Adding API Server, Client and documentation

composablesys · Apr 27, 2023 · 0798450 · 0798450
1 parent be64dcd
commit 0798450
Show file tree

Hide file tree

Showing 11 changed files with 5,086 additions and 51 deletions.
diff --git a/.DS_Store b/.DS_Store
diff --git a/API Server b/API Server
diff --git a/Client/visualize.html b/Client/visualize.html
@@ -0,0 +1,64 @@
+<!DOCTYPE html>
+<html>
+  <head>
+    <title>Embedding Vega-Lite</title>
+    <script src="https://cdn.jsdelivr.net/npm/[email protected]"></script>
+    <script src="https://cdn.jsdelivr.net/npm/[email protected]"></script>
+    <script src="https://cdn.jsdelivr.net/npm/[email protected]"></script>
+  </head>
+  <body>
+    <div id="chart"></div>
+
+    <script type="text/javascript">
+        var vlSpec = {
+            "$schema": "https://vega.github.io/schema/vega-lite/v5.json",
+            "concat": [{
+            "title": "Geometric mean of WASM engines",
+            "height": 650,
+            "width": 1000,
+            "data": {name: 'dbData'},
+            "transform": [{"filter": {"param": "suite"}}, {"filter": {"param": "metric"}}],
+            "encoding": {
+                "x": {"field": "Date", "type": "temporal", "title": "Year 2022"},
+                "y": {"aggregate": "mean", "field": "Value", "type": "quantitative", "title": "Metric Value"},
+                "color": {
+                    "field": "Engine"
+                }
+            },
+            "layer": [{"mark": "line"}, 
+                { "params": [{
+                "name": "suite",
+                "select": {"type": "point", "fields": ["Suite"]},
+                "bind": {"name": "Suite","input": "select", "options": ["PolybenchC", "Others"]}
+                },{ 
+                "name": "window", 
+                "select": {"type": "interval", "encodings": ["x"]
+                }},
+                {
+                "name": "metric",
+                "select": {"type": "point", "fields": ["MetricType"]},
+                "bind": {"name": "MetricType","input": "select", "options": ["execution_time", "Others"]}
+                }],
+                "mark": {"type": "circle", "tooltip": true},
+                "encoding": {
+                "opacity": {
+                    "value": 1
+                },
+                "size": {
+                    "value": 30
+                }
+                }
+            }]
+        }]
+     };
+        vegaEmbed('#chart', vlSpec).then(function (res) {
+            async function logJSONData() {
+                var response = await fetch("http://localhost:6363/get_all_mean");
+                var jsonData = await response.json();
+                res.view.insert('dbData', jsonData.data).run();
+            }
+            logJSONData();
+        });
+    </script>
+  </body>
+</html>
diff --git a/DBQueries.txt b/DBQueries.txt
@@ -0,0 +1,40 @@
+1. Granular view #1 - Display all geomean 
+SELECT 
+    exp_date, benchmark_suite, benchmark_item, engine, version, config,
+    machine, metric_type, avg FROM summary
+WHERE 
+    exp_label = 'NIGHTLYRUN' AND benchmark_item = "GEOMEAN"
+
+
+2. Fine view #2 - show for all selected data range
+SELECT 
+    exp_date, benchmark_suite, benchmark_item, engine, version, config,
+    machine, metric_type, avg FROM summary
+WHERE 
+    exp_data BETWEEN value1 AND value2 AND exp_label = 'NIGHTLYRUN' AND benchmark_item != "GEOMEAN"
+
+
+3. Fine view #3 - show violin plot for one DATE and one benchmark suite
+SELECT
+    percentile_95,
+    min,
+    max,
+    metric_type,
+    engine, 
+    version,
+    config,
+    machine,
+WHERE 
+    exp_data = value1 AND exp_label = 'NIGHTLYRUN' AND benchmark_suite = value2 AND benchmark_item != "GEOMEAN"
+
+
+Note: The API Server currently has slightly modified queries, since the database does not have "GEOMEAN" values yet.
+"GEOMEAN" is a tag assigned to "benchmark_item" column in the SUMMARY table, to store the geometric mean for each 
