Cost model 1.7 (#20188)

* Cost Model to limit transactions which are not parallelizeable (#16694)

* * Add following to banking_stage:
  1. CostModel as immutable ref shared between threads, to provide estimated cost for transactions.
  2. CostTracker which is shared between threads, tracks transaction costs for each block.

* replace hard coded program ID with id() calls

* Add Account Access Cost as part of TransactionCost. Account Access cost are weighted differently between read and write, signed and non-signed.

* Establish instruction_execution_cost_table, add function to update or insert instruction cost, unit tested. It is read-only for now; it allows Replay to insert realtime instruction execution costs to the table.

* add test for cost_tracker atomically try_add operation, serves as safety guard for future changes

* check cost against local copy of cost_tracker, return transactions that would exceed limit as unprocessed transaction to be buffered; only apply bank processed transactions cost to tracker;

* bencher to new banking_stage with max cost limit to allow cost model being hit consistently during bench iterations

* replay stage feed back program cost (#17731)

* replay stage feeds back realtime per-program execution cost to cost model;

* program cost execution table is initialized into empty table, no longer populated with hardcoded numbers;

* changed cost unit to microsecond, using value collected from mainnet;

* add ExecuteCostTable with fixed capacity for security concern, when its limit is reached, programs with old age AND less occurrence will be pushed out to make room for new programs.

* investigate system performance test degradation  (#17919)

* Add stats and counter around cost model ops, mainly:
- calculate transaction cost
- check transaction can fit in a block
- update block cost tracker after transactions are added to block
- replay_stage to update/insert execution cost to table

* Change mutex on cost_tracker to RwLock

* removed cloning cost_tracker for local use, as the metrics show clone is very expensive.

* acquire and hold locks for block of TXs, instead of acquire and release per transaction;

* remove redundant would_fit check from cost_tracker update execution path

* refactor cost checking with less frequent lock acquiring

* avoid many Transaction_cost heap allocation when calculate cost, which
is in the hot path - executed per transaction.

* create hashmap with new_capacity to reduce runtime heap realloc.

* code review changes: categorize stats, replace explicit drop calls, concisely initiate to default

* address potential deadlock by acquiring locks one at time

* Persist cost table to blockstore (#18123)

* Add `ProgramCosts` Column Family to blockstore, implement LedgerColumn; add `delete_cf` to Rocks
* Add ProgramCosts to compaction excluding list alone side with TransactionStatusIndex in one place: `excludes_from_compaction()`

* Write cost table to blockstore after `replay_stage` replayed active banks; add stats to measure persist time
* Deletes program from `ProgramCosts` in blockstore when they are removed from cost_table in memory
* Only try to persist to blockstore when cost_table is changed.
* Restore cost table during validator startup

* Offload `cost_model` related operations from replay main thread to dedicated service thread, add channel to send execute_timings between these threads;
* Move `cost_update_service` to its own module; replay_stage is now decoupled from cost_model.

* log warning when channel send fails (#18391)

* Aggregate cost_model into cost_tracker (#18374)

* * aggregate cost_model into cost_tracker, decouple it from banking_stage to prevent accidental deadlock. * Simplified code, removed unused functions

* review fixes

* update ledger tool to restore cost table from blockstore (#18489)

* update ledger tool to restore cost model from blockstore when compute-slot-cost

* Move initialize_cost_table into cost_model, so the function can be tested and shared between validator and ledger-tool

* refactor and simplify a test

* manually fix merge conflicts

* Per-program id timings (#17554)

* more manual fixing

* solve a merge conflict

* featurize cost model

* more merge fix

* cost model uses compute_unit to replace microsecond as cost unit
(#18934)

* Reject blocks for costs above the max block cost (#18994)

* Update block max cost limit to fix performance regession (#19276)

* replace function with const var for better readability (#19285)

* Add few more metrics data points (#19624)

* periodically report sigverify_stage stats (#19674)

* manual merge

* cost model nits (#18528)

* Accumulate consumed units (#18714)

* tx wide compute budget (#18631)

* more manual merge

* ignore zerorize drop security

* - update const cost values with data collected by #19627
- update cost calculation to closely proposed fee schedule #16984

