Google SWE teaches systems design | EP27: Search Indexes

preview_player
Показать описание
Could use a search index for my chode, its so small I can't even find it sometimes

Also, since I forgot to mention it: do not use ElasticSearch as your main database solution. It does not support multidocument transactions even on the same cluster. Furthermore, writes are not instantly available due to the second long period before they are flushed to indexes. Even further, the large amounts of caching results in some older data being returned sometimes. The biggest reason ultimately though is that updating a document means that all of the relevant inverse indexes have to be updated, which is a pain in the fucking ass.

Recommended Reading/Viewing:
  1. Jordan has no life
Рекомендации по теме
Google SWE teaches 0:16:05

Google SWE teaches systems design | EP1: Database Design

Google SWE teaches 0:14:28

Google SWE teaches systems design | EP19: Cassandra Deep Dive

Google SWE teaches 0:08:56

Google SWE teaches systems design | EP47: Collaborative Text Editing (operational transform v. CRDT)

Google SWE teaches 0:12:30

Google SWE teaches systems design | EP15: Batch Processing

Google SWE teaches 0:12:37

Google SWE teaches systems design | EP2: single leader replication

Google SWE teaches 0:07:06

Google SWE teaches systems design | EP24: Riak Explained

Google SWE teaches 0:10:24

Google SWE teaches systems design | EP4: Leaderless Replication

Google SWE teaches 0:18:06

Google SWE teaches systems design | EP16: Stream Processing

Google SWE teaches 0:06:09

Google SWE teaches systems design | EP10: Unreliable clocks

Google SWE teaches 0:12:13

Google SWE teaches systems design | EP21: Hadoop File System Design

Google SWE teaches 0:14:30

Google SWE teaches systems design | EP3: Multileader replication

Google SWE teaches 0:17:05

Google SWE teaches systems design | EP12: Linearizability and Ordering

Google SWE teaches 0:07:59

Google SWE teaches systems design | EP42: Merkle Trees

Google SWE teaches 0:09:05

Google SWE teaches systems design | EP50: Approaching Interview Questions

Google SWE teaches 0:10:54

Google SWE teaches systems design | EP34: Amazon Aurora

Google SWE teaches 0:15:16

Google SWE teaches systems design | EP29: Geospatial Indexes

Google SWE teaches 0:15:17

Google SWE teaches systems design | EP14: Raft in 15 minutes

Google SWE teaches 0:10:23

Google SWE teaches systems design | EP9: Data warehousing

Google SWE teaches 0:10:02

Google SWE teaches systems design | EP27: Search Indexes

Google SWE teaches 0:09:05

Google SWE teaches systems design | EP39: Apache Spark

Google SWE teaches 0:10:44

Google SWE teaches systems design | EP28: Time Series Databases

Google SWE teaches 0:09:39

Google SWE teaches systems design | EP41: Bloom Filters for set approximation

Google SWE teaches 0:13:20

Google SWE teaches systems design | EP7: Weak forms of isolation

Google SWE teaches 0:12:43

Google SWE teaches systems design | EP32: Google Spanner

Комментарии
Автор

Great video. Side note, I realized that some Leetcode questions help in understanding this concept (Edit Distance, Start and End of a value in a Sorted List, etc.)

mnchester
Автор

Hello, I believe elastic-search does not use LSM-tree for its architecture, but a similar approach using segment compaction on its inverted index data-structure. Can we correct this info for other users? Other than that great content! I've learned a lot from your videos.

mihnsmithvu
Автор

Can someone provide clarification on this - when we mention at 1:49 that we write first to an in-memory buffer, does it mean we're updating the LSM tree that we maintain in memory?

AnujNagpall