Adding launch symlink, helper run config, minor milvus workload updates

This commit is contained in:
Madhavan Sridharan 2024-03-19 14:42:20 -04:00
parent e31c2f02b1
commit 5bea0dac9e
5 changed files with 147 additions and 4 deletions

1
.gitignore vendored
View File

@ -2,6 +2,7 @@ exported_docs.zip
.nosqlbench/** .nosqlbench/**
workspaces/** workspaces/**
workshop/** workshop/**
local*
local/** local/**
metrics/** metrics/**
bin/** bin/**

View File

@ -0,0 +1,50 @@
<component name="ProjectRunConfigurationManager">
<configuration default="false" name="nosqlbench [clean,compile,package...]" type="MavenRunConfiguration" factoryName="Maven" nameIsGenerated="true">
<MavenSettings>
<option name="myGeneralSettings" />
<option name="myRunnerSettings">
<MavenRunnerSettings>
<option name="delegateBuildToMaven" value="false" />
<option name="environmentProperties">
<map />
</option>
<option name="jreName" value="#USE_PROJECT_JDK" />
<option name="mavenProperties">
<map />
</option>
<option name="passParentEnv" value="true" />
<option name="runMavenInBackground" value="true" />
<option name="skipTests" value="false" />
<option name="vmOptions" value="" />
</MavenRunnerSettings>
</option>
<option name="myRunnerParameters">
<MavenRunnerParameters>
<option name="cmdOptions" />
<option name="profiles">
<set />
</option>
<option name="goals">
<list>
<option value="clean" />
<option value="compile" />
<option value="package" />
<option value="-DskipTests" />
</list>
</option>
<option name="multimoduleDir" />
<option name="pomFileName" />
<option name="profilesMap">
<map />
</option>
<option name="projectsCmdOptionValues">
<list />
</option>
<option name="resolveToWorkspace" value="false" />
<option name="workingDirPath" value="$PROJECT_DIR$" />
</MavenRunnerParameters>
</option>
</MavenSettings>
<method v="2" />
</configuration>
</component>

View File

@ -0,0 +1,33 @@
<component name="ProjectRunConfigurationManager">
<configuration default="false" name="nosqlbench [clean,install]" type="MavenRunConfiguration" factoryName="Maven" nameIsGenerated="true">
<MavenSettings>
<option name="myGeneralSettings" />
<option name="myRunnerSettings" />
<option name="myRunnerParameters">
<MavenRunnerParameters>
<option name="cmdOptions" />
<option name="profiles">
<set />
</option>
<option name="goals">
<list>
<option value="clean" />
<option value="install" />
</list>
</option>
<option name="multimoduleDir" />
<option name="pomFileName" />
<option name="profilesMap">
<map />
</option>
<option name="projectsCmdOptionValues">
<list />
</option>
<option name="resolveToWorkspace" value="false" />
<option name="workingDirPath" value="$PROJECT_DIR$" />
</MavenRunnerParameters>
</option>
</MavenSettings>
<method v="2" />
</configuration>
</component>

View File

@ -0,0 +1,34 @@
<component name="ProjectRunConfigurationManager">
<configuration default="false" name="nosqlbench [clean,package,--debug]" type="MavenRunConfiguration" factoryName="Maven" nameIsGenerated="true">
<MavenSettings>
<option name="myGeneralSettings" />
<option name="myRunnerSettings" />
<option name="myRunnerParameters">
<MavenRunnerParameters>
<option name="cmdOptions" />
<option name="profiles">
<set />
</option>
<option name="goals">
<list>
<option value="clean" />
<option value="package" />
<option value="--debug" />
</list>
</option>
<option name="multimoduleDir" />
<option name="pomFileName" />
<option name="profilesMap">
<map />
</option>
<option name="projectsCmdOptionValues">
<list />
</option>
<option name="resolveToWorkspace" value="false" />
<option name="workingDirPath" value="$PROJECT_DIR$" />
</MavenRunnerParameters>
</option>
</MavenSettings>
<method v="2" />
</configuration>
</component>

