Baguette/dodb.cr
3
0
Fork 1
Code Issues Pull Requests Releases Wiki Activity

FIFO + paper

Browse Source
This commit is contained in:
Philippe PITTOLI 2024-05-24 00:25:28 +02:00
parent dc34e5b881
commit fefc4e9b26
9 changed files with 146 additions and 58 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
Makefile
View File

@ -5,9 +5,15 @@ Q ?= @
SHOULD_UPDATE = ./bin/should-update
DBDIR=/tmp/tests-on-dodb
RESULTS_DIR=results
benchmark-cars:
$(Q)crystal build spec/benchmark-cars.cr $(OPTS) --release
benchmark-cars-run: benchmark-cars
./benchmark-cars search # by default, test search durations
./bin/stats.sh $(RESULTS_DIR)
./bin/extract-data-benchmark-cars.sh $(RESULTS_DIR)
build: benchmark-cars
wipe-db:

21
paper/graph_query_partition.grap
View File

@ -7,8 +7,8 @@ ticks left out from 0 to 350 by 50
label left "Request duration" unaligned "for a partition (ms)" "(Median)" left 0.8
label bot "Number of cars matching the partition" down 0.1
obram = obuncache = obcache = obsemi = 0
cbram = cbuncache = cbcache = cbsemi = 0
obram = obuncache = obfifo = obcache = obsemi = 0
cbram = cbuncache = cbfifo = cbcache = cbsemi = 0
legendxleft = 1000
legendxright = 6500
@ -23,21 +23,26 @@ copy "../data/partitions.d" thru X
y_scale = 1000000
# ram cached semi uncached
# ram cached fifo semi uncached
line from cx,$2/y_scale to cx,$4/y_scale
line from cx,$5/y_scale to cx,$7/y_scale
line from cx,$8/y_scale to cx,$10/y_scale
line from cx,$11/y_scale to cx,$13/y_scale
line from cx,$14/y_scale to cx,$16/y_scale
#ty = $3
cbram = $3/y_scale
cbcache = $6/y_scale
cbsemi = $9/y_scale
cbuncache = $12/y_scale
cbfifo = $9/y_scale
cbsemi = $12/y_scale
cbuncache = $15/y_scale
if (obram > 0) then {line from cx,cbram to ox,obram}
if (obcache > 0) then {line from cx,cbcache to ox,obcache}
.gcolor pink
if (obfifo > 0) then {line from cx,cbfifo to ox,obfifo}
.gcolor
.gcolor blue
if (obsemi > 0) then {line from cx,cbsemi to ox,obsemi}
.gcolor
@ -47,15 +52,19 @@ copy "../data/partitions.d" thru X
obram = cbram
obcache = cbcache
obfifo = cbfifo
obsemi = cbsemi
obuncache = cbuncache
ox = cx
# ram cached semi uncached
# ram cached fifo semi uncached
.gcolor red
bullet at cx,cbram
.gcolor
bullet at cx,cbcache
.gcolor pink
bullet at cx,cbfifo
.gcolor
.gcolor blue
bullet at cx,cbsemi
.gcolor

