Adds more databases and benchmarks

[ribalba]

All the runs work on the cluster

[ArneTR]

@coderabbitai review

[coderabbitai]

✅ Action performed

Review finished.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

[ArneTR]

@coderabbitai re-review

[coderabbitai]

@ArneTR: Sure, I'll re-review the changes now.

✅ Action performed

Review finished.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

[ArneTR]

This PR is so big that I have quite some trouble reviewing it sadly :/

I also tried with coderabbit but even that failed as it, according to the support, has a 150 file limit.

Is there a way to split this PR up so that at least we can have it reviewed by AI?

One note I could find in my review attempt:

• I have read the benchmark design and tuning readme and find the approach plausible and accurate. But it is still very opaque to me on what grounds T1 and T2 optimizations happen. I do not feel I can reason about it or reproduce it. To give an example: It feels like for the DBs they are setting very different values - For instance benchmarks/chbenchmark/mysql.t1.yml and benchmarks/chbenchmark/oracle.t1.yml seem to set totally different settings. Is that a weird example I picked out or are the DB settings not that well comparable after all?

In essence I think this might be hard to maintain. Settings and tunings feel very unique to the databases and provide no abstraction. If some setting changes in behaviour we have no way to know or to debug.
A question comes to mind here: Should we really include the optimized versions at all? At least in the main branch. Maybe this is more a case for a branch that we allow to go stale?

How are the results and the data? Is there a clear relationship between settings and energy visible? Do T1 optimizations have similar energy improvements accross DBs?

[ribalba]

The idea with the tuning was more that we can have training data for the AI model.

[ribalba]

@ArneTR we should call about this.

[ribalba]

Superseded by a two-way split of this branch:

• Add Db2/MSSQL/Oracle engines and TPC-H, Wikipedia, YCSB, CH-benCHmark benchmarks #21 — Db2/MSSQL/Oracle engines and the TPC-H, Wikipedia, YCSB, CH-benCHmark benchmarks, all at default configuration (95 files, based on main).
• Add T1 (envelope-sized) and T2 (workload-aware) tuning tiers #22 — the T1/T2 tuning tiers and TUNING.md (66 files, stacked on Add Db2/MSSQL/Oracle engines and TPC-H, Wikipedia, YCSB, CH-benCHmark benchmarks #21; GitHub will retarget it to main once Add Db2/MSSQL/Oracle engines and TPC-H, Wikipedia, YCSB, CH-benCHmark benchmarks #21 merges).

The two branches together are byte-identical to new-dbs-and-tpch, so nothing is lost. Closing this in favour of the split.