[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.] [@BenjDicken](/creator/twitter/BenjDicken) "Today I'm reading the original log-structured merge-tree paper. LSM trees were designed to reduce IO cost for high-volume write-heavy database workloads. This came out in 1996 and today it's widely used by databases like DynamoDB ClickHouse and Google Bigtable"  [@BenjDicken](/creator/x/BenjDicken) on [X](/post/tweet/1937553103299203345) 2025-06-24 16:47:12 UTC 9796 followers, 15.1K engagements "@chris14978335 There's a difference between the AZ label you see in your AWS console and the actual physical availability zone. In that visual I used the physical AZ IDs. Read more here:"  [@BenjDicken](/creator/x/BenjDicken) on [X](/post/tweet/1947885969728016577) 2025-07-23 05:06:19 UTC 9792 followers, XX engagements "LSM trees use multiple layers and batched writes to improve efficiency. Data is first inserted into the sequential log (for recovery) and an in-memory tree. It only gets persisted to the on-disk B-tree in sequential batches"  [@BenjDicken](/creator/x/BenjDicken) on [X](/post/tweet/1937556932614848914) 2025-06-24 17:02:25 UTC 9807 followers, XXX engagements "@RichOBray That makes more sense 16mm is quite wide but with that level of crop it becomes more reasonable"  [@BenjDicken](/creator/x/BenjDicken) on [X](/post/tweet/1947716896696242229) 2025-07-22 17:54:29 UTC 9790 followers, XX engagements "@rixer_inc If your entire database can fit in RAM you won't notice as much of a difference except for writes. But this is rarely the case for large-scale databases"  [@BenjDicken](/creator/x/BenjDicken) on [X](/post/tweet/1947671361360732621) 2025-07-22 14:53:33 UTC 9792 followers, XXX engagements "The best way to maximize IOPS are using local NVMe SSDs. You get access to modern high-bandwidth and low-latency storage technology without no network or throttling in-between. This is what we use for @PlanetScale Metal to achieve best-in-class performance"  [@BenjDicken](/creator/x/BenjDicken) on [X](/post/tweet/1947344509362123010) 2025-07-21 17:14:45 UTC 9776 followers, 1069 engagements "Can't get over how great the postgres metrics reporting is. The observability we're giving our users is off the charts"  [@BenjDicken](/creator/x/BenjDicken) on [X](/post/tweet/1946303641478189280) 2025-07-18 20:18:43 UTC 9791 followers, 10.3K engagements "IOPS capacity is a critical consideration for high-performance databases. Yet many don't understand what they are how they impact performance and their costs. Let's take deep dive: What are IOPS"  [@BenjDicken](/creator/x/BenjDicken) on [X](/post/tweet/1947344505864393205) 2025-07-21 17:14:44 UTC 9807 followers, 17K engagements "The latency between even two "nearby" regions like us-east X and X is 18ms. us-east-1 to us-west-1 is closer to 60ms. This advice is spot on. Do yourself a favor and keep your database in the same region as your other infra. Better yet same AZ"  [@BenjDicken](/creator/x/BenjDicken) on [X](/post/tweet/1947641754640941353) 2025-07-22 12:55:54 UTC 9807 followers, 178.1K engagements
[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
@BenjDicken
"Today I'm reading the original log-structured merge-tree paper. LSM trees were designed to reduce IO cost for high-volume write-heavy database workloads. This came out in 1996 and today it's widely used by databases like DynamoDB ClickHouse and Google Bigtable" @BenjDicken on X 2025-06-24 16:47:12 UTC 9796 followers, 15.1K engagements
"@chris14978335 There's a difference between the AZ label you see in your AWS console and the actual physical availability zone. In that visual I used the physical AZ IDs. Read more here:" @BenjDicken on X 2025-07-23 05:06:19 UTC 9792 followers, XX engagements
"LSM trees use multiple layers and batched writes to improve efficiency. Data is first inserted into the sequential log (for recovery) and an in-memory tree. It only gets persisted to the on-disk B-tree in sequential batches" @BenjDicken on X 2025-06-24 17:02:25 UTC 9807 followers, XXX engagements
"@RichOBray That makes more sense 16mm is quite wide but with that level of crop it becomes more reasonable" @BenjDicken on X 2025-07-22 17:54:29 UTC 9790 followers, XX engagements
"@rixer_inc If your entire database can fit in RAM you won't notice as much of a difference except for writes. But this is rarely the case for large-scale databases" @BenjDicken on X 2025-07-22 14:53:33 UTC 9792 followers, XXX engagements
"The best way to maximize IOPS are using local NVMe SSDs. You get access to modern high-bandwidth and low-latency storage technology without no network or throttling in-between. This is what we use for @PlanetScale Metal to achieve best-in-class performance" @BenjDicken on X 2025-07-21 17:14:45 UTC 9776 followers, 1069 engagements
"Can't get over how great the postgres metrics reporting is. The observability we're giving our users is off the charts" @BenjDicken on X 2025-07-18 20:18:43 UTC 9791 followers, 10.3K engagements
"IOPS capacity is a critical consideration for high-performance databases. Yet many don't understand what they are how they impact performance and their costs. Let's take deep dive: What are IOPS" @BenjDicken on X 2025-07-21 17:14:44 UTC 9807 followers, 17K engagements
"The latency between even two "nearby" regions like us-east X and X is 18ms. us-east-1 to us-west-1 is closer to 60ms. This advice is spot on. Do yourself a favor and keep your database in the same region as your other infra. Better yet same AZ" @BenjDicken on X 2025-07-22 12:55:54 UTC 9807 followers, 178.1K engagements
/creator/twitter::1761964147510767616/posts