[test] add script and data to plot legacy-vs-blobs

18 files changed, 1369 insertions, 3 deletions

diff --git a/scripts/README.rst b/scripts/README.rst
index 37cf2c0e..a751832e 100644
--- a/scripts/README.rst
+++ b/scripts/README.rst
@@ -1,4 +1,7 @@
-Soledad Scripts
-===============
+Soledad Scripts directory
+=========================
 
-The scripts in this directory are meant to be used for development purposes.
+This directory contains scripts for different purposes, created in different
+times. They are here because at some point they were useful for development
+purposes. Please check each sub directory's readme file to know more about what
+is inside.
diff --git a/scripts/legacy-vs-blobs/README.rst b/scripts/legacy-vs-blobs/README.rst
new file mode 100644
index 00000000..3eedb3a4
--- /dev/null
+++ b/scripts/legacy-vs-blobs/README.rst
@@ -0,0 +1,43 @@
+Ploggin Legacy vs Blobs results
+===============================
+
+This script produces a bar graph comparing different implementations of the
+mail pipeline (legacy, blobs, blobs with session cache, blobs with session
+cache and persistent http connections) for different amount/sizes of incoming
+data (1 x 10M, 10 x 1M, 100 x 100K, 1000 x 10K).
+
+Data acquisition
+----------------
+
+Data was gathered through different runs of the code in
+`soledad/tests/benchmarks/test_legacy_vs_blobs.py`. As these tests are skipped
+by default, you have to (1) remove the @pytest.mark.skip decorator and (2)
+remove the second call to pytest in `tox.ini` that would cause a second run of
+the tests for measuring memory. This script only plots time, for now.
+
+Once you done the above, chdir into `soledad/tests` and do something like this,
+for each test group:
+
+  tox -e benchmarks -- -m benchmark_test_legacy_vs_blobs_10_1000k --benchmark-autosave
+
+Each run of tox as above produces a JSON file in `soledad/tests/.benchmarks`
+with stats. Each file has to be copied in a subdirectory of the `data/`
+directory of the folder where this script lies. Code has to be adapted for each
+extra set of results added to the `data/` directory.
+
+Currently, we are comparing 4 implementation variations:
+
+  - data/no-cache: contains legacy as baseline and blobs as the current master
+    implementation of blobs.
+  - data/cache: adds twisted session cache to blobs connections.
+  - data/persistent: adds session cache and persistent HTTP connections.
+
+Plotting
+--------
+
+Once you have all your data in subdirectories of the `data/` directory and the
+script is correctly tuned for your data, running `make` should be enough to
+create a virtualenvironment with dependencies and plotting data.
+
+By default, the script will generate a file called `./legacy-vs-blobs.png` in
+the current directory with the plot of the data.
diff --git a/scripts/legacy-vs-blobs/data/cache/1000_10k.json b/scripts/legacy-vs-blobs/data/cache/1000_10k.json
new file mode 100644
index 00000000..5be17758
--- /dev/null
+++ b/scripts/legacy-vs-blobs/data/cache/1000_10k.json
@@ -0,0 +1,111 @@
+{
+    "datetime": "2017-09-02T04:22:20.074434", 
+    "machine_info": {
+        "python_implementation": "CPython", 
+        "python_compiler": "GCC 6.3.0 20170118", 
+        "machine": "x86_64", 
+        "cpu": {
+            "hardware": "unknown", 
+            "brand": "Intel(R) Core(TM) i7-4600U CPU @ 2.10GHz", 
+            "vendor_id": "GenuineIntel"
+        }, 
+        "host": "pajeh", 
+        "system": "Linux", 
+        "processor": "", 
+        "python_implementation_version": "2.7.13", 
+        "release": "4.9.0-3-amd64", 
+        "python_version": "2.7.13", 
+        "python_build": [
+            "default", 
+            "Jan 19 2017 14:48:08"
+        ], 
+        "node": "pajeh"
+    }, 
+    "commit_info": {
+        "author_time": "2017-09-02T00:27:58-03:00", 
+        "time": "2017-09-02T00:39:13-03:00", 
+        "dirty": true, 
+        "project": "soledad", 
+        "id": "2dea7ebc513eaa506704fd7ae68bae94b220eea8", 
+        "branch": "8856"
+    }, 
+    "benchmarks": [
+        {
+            "name": "test_legacy_1000_10k", 
+            "stats": {
+                "median": 48.28695607185364, 
+                "rounds": 4, 