21
paper/graph_query_tag.grap
View File

@ -6,8 +6,8 @@ ticks left out from 0 to 170 by 20
label left "Request duration" unaligned "for a tag (ms)" "(Median)" left 0.8
label bot "Number of cars matching the tag" down 0.1
obram = obuncache = obcache = obsemi = 0
cbram = cbuncache = cbcache = cbsemi = 0
obram = obuncache = obfifo = obcache = obsemi = 0
cbram = cbuncache = cbfifo = cbcache = cbsemi = 0
legendxleft = 200
legendxright = 3000
@ -22,21 +22,26 @@ copy "../data/tags.d" thru X
y_scale = 1000000
# ram cached semi uncached
# ram cached fifo semi uncached
line from cx,$2/y_scale to cx,$4/y_scale
line from cx,$5/y_scale to cx,$7/y_scale
line from cx,$8/y_scale to cx,$10/y_scale
line from cx,$11/y_scale to cx,$13/y_scale
line from cx,$14/y_scale to cx,$16/y_scale
#ty = $3
cbram = $3/y_scale
cbcache = $6/y_scale
cbsemi = $9/y_scale
cbuncache = $12/y_scale
cbfifo = $9/y_scale
cbsemi = $12/y_scale
cbuncache = $15/y_scale
if (obram > 0) then {line from cx,cbram to ox,obram}
if (obcache > 0) then {line from cx,cbcache to ox,obcache}
.gcolor pink
if (obfifo > 0) then {line from cx,cbfifo to ox,obfifo}
.gcolor
.gcolor blue
if (obsemi > 0) then {line from cx,cbsemi to ox,obsemi}
.gcolor
@ -46,15 +51,19 @@ copy "../data/tags.d" thru X
obram = cbram
obcache = cbcache
obfifo = cbfifo
obsemi = cbsemi
obuncache = cbuncache
ox = cx
# ram cached semi uncached
# ram cached fifo semi uncached
.gcolor red
bullet at cx,cbram
.gcolor
bullet at cx,cbcache
.gcolor pink
bullet at cx,cbfifo
.gcolor
.gcolor blue
bullet at cx,cbsemi
.gcolor

9
paper/legend.grap
View File

@ -19,7 +19,7 @@ define legend {
diffx = xright - xleft
diffy = yup - ydown
hdiff = diffy/4.3
hdiff = diffy/5.7
cy = yup - (diffy/6)
cx = (diffx/20) + xleft
@ -27,6 +27,7 @@ define legend {
lendx = cx + diffx/8
tstartx = lendx + diffx/20
.ps -2
.gcolor red
line from lstartx,cy to lendx,cy
.gcolor
@ -35,6 +36,11 @@ define legend {
line from lstartx,cy to lendx,cy
"Cached db and index" ljust at tstartx,cy
cy = cy - hdiff
.gcolor pink
line from lstartx,cy to lendx,cy
.gcolor
"FIFO db and cached index" ljust at tstartx,cy
cy = cy - hdiff
.gcolor blue
line from lstartx,cy to lendx,cy
.gcolor
@ -44,4 +50,5 @@ define legend {
line from lstartx,cy to lendx,cy
.gcolor
"Uncached db and index" ljust at tstartx,cy
.ps +2
}