View File

@ -1,16 +1,41 @@
min_version: 5.21.1 min_version: 5.21.1
description: |
This is a template for live vector search testing.
schema: Install the schema required to run the test
rampup: Measure how long it takes to load a set of embeddings
search: Measure how the system responds to queries while it
is indexing recently ingested data.
search: Run vector search with a set of default (or overridden) parameters
In all of these phases, it is important to instance the metrics with distinct names.
Also, aggregates of recall should include total aggregate as well as a moving average.
scenarios: scenarios:
default: default:
drop_col: run tags==blocks:drop cycles===UNDEF threads===1 databaseName="baselines" uri="localhost:19530" token="root:Milvus" errors=counter,warn drop_col: run tags==blocks:drop cycles===UNDEF threads===1 databaseName="baselines" uri="localhost:19530" token="root:Milvus" errors=counter,warn
schema: run tags==blocks:schema cycles===2 threads===1 databaseName="baselines" uri="localhost:195 token="root:Milvus" errors=counter,warn schema: run tags==blocks:schema cycles===2 threads===1 databaseName="baselines" uri="localhost:195 token="root:Milvus" errors=counter,warn
rampup: run tags==blocks:rampup cycles===TEMPLATE(rampup_cycles,100) threads===TEMPLATE(rampup_threads,10) databaseName="baselines" uri="localhost:195 rampup: run tags==blocks:rampup cycles===TEMPLATE(rampup_cycles,100) threads===TEMPLATE(rampup_threads,10) databaseName="baselines" uri="localhost:195 errors=counter,warn
search: run tags==blocks:search cycles===TEMPLATE(search_cycles,100) threads===TEMPLATE(search_threads,10) databaseName="baselines" uri="localhost:195 search: run tags==blocks:search cycles===TEMPLATE(search_cycles,100) threads===TEMPLATE(search_threads,10) databaseName="baselines" uri="localhost:195 errors=counter,warn
params: params:
driver: milvus driver: milvus
instrument: true instrument: true
bindings:
row_key: ToString()
# filetype=hdf5 for TEMPLATE(filetype,hdf5)
test_floatlist_hdf5: HdfFileToFloatList("testdata/TEMPLATE(datafile).hdf5", "/test"); ToCqlVector();
relevant_indices_hdf5: HdfFileToIntArray("testdata/TEMPLATE(datafile).hdf5", "/neighbors")
distance_floatlist_hdf5: HdfFileToFloatList("testdata/TEMPLATE(datafile).hdf5", "/distance")
train_floatlist_hdf5: HdfFileToFloatList("testdata/TEMPLATE(datafile).hdf5", "/train"); ToCqlVector();
# filetype=fvec for TEMPLATE(filetype,fvec)
test_floatlist_fvec: FVecReader("testdata/TEMPLATE(datafile)_TEMPLATE(trainsize)_query_vectors.fvec"); ToCqlVector();
relevant_indices_fvec: IVecReader("testdata/TEMPLATE(datafile)_TEMPLATE(trainsize)_indices_query.ivec");
distance_floatlist_fvec: FVecReader("testdata/TEMPLATE(datafile)_TEMPLATE(testsize)_distances_count.fvec",TEMPLATE(dimensions),0);
train_floatlist_fvec: FVecReader("testdata/TEMPLATE(datafile)_TEMPLATE(trainsize)_base_vectors.fvec",TEMPLATE(dimensions),0); ToCqlVector();
# synthetic
# synthetic_vectors: HashedFloatVectors(TEMPLATE(dimensions));
blocks: blocks:
drop: drop:
ops: ops:
@ -66,7 +91,7 @@ blocks:
collection_name: "TEMPLATE(collection,vector)" collection_name: "TEMPLATE(collection,vector)"
fields: fields:
key: {row_key} key: {row_key}
value: {row_vector} value: {train_floatlist_TEMPLATE(filetype,hdf5)}
search: search:
ops: ops:
@ -79,6 +104,6 @@ blocks:
- key - key
- value - value
vector_field_name: "value" vector_field_name: "value"
vectors: {row_vector} vectors: {test_floatlist_TEMPLATE(filetype,hdf5)}
top_k: TEMPLATE(top_k,100) top_k: TEMPLATE(top_k,100)
consistency_level: "TEMPLATE(read_cl,EVENTUALLY)" consistency_level: "TEMPLATE(read_cl,EVENTUALLY)"