* add transaction cost histogram metrics (#20350)

* rebase to 1.7.15

* add tx count and thread id to stats (#20451)
each stat reports and resets when slot changes

* remove cost_model feature_set

* ignore vote transactions from cost model

Co-authored-by: sakridge <sakridge@gmail.com>
Co-authored-by: Jeff Biseda <jbiseda@gmail.com>
Co-authored-by: Jack May <jack@solana.com>
This commit is contained in:
Tao Zhu
2021-10-06 15:11:41 -05:00
committed by Trent Nelson
parent a4df784e82
commit db85d659b9
40 changed files with 3208 additions and 266 deletions

View File

@@ -0,0 +1,292 @@
//! this service receives instruction ExecuteTimings from replay_stage,
//! update cost_model which is shared with banking_stage to optimize
//! packing transactions into block; it also triggers persisting cost
//! table to blockstore.
use crate::cost_model::CostModel;
use solana_ledger::blockstore::Blockstore;
use solana_measure::measure::Measure;
use solana_runtime::bank::ExecuteTimings;
use solana_sdk::timing::timestamp;
use std::{
sync::{
atomic::{AtomicBool, Ordering},
mpsc::Receiver,
Arc, RwLock,
},
thread::{self, Builder, JoinHandle},
time::Duration,
};
#[derive(Default)]
pub struct CostUpdateServiceTiming {
last_print: u64,
update_cost_model_count: u64,
update_cost_model_elapsed: u64,
persist_cost_table_elapsed: u64,
}
impl CostUpdateServiceTiming {
fn update(
&mut self,
update_cost_model_count: u64,
update_cost_model_elapsed: u64,
persist_cost_table_elapsed: u64,
) {
self.update_cost_model_count += update_cost_model_count;
self.update_cost_model_elapsed += update_cost_model_elapsed;
self.persist_cost_table_elapsed += persist_cost_table_elapsed;
let now = timestamp();
let elapsed_ms = now - self.last_print;
if elapsed_ms > 1000 {
datapoint_info!(
"cost-update-service-stats",
("total_elapsed_us", elapsed_ms * 1000, i64),
(
"update_cost_model_count",
self.update_cost_model_count as i64,
i64
),
(
"update_cost_model_elapsed",
self.update_cost_model_elapsed as i64,
i64
),
(
"persist_cost_table_elapsed",
self.persist_cost_table_elapsed as i64,
i64
),
);
*self = CostUpdateServiceTiming::default();
self.last_print = now;
}
}
}
pub type CostUpdateReceiver = Receiver<ExecuteTimings>;
pub struct CostUpdateService {
thread_hdl: JoinHandle<()>,
}
impl CostUpdateService {
#[allow(clippy::new_ret_no_self)]
pub fn new(
exit: Arc<AtomicBool>,
blockstore: Arc<Blockstore>,
cost_model: Arc<RwLock<CostModel>>,
cost_update_receiver: CostUpdateReceiver,
) -> Self {
let thread_hdl = Builder::new()
.name("solana-cost-update-service".to_string())
.spawn(move || {
Self::service_loop(exit, blockstore, cost_model, cost_update_receiver);
})
.unwrap();
Self { thread_hdl }
}
pub fn join(self) -> thread::Result<()> {
self.thread_hdl.join()
}
fn service_loop(
exit: Arc<AtomicBool>,
blockstore: Arc<Blockstore>,
cost_model: Arc<RwLock<CostModel>>,
cost_update_receiver: CostUpdateReceiver,
) {
let mut cost_update_service_timing = CostUpdateServiceTiming::default();
let mut dirty: bool;
let mut update_count: u64;
let wait_timer = Duration::from_millis(100);
loop {
if exit.load(Ordering::Relaxed) {
break;
}
dirty = false;
update_count = 0_u64;
let mut update_cost_model_time = Measure::start("update_cost_model_time");
for cost_update in cost_update_receiver.try_iter() {
dirty |= Self::update_cost_model(&cost_model, &cost_update);
update_count += 1;
}
update_cost_model_time.stop();
let mut persist_cost_table_time = Measure::start("persist_cost_table_time");
if dirty {
Self::persist_cost_table(&blockstore, &cost_model);
}