+                "ld15iqr": 36.51543307304382, 
+                "iqr_outliers": 0, 
+                "stddev": 10.431966763848191, 
+                "q1": 40.382596015930176, 
+                "min": 36.51543307304382, 
+                "q3": 56.539175629615784, 
+                "total": 193.84354329109192, 
+                "iterations": 1, 
+                "max": 60.75419807434082, 
+                "stddev_outliers": 2, 
+                "mean": 48.46088582277298, 
+                "ops": 0.020635198532216573, 
+                "hd15iqr": 60.75419807434082, 
+                "outliers": "2;0", 
+                "iqr": 16.156579613685608
+            }, 
+            "extra_info": {
+                "cpu_percent": 26.6, 
+                "doc": ""
+            }, 
+            "params": null, 
+            "param": null, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_legacy_1000_10k", 
+            "group": "test_legacy_vs_blobs_1000_10k", 
+            "options": {
+                "timer": "time", 
+                "min_rounds": 5, 
+                "warmup": false, 
+                "disable_gc": false, 
+                "min_time": 5e-06, 
+                "max_time": 1.0
+            }
+        }, 
+        {
+            "name": "test_blobs_1000_10k", 
+            "stats": {
+                "median": 87.17049705982208, 
+                "rounds": 4, 
+                "ld15iqr": 65.08280301094055, 
+                "iqr_outliers": 0, 
+                "stddev": 19.00300388729206, 
+                "q1": 72.67573952674866, 
+                "min": 65.08280301094055, 
+                "q3": 101.80426609516144, 
+                "total": 348.9600112438202, 
+                "iterations": 1, 
+                "max": 109.53621411323547, 
+                "stddev_outliers": 2, 
+                "mean": 87.24000281095505, 
+                "ops": 0.01146263145092914, 
+                "hd15iqr": 109.53621411323547, 
+                "outliers": "2;0", 
+                "iqr": 29.12852656841278
+            }, 
+            "extra_info": {
+                "cpu_percent": 19.4, 
+                "doc": ""
+            }, 
+            "params": null, 
+            "param": null, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_blobs_1000_10k", 
+            "group": "test_legacy_vs_blobs_1000_10k", 
+            "options": {
+                "timer": "time", 
+                "min_rounds": 5, 
+                "warmup": false, 
+                "disable_gc": false, 
+                "min_time": 5e-06, 
+                "max_time": 1.0
+            }
+        }
+    ], 
+    "version": "3.1.1"
+}
\ No newline at end of file
diff --git a/scripts/legacy-vs-blobs/data/cache/100_100k.json b/scripts/legacy-vs-blobs/data/cache/100_100k.json
new file mode 100644
index 00000000..17fc598b
--- /dev/null
+++ b/scripts/legacy-vs-blobs/data/cache/100_100k.json
@@ -0,0 +1,111 @@
+{
+    "benchmarks": [
+        {
+            "params": null, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_legacy_100_100k", 
+            "extra_info": {
+                "doc": "", 
+                "cpu_percent": 25.1
+            }, 
+            "group": "test_legacy_vs_blobs_100_100k", 
+            "stats": {
+                "q1": 11.39308750629425, 
+                "rounds": 4, 
+                "q3": 12.677581906318665, 
+                "hd15iqr": 12.946987867355347, 
+                "median": 12.090978026390076, 
+                "stddev_outliers": 2, 
+                "mean": 12.035334706306458, 
+                "iqr_outliers": 0, 
+                "iterations": 1, 
+                "total": 48.14133882522583, 
+                "min": 11.012394905090332, 
+                "outliers": "2;0", 
+                "ops": 0.08308867384269794, 
+                "ld15iqr": 11.012394905090332, 
+                "stddev": 0.833654348956266, 
+                "iqr": 1.284494400024414, 
+                "max": 12.946987867355347
+            }, 
+            "name": "test_legacy_100_100k", 
+            "param": null, 
+            "options": {
+                "min_rounds": 5, 
+                "timer": "time", 
+                "max_time": 1.0, 
