nimbus-eth1/nimbus/db/aristo/aristo_delta/delta_reverse.nim

105 lines
3.0 KiB
Nim
Raw Normal View History

# nimbus-eth1
# Copyright (c) 2023-2024 Status Research & Development GmbH
# Licensed under either of
# * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE) or
# http://www.apache.org/licenses/LICENSE-2.0)
# * MIT license ([LICENSE-MIT](LICENSE-MIT) or
# http://opensource.org/licenses/MIT)
# at your option. This file may not be copied, modified, or distributed
# except according to those terms.
import
std/tables,
eth/common,
results,
".."/[aristo_desc, aristo_get, aristo_utils]
# ------------------------------------------------------------------------------
# Private functions
# ------------------------------------------------------------------------------
proc revSubTree(
db: AristoDbRef;
rev: LayerRef;
rvid: RootedVertexID;
): Result[void,(VertexID,AristoError)] =
## Collect subtrees marked for deletion
let
vtx = block:
let rc = db.getVtxUbe rvid
if rc.isOk:
rc.value
elif rc.error == GetVtxNotFound:
VertexRef(nil)
else:
return err((rvid.vid,rc.error))
key = block:
Pre-allocate vids for branches (#2882) Each branch node may have up to 16 sub-items - currently, these are given VertexID based when they are first needed leading to a mostly-random order of vertexid for each subitem. Here, we pre-allocate all 16 vertex ids such that when a branch subitem is filled, it already has a vertexid waiting for it. This brings several important benefits: * subitems are sorted and "close" in their id sequencing - this means that when rocksdb stores them, they are likely to end up in the same data block thus improving read efficiency * because the ids are consequtive, we can store just the starting id and a bitmap representing which subitems are in use - this reduces disk space usage for branches allowing more of them fit into a single disk read, further improving disk read and caching performance - disk usage at block 18M is down from 84 to 78gb! * the in-memory footprint of VertexRef reduced allowing more instances to fit into caches and less memory to be used overall. Because of the increased locality of reference, it turns out that we no longer need to iterate over the entire database to efficiently generate the hash key database because the normal computation is now faster - this significantly benefits "live" chain processing as well where each dirtied key must be accompanied by a read of all branch subitems next to it - most of the performance benefit in this branch comes from this locality-of-reference improvement. On a sample resync, there's already ~20% improvement with later blocks seeing increasing benefit (because the trie is deeper in later blocks leading to more benefit from branch read perf improvements) ``` blocks: 18729664, baseline: 190h43m49s, contender: 153h59m0s Time (total): -36h44m48s, -19.27% ``` Note: clients need to be resynced as the PR changes the on-disk format R.I.P. little bloom filter - your life in the repo was short but valuable
2024-12-04 10:42:04 +00:00
let rc = db.getKeyUbe(rvid, {})
if rc.isOk:
Pre-allocate vids for branches (#2882) Each branch node may have up to 16 sub-items - currently, these are given VertexID based when they are first needed leading to a mostly-random order of vertexid for each subitem. Here, we pre-allocate all 16 vertex ids such that when a branch subitem is filled, it already has a vertexid waiting for it. This brings several important benefits: * subitems are sorted and "close" in their id sequencing - this means that when rocksdb stores them, they are likely to end up in the same data block thus improving read efficiency * because the ids are consequtive, we can store just the starting id and a bitmap representing which subitems are in use - this reduces disk space usage for branches allowing more of them fit into a single disk read, further improving disk read and caching performance - disk usage at block 18M is down from 84 to 78gb! * the in-memory footprint of VertexRef reduced allowing more instances to fit into caches and less memory to be used overall. Because of the increased locality of reference, it turns out that we no longer need to iterate over the entire database to efficiently generate the hash key database because the normal computation is now faster - this significantly benefits "live" chain processing as well where each dirtied key must be accompanied by a read of all branch subitems next to it - most of the performance benefit in this branch comes from this locality-of-reference improvement. On a sample resync, there's already ~20% improvement with later blocks seeing increasing benefit (because the trie is deeper in later blocks leading to more benefit from branch read perf improvements) ``` blocks: 18729664, baseline: 190h43m49s, contender: 153h59m0s Time (total): -36h44m48s, -19.27% ``` Note: clients need to be resynced as the PR changes the on-disk format R.I.P. little bloom filter - your life in the repo was short but valuable
2024-12-04 10:42:04 +00:00
rc.value[0]
elif rc.error == GetKeyNotFound:
VOID_HASH_KEY
else:
return err((rvid.vid,rc.error))
if vtx.isValid:
for vid in vtx.subVids:
? db.revSubTree(rev, (rvid.root,vid))
rev.sTab[rvid] = vtx
if key.isValid:
rev.kMap[rvid] = key
ok()