persist_cost_table_time.stop();
cost_update_service_timing.update(
update_count,
update_cost_model_time.as_us(),
persist_cost_table_time.as_us(),
);
thread::sleep(wait_timer);
}
}
fn update_cost_model(cost_model: &RwLock<CostModel>, execute_timings: &ExecuteTimings) -> bool {
let mut dirty = false;
{
let mut cost_model_mutable = cost_model.write().unwrap();
for (program_id, timing) in &execute_timings.details.per_program_timings {
if timing.count < 1 {
continue;
}
let units = timing.accumulated_units / timing.count as u64;
match cost_model_mutable.upsert_instruction_cost(program_id, units) {
Ok(c) => {
debug!(
"after replayed into bank, instruction {:?} has averaged cost {}",
program_id, c
);
dirty = true;
}
Err(err) => {
debug!(
"after replayed into bank, instruction {:?} failed to update cost, err: {}",
program_id, err
);
}
}
}
}
debug!(
"after replayed into bank, updated cost model instruction cost table, current values: {:?}",
cost_model.read().unwrap().get_instruction_cost_table()
);
dirty
}
fn persist_cost_table(blockstore: &Blockstore, cost_model: &RwLock<CostModel>) {
let cost_model_read = cost_model.read().unwrap();
let cost_table = cost_model_read.get_instruction_cost_table();
let db_records = blockstore.read_program_costs().expect("read programs");
// delete records from blockstore if they are no longer in cost_table
db_records.iter().for_each(|(pubkey, _)| {
if cost_table.get(pubkey).is_none() {
blockstore
.delete_program_cost(pubkey)
.expect("delete old program");
}
});
for (key, cost) in cost_table.iter() {
blockstore
.write_program_cost(key, cost)
.expect("persist program costs to blockstore");
}
}
}
#[cfg(test)]
mod tests {
use super::*;
use solana_runtime::message_processor::ProgramTiming;
use solana_sdk::pubkey::Pubkey;
#[test]
fn test_update_cost_model_with_empty_execute_timings() {
let cost_model = Arc::new(RwLock::new(CostModel::default()));
let empty_execute_timings = ExecuteTimings::default();
CostUpdateService::update_cost_model(&cost_model, &empty_execute_timings);
assert_eq!(
0,
cost_model
.read()
.unwrap()
.get_instruction_cost_table()
.len()
);
}
#[test]
fn test_update_cost_model_with_execute_timings() {
let cost_model = Arc::new(RwLock::new(CostModel::default()));
let mut execute_timings = ExecuteTimings::default();
let program_key_1 = Pubkey::new_unique();
let mut expected_cost: u64;
// add new program
{
let accumulated_us: u64 = 1000;
let accumulated_units: u64 = 100;
let count: u32 = 10;
expected_cost = accumulated_units / count as u64;
execute_timings.details.per_program_timings.insert(
program_key_1,
ProgramTiming {
accumulated_us,
accumulated_units,
count,
},
);
CostUpdateService::update_cost_model(&cost_model, &execute_timings);
assert_eq!(
1,
cost_model
.read()
.unwrap()
.get_instruction_cost_table()
.len()
);
assert_eq!(
Some(&expected_cost),
cost_model
.read()
.unwrap()
.get_instruction_cost_table()
.get(&program_key_1)
);
}
// update program
{
let accumulated_us: u64 = 2000;
let accumulated_units: u64 = 200;
let count: u32 = 10;
// to expect new cost is Average(new_value, existing_value)
expected_cost = ((accumulated_units / count as u64) + expected_cost) / 2;
execute_timings.details.per_program_timings.insert(
program_key_1,
ProgramTiming {
accumulated_us,
accumulated_units,
count,
},
);
CostUpdateService::update_cost_model(&cost_model, &execute_timings);
assert_eq!(
1,
cost_model
.read()
.unwrap()
.get_instruction_cost_table()
.len()
);
assert_eq!(
Some(&expected_cost),
cost_model
.read()
.unwrap()
.get_instruction_cost_table()
.get(&program_key_1)
);
}
}
}