104
spec/benchmark-cars.cr
View File

@ -13,13 +13,6 @@ require "./db-cars.cr"
# ENV["NBRUN"] rescue 100
# ENV["MAXINDEXES"] rescue 5_000
class DODB::Storage(V)
def empty_db
while pop
end
end
end
class Context
class_property report_dir = "results"
class_property max_indexes = 5_000
@ -85,10 +78,36 @@ def prepare_env(storage, name, s_index, s_partition, s_tags, &)
long_operation "removing #{name} data" { storage.rm_storage_dir }
end
def batch()
def search_benchmark(storage : DODB::Storage(Car),
current_db_size : Int32,
name : String,
search_name : DODB::Index::Basic(Car),
search_color : DODB::Index::Partition(Car),
search_keywords : DODB::Index::Tags(Car))
name_to_search = ENV["CARNAME"] rescue "Corvet-#{(current_db_size/2).to_i}"
color_to_search = ENV["CARCOLOR"] rescue "red"
keyword_to_search = ENV["CARKEYWORD"] rescue "spacious"
puts "NEW BATCH: db-size #{current_db_size}, name: '#{name_to_search}', color: '#{color_to_search}', tag: '#{keyword_to_search}'"
report(storage, "#{name}_#{current_db_size}_index") do
corvet = search_name.get name_to_search
end
if current_db_size <= Context.max_indexes
report(storage, "#{name}_#{current_db_size}_partitions") do
corvet = search_color.get? color_to_search
end
report(storage, "#{name}_#{current_db_size}_tags") do
corvet = search_keywords.get? keyword_to_search
end
else
fake_report("#{name}_#{current_db_size}_partitions")
fake_report("#{name}_#{current_db_size}_tags")
end
end
def bench_searches()
cars_ram = SPECDB::RAMOnly(Car).new
cars_cached = SPECDB::Cached(Car).new
cars_fifo = DODB::FIFOSpecDataBase(Car).new
cars_fifo = SPECDB::FIFO(Car).new "", 5000 # With only 5_000 entries
cars_semi = SPECDB::Uncached(Car).new "-semi"
cars_uncached = SPECDB::Uncached(Car).new
@ -98,32 +117,7 @@ def batch()
semi_Sby_name, semi_Sby_color, semi_Sby_keywords = cached_indexes cars_semi
uncached_Sby_name, uncached_Sby_color, uncached_Sby_keywords = uncached_indexes cars_uncached
fn = ->(storage : DODB::Storage(Car),
current_db_size : Int32,
name : String,
search_name : DODB::Index::Basic(Car),
search_color : DODB::Index::Partition(Car),
search_keywords : DODB::Index::Tags(Car)) {
name_to_search = ENV["CARNAME"] rescue "Corvet-#{(current_db_size/2).to_i}"
color_to_search = ENV["CARCOLOR"] rescue "red"
keyword_to_search = ENV["CARKEYWORD"] rescue "spacious"
puts "NEW BATCH: db-size #{current_db_size}, name: '#{name_to_search}', color: '#{color_to_search}', tag: '#{keyword_to_search}'"
report(storage, "#{name}_#{current_db_size}_index") do
corvet = search_name.get name_to_search
end
if current_db_size <= Context.max_indexes
report(storage, "#{name}_#{current_db_size}_partitions") do
corvet = search_color.get? color_to_search
end
report(storage, "#{name}_#{current_db_size}_tags") do
corvet = search_keywords.get? keyword_to_search
end
else
fake_report("#{name}_#{current_db_size}_partitions")
fake_report("#{name}_#{current_db_size}_tags")
end
}
fn = ->search_benchmark(DODB::Storage(Car), Int32, String, DODB::Index::Basic(Car), DODB::Index::Partition(Car), DODB::Index::Tags(Car))
prepare_env cars_ram, "ram", ram_Sby_name, ram_Sby_color, ram_Sby_keywords, &fn
prepare_env cars_cached, "cached", cached_Sby_name, cached_Sby_color, cached_Sby_keywords, &fn
@ -143,14 +137,16 @@ def perform_add(storage : DODB::Storage(Car))
end
end
def batch_add()
def bench_add()
cars_ram = SPECDB::RAMOnly(Car).new
cars_cached = SPECDB::Cached(Car).new
cars_fifo = SPECDB::FIFO(Car).new "", 5_000
cars_semi = SPECDB::Uncached(Car).new "-semi"
cars_uncached = SPECDB::Uncached(Car).new
ram_indexes cars_ram
cached_indexes cars_cached
cached_indexes cars_fifo
cached_indexes cars_semi
uncached_indexes cars_uncached
@ -160,6 +156,9 @@ def batch_add()
avr = perform_add(cars_cached)
puts "(cached db and indexes) add a value (average on #{Context.nb_run} tries): #{avr}"
avr = perform_add(cars_fifo)
puts "(fifo db and cached indexes) add a value (average on #{Context.nb_run} tries): #{avr}"
avr = perform_add(cars_semi)
puts "(uncached db but cached indexes) add a value (average on #{Context.nb_run} tries): #{avr}"
@ -172,6 +171,22 @@ def batch_add()
cars_uncached.rm_storage_dir
end
def bench_50_shades_of_fifo()
cars_fifo1 = SPECDB::FIFO(Car).new "", 1_000
cars_fifo5 = SPECDB::FIFO(Car).new "", 5_000
cars_fifo10 = SPECDB::FIFO(Car).new "", 10_000
fifo_Sby_name1, fifo_Sby_color1, fifo_Sby_keywords1 = cached_indexes cars_fifo1
fifo_Sby_name5, fifo_Sby_color5, fifo_Sby_keywords5 = cached_indexes cars_fifo5
fifo_Sby_name10, fifo_Sby_color10, fifo_Sby_keywords10 = cached_indexes cars_fifo10
fn = ->search_benchmark(DODB::Storage(Car), Int32, String, DODB::Index::Basic(Car), DODB::Index::Partition(Car), DODB::Index::Tags(Car))
prepare_env cars_fifo1, "fifo1", fifo_Sby_name1, fifo_Sby_color1, fifo_Sby_keywords1, &fn
prepare_env cars_fifo5, "fifo5", fifo_Sby_name5, fifo_Sby_color5, fifo_Sby_keywords5, &fn
prepare_env cars_fifo10, "fifo10", fifo_Sby_name10, fifo_Sby_color10, fifo_Sby_keywords10, &fn
end
ENV["REPORT_DIR"]?.try { |report_dir| Context.report_dir = report_dir }
Dir.mkdir_p Context.report_dir
@ -187,5 +202,18 @@ pp! Context.to
pp! Context.incr
pp! Context.max_indexes
batch
batch_add
if ARGV.size == 0
puts "Usage: benchmark-cars (fifo|search|add)"
exit 0
end
case ARGV[0]
when /fifo/
bench_50_shades_of_fifo
when /search/
bench_searches
when /add/
bench_add
else
puts "Usage: benchmark-cars (fifo|search|add)"
end

