If You Care About Performance, Use User Defined Types
•
0 likes•2,477 views
Shlomi Livne, VP of R&D at ScyllaDB, presented on the performance benefits of using user-defined types (UDTs) in ScyllaDB. He explained that with traditional columns, each column has overhead and flexibility comes at a price. However, with frozen UDTs, the columns are treated as a single unit, sharing metadata and improving performance. Livne showed results of a test where UDTs with many fields outperformed traditional columns with the same number of fields. However, he noted that Scylla's row cache and Java driver performance need improvement for UDTs.
Report
Share
More Related Content
What's hot
What's hot (20)
Viewers also liked
Viewers also liked (8)
Similar to If You Care About Performance, Use User Defined Types
Similar to If You Care About Performance, Use User Defined Types (7)
More from ScyllaDB
More from ScyllaDB (20)
Recently uploaded
Recently uploaded (20)
If You Care About Performance, Use User Defined Types
- 1. PRESENTATION TITLE ON ONE LINE AND ON TWO LINES First and last name Position, company If You Care About Performance Use User Defined Types VP of R&D, ScyllaDB Shlomi Livne
- 2. PRESENTATION TITLE ON ONE LINE AND ON TWO LINES First and last name Position, company Shlomi Livne 2 Shlomi is VP of R&D at ScyllaDB. Prior to ScyllaDB, he led the research and development team at Convergin, which was acquired by Oracle.
- 3. PRESENTATION TITLE ON ONE LINE AND ON TWO LINES First and last name Position, company Every column has an overhead For each Column/Cell in a Row you can ▪ Insert/Update independently ▪ Set a different TTL o CQL: INSERT INTO … USING TTL X; ▪ Set a different Writetime o CQL: INSERT INTO … USING WRITETIME X; ▪ Delete independently This “flexibility” comes with a price - do you really need it ? 3
- 4. PRESENTATION TITLE ON ONE LINE AND ON TWO LINES First and last name Position, company Frozen User Defined Types ▪ User Defined Types - allows the user to define a structure ▪ In Scylla frozen<UDT> are supported ▪ frozen<UDT> are treated as a single column o they share all the metadata o an insert/update will set the complete cell (till we implement https://github.com/scylladb/scylla/issues/2201) o select can extract a subset of a UDT value 4
- 5. PRESENTATION TITLE ON ONE LINE AND ON TWO LINES First and last name Position, company So how much “better” are UDTs ? Example: ▪ Compares o Tables of 10,20,30,40,50 blob columns o Tables with a frozen udt with 10,20,30,40,50 blob fields ▪ Write workload: all columns/fields are populated with 10 bytes ▪ Read workload: all columns/fields are retrieved 5
- 6. PRESENTATION TITLE ON ONE LINE AND ON TWO LINES First and last name Position, company 6
- 7. PRESENTATION TITLE ON ONE LINE AND ON TWO LINES First and last name Position, company 7
- 8. PRESENTATION TITLE ON ONE LINE AND ON TWO LINES First and last name Position, company 8
- 9. PRESENTATION TITLE ON ONE LINE AND ON TWO LINES First and last name Position, company 9
- 10. PRESENTATION TITLE ON ONE LINE AND ON TWO LINES First and last name Position, company 2 last comments: ▪ Scylla’s row cache - caches complete rows o It will cache all the row columns - so retrieving a partial set of columns still forces scylla to read from disk the complete row ▪ There’s an additional “overhead” on the client side, a sample showed that the Java driver is not as performant in UDTs (test it out) 10
- 11. PRESENTATION TITLE ON ONE LINE AND ON TWO LINES First and last name Position, company THANK YOU shlomi@scylladb.com @ShlomiLivne Any questions?