# ------------------------------------------------------------------------------
# Public functions
# ------------------------------------------------------------------------------
proc revFilter*(
db: AristoDbRef; # Database
filter: LayerRef; # Filter to revert
): Result[LayerRef,(VertexID,AristoError)] =
## Assemble reverse filter for the `filter` argument, i.e. changes to the
## backend that reverse the effect of applying the this read-only filter.
##
## This read-only filter is calculated against the current unfiltered
## backend (excluding optionally installed read-only filter.)
##
let rev = LayerRef()
# Get vid generator state on backend
block:
2024-06-04 15:05:13 +00:00
let rc = db.getTuvUbe()
if rc.isOk:
2024-06-04 15:05:13 +00:00
rev.vTop = rc.value
elif rc.error != GetTuvNotFound:
return err((VertexID(0), rc.error))
# Calculate reverse changes for the `sTab[]` structural table
for rvid in filter.sTab.keys:
let rc = db.getVtxUbe rvid
if rc.isOk:
rev.sTab[rvid] = rc.value
elif rc.error == GetVtxNotFound:
rev.sTab[rvid] = VertexRef(nil)
else:
return err((rvid.vid,rc.error))
# Calculate reverse changes for the `kMap[]` structural table.
for rvid in filter.kMap.keys:
Pre-allocate vids for branches (#2882) Each branch node may have up to 16 sub-items - currently, these are given VertexID based when they are first needed leading to a mostly-random order of vertexid for each subitem. Here, we pre-allocate all 16 vertex ids such that when a branch subitem is filled, it already has a vertexid waiting for it. This brings several important benefits: * subitems are sorted and "close" in their id sequencing - this means that when rocksdb stores them, they are likely to end up in the same data block thus improving read efficiency * because the ids are consequtive, we can store just the starting id and a bitmap representing which subitems are in use - this reduces disk space usage for branches allowing more of them fit into a single disk read, further improving disk read and caching performance - disk usage at block 18M is down from 84 to 78gb! * the in-memory footprint of VertexRef reduced allowing more instances to fit into caches and less memory to be used overall. Because of the increased locality of reference, it turns out that we no longer need to iterate over the entire database to efficiently generate the hash key database because the normal computation is now faster - this significantly benefits "live" chain processing as well where each dirtied key must be accompanied by a read of all branch subitems next to it - most of the performance benefit in this branch comes from this locality-of-reference improvement. On a sample resync, there's already ~20% improvement with later blocks seeing increasing benefit (because the trie is deeper in later blocks leading to more benefit from branch read perf improvements) ``` blocks: 18729664, baseline: 190h43m49s, contender: 153h59m0s Time (total): -36h44m48s, -19.27% ``` Note: clients need to be resynced as the PR changes the on-disk format R.I.P. little bloom filter - your life in the repo was short but valuable
2024-12-04 10:42:04 +00:00
let rc = db.getKeyUbe(rvid, {})
if rc.isOk:
Pre-allocate vids for branches (#2882) Each branch node may have up to 16 sub-items - currently, these are given VertexID based when they are first needed leading to a mostly-random order of vertexid for each subitem. Here, we pre-allocate all 16 vertex ids such that when a branch subitem is filled, it already has a vertexid waiting for it. This brings several important benefits: * subitems are sorted and "close" in their id sequencing - this means that when rocksdb stores them, they are likely to end up in the same data block thus improving read efficiency * because the ids are consequtive, we can store just the starting id and a bitmap representing which subitems are in use - this reduces disk space usage for branches allowing more of them fit into a single disk read, further improving disk read and caching performance - disk usage at block 18M is down from 84 to 78gb! * the in-memory footprint of VertexRef reduced allowing more instances to fit into caches and less memory to be used overall. Because of the increased locality of reference, it turns out that we no longer need to iterate over the entire database to efficiently generate the hash key database because the normal computation is now faster - this significantly benefits "live" chain processing as well where each dirtied key must be accompanied by a read of all branch subitems next to it - most of the performance benefit in this branch comes from this locality-of-reference improvement. On a sample resync, there's already ~20% improvement with later blocks seeing increasing benefit (because the trie is deeper in later blocks leading to more benefit from branch read perf improvements) ``` blocks: 18729664, baseline: 190h43m49s, contender: 153h59m0s Time (total): -36h44m48s, -19.27% ``` Note: clients need to be resynced as the PR changes the on-disk format R.I.P. little bloom filter - your life in the repo was short but valuable
2024-12-04 10:42:04 +00:00
rev.kMap[rvid] = rc.value[0]
elif rc.error == GetKeyNotFound:
rev.kMap[rvid] = VOID_HASH_KEY
else:
return err((rvid.vid,rc.error))
ok(rev)
# ------------------------------------------------------------------------------
# End
# ------------------------------------------------------------------------------