+                "min_time": 5e-06, 
+                "warmup": false, 
+                "disable_gc": false
+            }
+        }, 
+        {
+            "params": null, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_blobs_100_100k", 
+            "extra_info": {
+                "doc": "", 
+                "cpu_percent": 24.9
+            }, 
+            "group": "test_legacy_vs_blobs_100_100k", 
+            "stats": {
+                "q1": 5.015861988067627, 
+                "rounds": 4, 
+                "q3": 6.166973114013672, 
+                "hd15iqr": 6.362173080444336, 
+                "median": 5.64124858379364, 
+                "stddev_outliers": 2, 
+                "mean": 5.591417551040649, 
+                "iqr_outliers": 0, 
+                "iterations": 1, 
+                "total": 22.365670204162598, 
+                "min": 4.7209999561309814, 
+                "outliers": "2;0", 
+                "ops": 0.1788455236747405, 
+                "ld15iqr": 4.7209999561309814, 
+                "stddev": 0.7246033659461454, 
+                "iqr": 1.151111125946045, 
+                "max": 6.362173080444336
+            }, 
+            "name": "test_blobs_100_100k", 
+            "param": null, 
+            "options": {
+                "min_rounds": 5, 
+                "timer": "time", 
+                "max_time": 1.0, 
+                "min_time": 5e-06, 
+                "warmup": false, 
+                "disable_gc": false
+            }
+        }
+    ], 
+    "datetime": "2017-09-02T03:53:29.055256", 
+    "version": "3.1.1", 
+    "machine_info": {
+        "node": "pajeh", 
+        "python_build": [
+            "default", 
+            "Jan 19 2017 14:48:08"
+        ], 
+        "python_version": "2.7.13", 
+        "machine": "x86_64", 
+        "python_implementation": "CPython", 
+        "cpu": {
+            "brand": "Intel(R) Core(TM) i7-4600U CPU @ 2.10GHz", 
+            "hardware": "unknown", 
+            "vendor_id": "GenuineIntel"
+        }, 
+        "python_implementation_version": "2.7.13", 
+        "system": "Linux", 
+        "host": "pajeh", 
+        "python_compiler": "GCC 6.3.0 20170118", 
+        "release": "4.9.0-3-amd64", 
+        "processor": ""
+    }, 
+    "commit_info": {
+        "project": "soledad", 
+        "time": "2017-09-02T00:39:13-03:00", 
+        "author_time": "2017-09-02T00:27:58-03:00", 
+        "branch": "8856", 
+        "id": "2dea7ebc513eaa506704fd7ae68bae94b220eea8", 
+        "dirty": true
+    }
+}
\ No newline at end of file
diff --git a/scripts/legacy-vs-blobs/data/cache/10_1000k.json b/scripts/legacy-vs-blobs/data/cache/10_1000k.json
new file mode 100644
index 00000000..1a733f2d
--- /dev/null
+++ b/scripts/legacy-vs-blobs/data/cache/10_1000k.json
@@ -0,0 +1,111 @@
+{
+    "benchmarks": [
+        {
+            "params": null, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_legacy_10_1000k", 
+            "extra_info": {
+                "doc": "", 
+                "cpu_percent": 27.7
+            }, 
+            "group": "test_legacy_vs_blobs_10_1000k", 
+            "stats": {
+                "q1": 8.741152048110962, 
+                "rounds": 4, 
+                "q3": 8.863584876060486, 
+                "median": 8.83247196674347, 
+                "stddev_outliers": 1, 
+                "iqr": 0.12243282794952393, 
+                "mean": 8.802368462085724, 
+                "iqr_outliers": 0, 
+                "iterations": 1, 
+                "total": 35.209473848342896, 
+                "min": 8.67733097076416, 
+                "outliers": "1;0", 
+                "ops": 0.11360578738634734, 
+                "ld15iqr": 8.67733097076416, 
+                "stddev": 0.08786766595777334, 
+                "hd15iqr": 8.867198944091797, 
+                "max": 8.867198944091797
+            }, 
+            "name": "test_legacy_10_1000k", 
+            "param": null, 
+            "options": {
+                "min_rounds": 5, 
+                "timer": "time", 
+                "max_time": 1.0, 
+                "min_time": 5e-06, 
+                "warmup": false, 
+                "disable_gc": false
+            }
+        }, 
