Khalil

@hellopivot

[Khalil @_khalilomer]

@jayair @berenddeboer Typesense is in-memory. Great if your dataset can fit in memory, but not relevant to the majority of elastic use cases.

[Khalil @_khalilomer]

@thdxr Publicly its still in 'researching' status. But my understanding is that the team that shipped ephemeral EBS has been assigned persistent EBS.
https://t.co/BmOxgKecIs

[Khalil @_khalilomer]

@thdxr Maybe there's a way to avoid the internal facing LB and just use service discovery?

[Khalil @_khalilomer]

@samlambert Yeah, 'distributed KV store with two-phase commit' doesn't sound as good.

[Khalil @_khalilomer]

@thdxr someone needs to ask lebron whether fargate is serverless

[Khalil @_khalilomer]

@dreamsofcode_io @thdxr Yeah we are doing the same thing, with NATS rather than Kafka. github .com/gobwas/ws is an extremely simple and high performance Go ws library.

[Khalil @_khalilomer]

@jayair Elasticache Serverless is much better with Valkey. 100MB minimum cache size rather than 1 GB.

[Khalil @_khalilomer]

@colin_fraser This is like an LLM autoimmune disease. The very probabilistic modeling that makes it tick is the thing that prevents it from getting things like this and the boat/goat test right.

[Khalil @_khalilomer]

@rafalwilinski > has joins

If you combine Citus with pgvector, this is theoretically possible with Postgres. However it is probably easier and fast enough in most cases to do the joins application-side and use Elastic for the vectors.

[Khalil @_khalilomer]

@c_hgungus @OSthatguy This is the best possible version of this meme for the 10 people who will get it.

[Khalil @_khalilomer]

@samlambert The best part is that they aren't a PlanetScale alternative. No serious buyer is unclear about whether they need row or column storage.

[Khalil @_khalilomer]

@swyx Some really good references here

[Khalil @_khalilomer]

@soleio Has got to be Sonoma, right?

[Khalil @_khalilomer]

@jdan [replying 2 hours later, not threaded]

ok can you please add that to linear for yourself?

[Khalil @_khalilomer]

@ankurnagpal ✋

[Khalil @_khalilomer]

@samlambert What's your experience with EBS at scale? I've heard conflicting things re: performance tradeoff between EBS and local disks for replicated database workloads where you aren't totally reliant on EBS replication for HA.

[Khalil @_khalilomer]

@raroque https://t.co/41F2KPm5cH

[Khalil @_khalilomer]

@andrii_sherman Better to just .gitignore in my opinion

[Khalil @_khalilomer]

@jdan @ValDotTown Big congrats to Fitzpatrick.