๐๐ฒ๐ฟ๐ฒ'๐ ๐ฎ ๐ฐ๐ผ๐๐ ๐๐ฟ๐ถ๐ฐ๐ธ ๐บ๐ผ๐๐ ๐๐ฒ๐ฎ๐บ๐ ๐บ๐ถ๐๐ ๐๐ถ๐๐ต ๐๐ต๐ฒ๐ถ๐ฟ ๐๐ฒ๐ฐ๐๐ผ๐ฟ ...
TL;DR ยท AI Summary
Milvus ๆๅบ้่ฟ compaction๏ผๆฎตๅๅนถไธ็ฉ็ๅ ้ค๏ผๅ TTL๏ผ่ชๅจ่ฟๆ๏ผไธค้กนๅ ็ฝฎๆบๅถ๏ผๅฏๆพ่้ไฝๅ้ๆฐๆฎๅบๅญๅจๆๆฌ๏ผๅฐคๅ ถ้็จไบไผ่ฏๆฐๆฎใๆถๆๆง RAG ็ญๆ็ๅฝๅจๆ็ๆฐๆฎๅบๆฏใ
Key Takeaways
- ๅ้ๆฐๆฎๅบไธญ้ป่พๅ ้คไธ้ๆพ็ฃ็็ฉบ้ด๏ผๅฏผ่ดๅญๅจ่จ่่พพ2โ5ๅ
- Compaction ็ฉ็ๅๅนถๆฎตๅนถๆธ ้คๅทฒๅ ๅ้๏ผ้้ฟๅผไธๅก้ซๅณฐๆง่ก
- TTL ๅฏ่ชๅจๆธ ็่ฟๆๆฐๆฎ๏ผ้ ๅ compaction ๅฎ็ฐ้ถๆๅจ่ฟ็ปด็ๆๆฌไผๅ
Outline
Jump quickly between sections.
ๆๅบๅคๆฐๅข้ๅฟฝ่งๅ้ๆฐๆฎๅบๅญๅจๆๆฌ๏ผ็ดๅฐ่ดฆๅๆฟๅขๆๅ็ฐ็ฃ็ๆ็ปญ่จ่ใ
Milvus ็ญ่ฟฝๅ ๅ็ณป็ปไธญๅ ้คไป ๆ ่ฎฐ่้็ฉ็็งป้ค๏ผๆงๆฎตๆฎ็ๅฏผ่ดๆฅ่ฏขไฝๆไธ็ฉบ้ดๆตช่ดนใ
ๅๅนถๅฐๆฎตใ็ฉ็ๅ ้คๅทฒๆ ่ฎฐ่ก๏ผไฝ I/O ๅฏ้๏ผๅปบ่ฎฎ็ฆปๅณฐ่ฐๅบฆใ
ๆๆถ้ด่ชๅจ่ฟๆๆฐๆฎ๏ผๅฝปๅบ่ง้ฟๆๅจๆธ ็๏ผๅคฉ็ถ้้ ไผ่ฏ/ๆถๆๆง RAG ๅบๆฏใ
ไบ่ ็ปๅๅฏๅคงๅน ๅๅๅฎ้ ๅญๅจๅผ้๏ผ็นๅซ้ๅๆๆ็กฎ็ๅฝๅจๆ็ๆฐๆฎๅทฅไฝ่ด่ฝฝใ
Mindmap
See how the topics connect at a glance.
ๆฅ็ๅคง็บฒๆๆฌ๏ผๆ ้็ข / ๆ JS ๅๅฅฝ๏ผ
- Milvus ๅญๅจๆๆฌไผๅ
- ้ฎ้ขๆ นๆบ
- ่ฟฝๅ ๅ็ณป็ป้ป่พๅ ้คไธ้ๆพ็ฃ็
- ๆฎตๅ ็งฏ โ ๆฅ่ฏขๆซๆๆ ๆ่ก โ ๅญๅจ่จ่2โ5x
- ๅ ณ้ฎๆๆฏ
- Compaction๏ผๅๅนถๆฎต + ็ฉ็ๅ ้ค
- TTL๏ผๆๆถ้ด่ชๅจ่ฟๆ
- ้็จๅบๆฏ
- ไผ่ฏๆฐๆฎ
- ๆถๆๆง RAG
- ๆ็ๅฝๅจๆ็ๅทฅไฝ่ด่ฝฝ
Highlights
Key sentences worth saving and sharing.
In append-only systems like Milvus, deleting a vector only flags it: the data stays on disk.
โ ๅๆ็ฌฌ2ๅฅ
Segments keep piling up, queries scan rows that no longer exist, and storage quietly grows 2โ5x beyond your actual live data.
โ ๅๆ็ฌฌ3ๅฅ
Compaction handles the rest. If you're running session data, time-sensitive RAG, or any workload where data has a shelf life, these two settings alone can cut storage costs dramatically.