6
spec/spec-database.cr
View File

@ -24,10 +24,10 @@ class SPECDB::Cached(V) < DODB::Storage::Cached(V)
end
end
class DODB::FIFOSpecDataBase(V) < DODB::Storage::Stacked(V)
class SPECDB::FIFO(V) < DODB::Storage::Stacked(V)
property storage_dir : String
def initialize(storage_ext = "", remove_previous_data = true)
@storage_dir = "specdb-storage-fifo#{storage_ext}"
def initialize(storage_ext = "", @max_entries = 100_000, remove_previous_data = true)
@storage_dir = "specdb-storage-fifo-#{@max_entries}#{storage_ext}"
::FileUtils.rm_rf storage_dir if remove_previous_data
super storage_dir
end

1
src/dodb/storage/cached.cr
View File

@ -73,7 +73,6 @@ class DODB::Storage::Cached(V) < DODB::Storage(V)
@data[key] rescue raise MissingEntry.new(key)
end
# :inherit:
def []=(key : Int32, value : V)
old_value = self.[key]?

29
src/dodb/storage/stacked.cr
View File

@ -31,7 +31,6 @@
# ```
#
# NOTE: fast for frequently requested data and requires a stable (and configurable) amount of memory.
# TODO: not yet implemented.
class DODB::Storage::Stacked(V) < DODB::Storage::Cached(V)
# The *stack* a simple FIFO instance where the key of the requested data is pushed.
# In case the number of stored entries exceeds what is allowed, the least recently used entry is removed.
@ -52,9 +51,35 @@ class DODB::Storage::Stacked(V) < DODB::Storage::Cached(V)
def [](key : Int32) : V
val = @data[key] rescue raise MissingEntry.new(key)
push_fifo key
val
end
# Assumes new entries are more requested than old ones.
def []=(key : Int32, value : V)
super key, value
push_fifo key
end
# :inherit:
#
# Assumes new entries are more requested than old ones.
def <<(item : V)
key = super item
push_fifo key
end
def unsafe_delete(key : Int32)
@stack.delete key if super key
end
def delete(key : Int32)
@stack.delete key if super key
end
private def push_fifo(key : Int32)
if entry_to_remove = @stack << key
@data.delete entry_to_remove
end
val
end
end

7
src/fifo.cr
View File

@ -27,11 +27,16 @@ class FIFO(V)
end
# Pushes a value in the FIFO and gets the oldest value whether it exceeds the allowed number of entries.
# NOTE: `#<<(v : V)` is the only function since it's enough for the intended use, feel free to improve this.
# NOTE: `#<<(v : V)` is (almost) the only function since it's enough for the intended use, feel free to improve this.
# WARNING: implementation is extremely simple (3 lines) and not designed to be highly efficient.
def <<(v : V) : V?
@data.select! { |x| v != v } # remove dups
@data.unshift v # push on top of the stack
@data.pop if @data.size > @max_entries # remove least recently used entry if `@data` is too big
end
# Removes a value.
def delete(v : V)
@data.select! { |x| v != v }
end
end