q04p7ghp0
See how it is built: hubs.la/Q04p7C0G0
1
109
⚠️ Every upgrade carries regression risk.
Upstream #ClickHouse® ships monthly, but only 3 latest releases are supported & LTS lasts a year.
#AltinityStableBuilds go through build, test & security validation, with up to 3 years of support. Try the #demo:
hubs.ly/Q04nTc200
1
92
🧱 Fast storage gets expensive.
Object storage is cheaper, but the query path gets more complicated.
#ProjectAntalya's Hybrid tables query #ClickHouse® MergeTree & #Iceberg with one SQL statement, tested here on NYC taxi & US flight data. Try the demo: hubs.la/Q04nMqZY0
1
92
Your IdP already controls access to everything else. Why should #ClickHouse® be the odd one out?
#AltinityAntalya builds now support #OAuth token-based auth w/ any OIDC provider. Short-lived tokens, IdP role mapping, no shared passwords. Now GA in 26.3: hubs.la/Q04njSJh0
1
85