โ ๅๆๆซๆฎต
๐ช๐ฒ'๐๐ฒ ๐ป๐ผ๐๐ถ๐ฐ๐ฒ๐ฑ ๐บ๐ผ๐๐ ๐๐ฒ๐ฎ๐บ๐ ๐ป๐ฒ๐๐ฒ๐ฟ ๐๐ต๐ถ๐ป๐ธ ๐ฎ๐ฏ๐ผ๐๐ ๐๐ฒ๐ฐ๐๐ผ๐ฟ ๐๐๐ผ๐ฟ๐ฎ๐ด๐ฒ ๐ฐ๐ผ๐๐๐ https://t.co/s0PCsgxGIs" / X
๐๐ฒ๐ฟ๐ฒ'๐ ๐ฎ ๐ฐ๐ผ๐๐ ๐๐ฟ๐ถ๐ฐ๐ธ ๐บ๐ผ๐๐ ๐๐ฒ๐ฎ๐บ๐ ๐บ๐ถ๐๐ ๐๐ถ๐๐ต ๐๐ต๐ฒ๐ถ๐ฟ ๐๐ฒ๐ฐ๐๐ผ๐ฟ ๐ฑ๐ฎ๐๐ฎ๐ฏ๐ฎ๐๐ฒ: ๐๐๐ถ๐ป๐ด ๐ฐ๐ผ๐บ๐ฝ๐ฎ๐ฐ๐๐ถ๐ผ๐ป ๐ฎ๐ป๐ฑ ๐ง๐ง๐. ๐ช๐ฒ'๐๐ฒ ๐ป๐ผ๐๐ถ๐ฐ๐ฒ๐ฑ ๐บ๐ผ๐๐ ๐๐ฒ๐ฎ๐บ๐ ๐ป๐ฒ๐๐ฒ๐ฟ ๐๐ต๐ถ๐ป๐ธ ๐ฎ๐ฏ๐ผ๐๐ ๐๐ฒ๐ฐ๐๐ผ๐ฟ ๐๐๐ผ๐ฟ๐ฎ๐ด๐ฒ ๐ฐ๐ผ๐๐๐ ๐๐ป๐๐ถ๐น ๐๐ต๐ฒ ๐ฏ๐ถ๐น๐น ๐๐ฝ๐ถ๐ธ๐ฒ๐. In append-only systems like Milvus, deleting a vector only flags it: the data stays on disk. Segments keep piling up, queries scan rows that no longer exist, and storage quietly grows 2โ5x beyond your actual live data. ๐ง๐๐ผ ๐ฏ๐๐ถ๐น๐-๐ถ๐ป ๐ณ๐ฒ๐ฎ๐๐๐ฟ๐ฒ๐ ๐ณ๐ถ๐
๐๐ต๐ถ๐: โข ๐๐ผ๐บ๐ฝ๐ฎ๐ฐ๐๐ถ๐ผ๐ป โ merges small segments and physically drops deleted rows. Schedule it off-peak since it's I/O-heavy. โข ๐ง๐ง๐ โ auto-expires old data so you skip manual cleanup entirely. Compaction handles the rest. ๐๐ณ ๐๐ผ๐'๐ฟ๐ฒ ๐ฟ๐๐ป๐ป๐ถ๐ป๐ด ๐๐ฒ๐๐๐ถ๐ผ๐ป ๐ฑ๐ฎ๐๐ฎ, ๐๐ถ๐บ๐ฒ-๐๐ฒ๐ป๐๐ถ๐๐ถ๐๐ฒ ๐ฅ๐๐, ๐ผ๐ฟ ๐ฎ๐ป๐ ๐๐ผ๐ฟ๐ธ๐น๐ผ๐ฎ๐ฑ ๐๐ต๐ฒ๐ฟ๐ฒ ๐ฑ๐ฎ๐๐ฎ ๐ต๐ฎ๐ ๐ฎ ๐๐ต๐ฒ๐น๐ณ ๐น๐ถ๐ณ๐ฒ, ๐๐ต๐ฒ๐๐ฒ ๐๐๐ผ ๐๐ฒ๐๐๐ถ๐ป๐ด๐ ๐ฎ๐น๐ผ๐ป๐ฒ ๐ฐ๐ฎ๐ป ๐ฐ๐๐ ๐๐๐ผ๐ฟ๐ฎ๐ด๐ฒ ๐ฐ๐ผ๐๐๐ ๐ฑ๐ฟ๐ฎ๐บ๐ฎ๐๐ถ๐ฐ๐ฎ๐น๐น๐. ๐๐๐น๐น ๐ด๐๐ถ๐ฑ๐ฒ ๐ผ๐ป ๐๐ต๐ฒ ๐ ๐ถ๐น๐๐๐ ๐ฏ๐น๐ผ๐ด 
milvus.io/blog/how-to-cu
