* async batch verification
When batch verification is done, the main thread is blocked reducing
concurrency.
With this PR, the new thread signalling primitive in chronos is used to
offload the full batch verification process to a separate thread
allowing the main threads to continue async operations while the other
threads verify signatures.
Similar to previous behavior, the number of ongoing batch verifications
is capped to prevent runaway resource usage.
In addition to the asynchronous processing, 3 addition changes help
drive throughput:
* A loop is used for batch accumulation: this prevents a stampede of
small batches in eager mode where both the eager and the scheduled batch
runner would pick batches off the queue, prematurely picking "fresh"
batches off the queue
* An additional small wait is introduced for small batches - this helps
create slightly larger batches which make better used of the increased
concurrency
* Up to 2 batches are scheduled to the threadpool during high pressure,
reducing startup latency for the threads
Together, these changes increase attestation verification throughput
under load up to 30%.
* fixup
* Update submodules
* fix blst build issues (and a PIC warning)
* bump
---------
Co-authored-by: Zahary Karadjov <zahary@gmail.com>
This PR removes a few hundred thousand temporary seq allocations during
state transition - in particular, the flag seq was allocated per
validator while committees are computed per attestation.
Split up the `ShufflingRef` acceleration logic into generically usable
parts and attester shuffling specific parts. The generic parts could be
used to accelerate other purposes, e.g., REST `/states/xxx/randao` API.
* speed up state/block loading
When loading blocks and states from db/era, we currently redundantly
check their CRC32 - for a state, this costs 50ms of loading time
presently (110mb uncompressed size) on a decent laptop.
* remove `maxDecompressedDbRecordSize` - not actually used on recent
data since we store the framed format - also, we're in luck: we blew
past the limit quite some time ago
* fix obsolete exception-based error checking
* avoid `zeroMem` when reading from era store
see https://github.com/status-im/nim-snappy/pull/22 for benchmarks
* bump snappy
* Add new REST endpoints to monitor REST server connections and new chronos metrics.
* Bump head versions of chronos and presto.
* Bump chronos with regression fix.
* Remove outdated tests which was supposed to test pipeline mode.
* Disable pipeline mode in resttest.
* Update copyright year.
* Upgrade test_signing_node to start use AsyncProcess instead of std library's osproc.
Bump chronos to check graceful shutdown.
* Update AllTests.
* Bump chronos.
Split up the `ShufflingRef` acceleration logic into generically usable
parts and attester shuffling specific parts. The generic parts could be
used to accelerate other purposes, e.g., REST `/states/xxx/randao` API.
When producing a local block and `runProposalForkchoiceUpdated` was
missed, `getPayloadFromSingleEL` adds additional ~500ms of latency.
Quick fix to avoid missing blocks to that.
We currently call `onSlotEnd` whenever all in-BN validator duties are
completed. VC validator duties are not awaited. When `onSlotEnd` is
processed close to the slot start, a VC may therefore miss duties.
Adding a delay before `onSlotEnd` improves this situation.
The logic can be optimized further if `ActionTracker` would track
`knownValidators` from REST separately from in-process ones.
To enable additional use cases, e.g., `/states/###/randao` beacon API,
`ShufflingRef` acceleration logic needs to be able to operate on parts
of the DAG that do not have `BlockRef`. Changing `commonAncestor` to
act on `BlockId` instead of `BlockRef` is a step toward that and also
simplifies the logic some more.
* fall back to non-fcu fork choice on epoch boundaries
* Future[bool]
* fix
* Update beacon_chain/consensus_object_pools/consensus_manager.nim
Co-authored-by: Etan Kissling <etan@status.im>
* make things consistent with Opt[void] return
---------
Co-authored-by: Etan Kissling <etan@status.im>
Post-merge blocks contain all information to directly obtain RANDAO
without having to load any additional info. Take advantage of that to
further accelerate `ShufflingRef` computation. Note that it is still
necessary to verify that `blck` / `state` share a sufficiently recent
ancestor for the purpose of computing attester shufflings.
- new: 243.71s, 239.67s, 237.32s, 238.36s, 239.57s
- old: 251.33s, 234.29s, 249.28s, 237.03s, 236.78s
Current RANDAO recovery logic is quite complex as it optimizes for the
minimum amount of database reads. Loading blocks isn't the bottleneck
though, so rather make the implementation more concise by avoiding the
complex strategy planning step. Note that this also prepares for an even
faster implementation for post-merge blocks in the future that extracts
RANDAO from `ExecutionPayload` directly if available, so even in cases
where efficiency is slightly lower, only historical data is affected.
`time nim c -r tests/test_blockchain_dag` (cached binary):
- new: 145.45s, 133.59s, 144.65s, 127.69s, 136.14s
- old: 149.15s, 150.84s, 135.77s, 137.49s, 133.89s
* Perform block pre-check before validating execution
When syncing, blocks have not been gossip-validated and are therefore
prone to trivial faults like being known-unviable, duplicate or missing
their parent.
In addition, the duplicate-block check in BlockProcessor was not
considering the quarantine flow and would therefore cause
recently-quarantined blocks to be silenty dropped when their parent
appears delaying the sync end-game and thus causing longer startup
resync time.
This PR verifies trivial conditions before performing execution
validation thus avoiding duplicates and missing parents alike.
It also ensures that the fast-sync EL mode is used for finalized blocks
even if the EL is timing out / slow to respond - this allows the CL to
complete its sync faster and switch to "normal" lock-step at the head of
the chain more quickly, thus also allowing the EL to access the latest
consensensus information earlier.
* oops
* remove unused constant
When the requestmanager is busy fetching blocks, the queue might get
filled with multiple entries of the same root - since there is no
deduplication, requests containing the same root multiple times will be
sent out.
Also, because the items sit in the queue for a long time potentially,
the request might be stale by the time that the manager is ready with
the previous request.
This PR removes the queue and directly fetches the blocks to download
from the quarantine which solves both problems (the quarantine already
de-duplicates and is clean of stale information).
Removing the queue for blobs is left for a future PR.
Co-authored-by: tersec <tersec@users.noreply.github.com>
* early exit `commonAncestor` when comparing with `finalizedHead`
As all `BlockRef` lead to `finalizedHead` (`parent == nil`),
can shortcut in that situation and immediately return `finalizedHead`
if passed as one of the arguments.
* typo in comment
* add test from #5152
Co-authored-by: tersec <tersec@users.noreply.github.com>
* add note about test complexity
* regenerate test summary
---------
Co-authored-by: tersec <tersec@users.noreply.github.com>
tersec
Jacek Sieka
Etan Kissling
henridf
Philippe Schommers
Eugene Kabanov
Zahary Karadjov