Most nodes in the cluster receive the same shred from two different
nodes: parent, and the first node of their neighborhood:
https://github.com/solana-labs/solana/blob/a8c695ba5/core/src/cluster_nodes.rs#L178-L197
Because of the erasure codings, half of the shreds are already
redundant. So this redundant propagation path will only add extra
overhead.
Additionally the very first node of the broadcast tree has 2x fanout
(i.e. 400 nodes) which adds too much load at one node.
This commit simplifies the broadcast tree by dropping the redundant
propagation path and removing the 2x fanout at root node.
- Add write-lock-contention option, replacing same_payer
- write-lock-contention also has a same-batch-only value, where
contention happens only inside batches, not between them
- Rename num-threads to batches-per-iteration, which is closer to what
it is actually doing.
- Add num-banking-threads as a new option
- Rename packets-per-chunk to packets-per-batch, because this is closer
to what's happening; and it was previously confusing that num-chunks
had little to do with packets-per-chunk.
Example output for a iterations=100 and a permutation of inputs:
contention,threads,batchsize,batchcount,tps
none, 3,192, 4,65290.30
none, 4,192, 4,77358.06
none, 5,192, 4,86436.65
none, 3, 12,64,43944.57
none, 4, 12,64,65852.15
none, 5, 12,64,70674.37
same-batch-only,3,192, 4,3928.21
same-batch-only,4,192, 4,6460.15
same-batch-only,5,192, 4,7242.85
same-batch-only,3, 12,64,11377.58
same-batch-only,4, 12,64,19582.79
same-batch-only,5, 12,64,24648.45
full, 3,192, 4,3914.26
full, 4,192, 4,2102.99
full, 5,192, 4,3041.87
full, 3, 12,64,11316.17
full, 4, 12,64,2224.99
full, 5, 12,64,5240.32
1. use was_executed to correctly identify transactions requires cost adjustment;
2. add function to specifically handle executino cost adjustment without have to copy accounts
* Increase connection timeouts
* Bump quic connection cache to 1024
* Use constant for quic connection timeout and add warm cache service
* Fixes to QUIC warmup service
* fix check failure
* fixes after rebase
* fix timeout test
Co-authored-by: Pankaj Garg <pankaj@solana.com>
Benchmarks show roughly a 6% improvement. The impact could be more
significant when transactions need to be retried a lot.
after patch:
{'name': 'banking_bench_total', 'median': '72767.43'}
{'name': 'banking_bench_tx_total', 'median': '80240.38'}
{'name': 'banking_bench_success_tx_total', 'median': '72767.43'}
test bench_banking_stage_multi_accounts
... bench: 6,137,264 ns/iter (+/- 1,364,111)
test bench_banking_stage_multi_programs
... bench: 10,086,435 ns/iter (+/- 2,921,440)
before patch:
{'name': 'banking_bench_total', 'median': '68572.26'}
{'name': 'banking_bench_tx_total', 'median': '75704.75'}
{'name': 'banking_bench_success_tx_total', 'median': '68572.26'}
test bench_banking_stage_multi_accounts
... bench: 6,521,007 ns/iter (+/- 1,926,741)
test bench_banking_stage_multi_programs
... bench: 10,526,433 ns/iter (+/- 2,736,530)
- don't store pending tx signatures and costs in CostTracker
- apply tx costs to global state immediately again
- go from commit_or_cancel to update_or_remove, where the cost tracker
is either updated with the true costs for successful tx, or the costs
of a retryable tx is removed
- move the function into qos_service and hold the cost tracker lock for
the whole loop
* panic when test timeout
* nonblocking send when when droping banks
* debug log
* timeout for tvu
* unused varaible
* timeout for tpu
* Revert "debug log"
This reverts commit da780a3301a51d7c496141a85fcd35014fe6dff5.
* add timeout const
* fix typo
* Revert "nonblocking send when when droping banks".
I will create another pull request for this.
This reverts commit 088c98ec0facf825b5eca058fb860deba6d28888.
* Update core/src/tpu.rs
Co-authored-by: Trent Nelson <trent.a.b.nelson@gmail.com>
* Update core/src/tpu.rs
Co-authored-by: Trent Nelson <trent.a.b.nelson@gmail.com>
* Update core/src/tvu.rs
Co-authored-by: Trent Nelson <trent.a.b.nelson@gmail.com>
* Update core/src/tvu.rs
Co-authored-by: Trent Nelson <trent.a.b.nelson@gmail.com>
* Update core/src/validator.rs
Co-authored-by: Trent Nelson <trent.a.b.nelson@gmail.com>
Co-authored-by: Trent Nelson <trent.a.b.nelson@gmail.com>
test_skip_repair in retransmit-stage is no longer relevant because
following: https://github.com/solana-labs/solana/pull/19233
repair packets are filtered out earlier in window-service and so
retransmit stage does not know if a shred is repaired or not.
Also, following turbine peer shuffle changes:
https://github.com/solana-labs/solana/pull/24080
the test has become flaky since it does not take into account how peers
are shuffled for each shred.
behzad nouri
Tao Zhu
Jeff Washington (jwash)
Jason Davis
Christian Kamm
steviez
Tao Zhu
sakridge
Jon Cinque
Michael Vines
Giorgio Gambino
carllin
Tyera Eulberg
HaoranYi
Brooks Prumo
Jeff Biseda