+engine + config across all "benchmark_item" in a "benchmark_suite".
+
+Once the geometric mean values are added to the database, with the corresponding "benchmark_item" column storing string
+"GEOMEAN", the API queries should be replaced with the above mentioned queries.
+
+
+
+
diff --git a/DBSchema.txt b/DBSchema.txt
@@ -0,0 +1,41 @@
+PostgresQL DB Schema for WASM benchmarking
+
+1. Create a database 'wasm_benchmark' on PostgresQL
+
+2. Create a summary table using the following syntax
+CREATE TABLE summary (
+    exp_date DATE NOT NULL DEFAULT CURRENT_DATE,
+    exp_label VARCHAR NOT NULL,
+    benchmark_suite VARCHAR(100) NOT NULL,
+    benchmark_item VARCHAR(100) NOT NULL,
+    engine VARCHAR(100) NOT NULL,
+    version VARCHAR(100) NOT NULL,
+    config VARCHAR(40) NOT NULL,
+    machine VARCHAR(25) NOT NULL,
+    metric_type metric NOT NULL,
+    avg NUMERIC(10,7) NOT NULL,
+    percentile_5 NUMERIC(10,7) NOT NULL,
+    percentile_95 NUMERIC(10,7) NOT NULL,
+    min NUMERIC(10,7) NOT NULL,
+    max NUMERIC(10,7) NOT NULL,
+    time TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
+)
+
+3. Create a raw data table with the following syntax
+CREATE TABLE raw (
+    exp_date DATE NOT NULL DEFAULT CURRENT_DATE,
+    exp_label VARCHAR NOT NULL,
+    benchmark_suite VARCHAR(100) NOT NULL,
+    benchmark_item VARCHAR(100) NOT NULL,
+    engine VARCHAR(100) NOT NULL,
+    version VARCHAR(100) NOT NULL,
+    config VARCHAR(40) NOT NULL,
+    machine VARCHAR(25) NOT NULL,
+    execution_samples NUMERIC(10,7)[] NOT NULL,
+    memory_samples NUMERIC(10,7)[] NOT NULL,
+    preprocess_samples NUMERIC(10,7)[] NOT NULL,
+    aot_samples NUMERIC(10,7)[] NOT NULL,
+    jit_samples NUMERIC(10,7)[] NOT NULL,
+    time TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
+)
+
diff --git a/README.md b/README.md
@@ -1 +1,38 @@
 # wish-you-were-fast
+
+Wish-you-were-fast is a real-time wasm benchmarking + visualization plaftorm. This README attempts to explain the 
+various directories and the parts of the platform they represent.
+
+* API Server
+
+    Hosts a Flask Server acting as an API Gateway microservice between the WASM visualization frontend and
+    the PostgresQL database (storing the benchmarking results). Currently, the API Server provides 3 REST APIs necessary to provide the initial 3 views of the frontend visualization. It receives request parameters from the 
+    frontend service, queries the database, parses it and returns a JSON result containing the requested data. 
+
+* Client
+
+    Provides a barebones html file which hosts the vegalite-based visualizations. Easiest way to get this running locally is to go to the directory on cmdline and start a http server using 
+    "python -m http.server". The visualization would then be available on localhost:8000/visualize.html.
+
+    Note: The visualization currently provides only the first view. While the API Server supports the necessary APIs for all 3 views, we are yet to figure out how an interactive datarange
+    on the first vegalite view (i.e the x-axis select interval) can be converted from vegalite object to a dynamic REST API parameter. Once done, the remaining 2 views can be impleted similar to the multiview-exec.vl.json specification in the vegalite folder.
+
+* historical-data
+
+    Contains raw data for each of the 6 engines on PolyBenchC for a 6 month duration from Sept '22 to Feb '23, sampled at a monthly interval. Also contains a consolidated csv which can be used to load data directly to the PostgresQL database.
+
+* vegalite
+
+    Contains examples and test-scripts for visualizing and learning vegalite
+
+* DBSchema.txt
+
+    Lays down the DB schema to create the PostgresQL database
+
+* DBQueries.txt
+
+    Documents the queries required to create the 3 initial vegalite views
+
+
+
+
diff --git a/historical-data/.DS_Store b/historical-data/.DS_Store