+        {
+            "params": null, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_blobs_10_1000k", 
+            "extra_info": {
+                "doc": "", 
+                "cpu_percent": 37.0
+            }, 
+            "group": "test_legacy_vs_blobs_10_1000k", 
+            "stats": {
+                "q1": 1.8672034740447998, 
+                "rounds": 4, 
+                "q3": 1.9366559982299805, 
+                "median": 1.8831886053085327, 
+                "stddev_outliers": 1, 
+                "iqr": 0.06945252418518066, 
+                "mean": 1.9019297361373901, 
+                "iqr_outliers": 0, 
+                "iterations": 1, 
+                "total": 7.6077189445495605, 
+                "min": 1.8517439365386963, 
+                "outliers": "1;0", 
+                "ops": 0.5257817788952025, 
+                "ld15iqr": 1.8517439365386963, 
+                "stddev": 0.06029736288568101, 
+                "hd15iqr": 1.9895977973937988, 
+                "max": 1.9895977973937988
+            }, 
+            "name": "test_blobs_10_1000k", 
+            "param": null, 
+            "options": {
+                "min_rounds": 5, 
+                "timer": "time", 
+                "max_time": 1.0, 
+                "min_time": 5e-06, 
+                "warmup": false, 
+                "disable_gc": false
+            }
+        }
+    ], 
+    "datetime": "2017-09-02T03:48:24.650044", 
+    "version": "3.1.1", 
+    "machine_info": {
+        "python_implementation_version": "2.7.13", 
+        "system": "Linux", 
+        "host": "pajeh", 
+        "python_compiler": "GCC 6.3.0 20170118", 
+        "release": "4.9.0-3-amd64", 
+        "python_implementation": "CPython", 
+        "processor": "", 
+        "node": "pajeh", 
+        "python_build": [
+            "default", 
+            "Jan 19 2017 14:48:08"
+        ], 
+        "python_version": "2.7.13", 
+        "machine": "x86_64", 
+        "cpu": {
+            "brand": "Intel(R) Core(TM) i7-4600U CPU @ 2.10GHz", 
+            "hardware": "unknown", 
+            "vendor_id": "GenuineIntel"
+        }
+    }, 
+    "commit_info": {
+        "author_time": "2017-09-02T00:27:58-03:00", 
+        "time": "2017-09-02T00:39:13-03:00", 
+        "project": "soledad", 
+        "branch": "8856", 
+        "id": "2dea7ebc513eaa506704fd7ae68bae94b220eea8", 
+        "dirty": true
+    }
+}
\ No newline at end of file
diff --git a/scripts/legacy-vs-blobs/data/cache/1_10000k.json b/scripts/legacy-vs-blobs/data/cache/1_10000k.json
new file mode 100644
index 00000000..2d970c6e
--- /dev/null
+++ b/scripts/legacy-vs-blobs/data/cache/1_10000k.json
@@ -0,0 +1,111 @@
+{
+    "datetime": "2017-09-02T03:46:17.875002", 
+    "machine_info": {
+        "processor": "", 
+        "python_implementation_version": "2.7.13", 
+        "python_version": "2.7.13", 
+        "machine": "x86_64", 
+        "python_build": [
+            "default", 
+            "Jan 19 2017 14:48:08"
+        ], 
+        "python_implementation": "CPython", 
+        "python_compiler": "GCC 6.3.0 20170118", 
+        "cpu": {
+            "vendor_id": "GenuineIntel", 
+            "brand": "Intel(R) Core(TM) i7-4600U CPU @ 2.10GHz", 
+            "hardware": "unknown"
+        }, 
+        "release": "4.9.0-3-amd64", 
+        "system": "Linux", 
+        "host": "pajeh", 
+        "node": "pajeh"
+    }, 
+    "version": "3.1.1", 
+    "commit_info": {
+        "time": "2017-09-02T00:39:13-03:00", 
+        "id": "2dea7ebc513eaa506704fd7ae68bae94b220eea8", 
+        "project": "soledad", 
+        "author_time": "2017-09-02T00:27:58-03:00", 
+        "branch": "8856", 
+        "dirty": true
+    }, 
+    "benchmarks": [
+        {
+            "extra_info": {
+                "cpu_percent": 25.8, 
+                "doc": ""
+            }, 
+            "name": "test_legacy_1_10000k", 
+            "params": null, 
+            "options": {
+                "max_time": 1.0, 
+                "min_rounds": 5, 
+                "min_time": 5e-06, 
+                "warmup": false, 
+                "timer": "time", 
+                "disable_gc": false
+            }, 
+            "group": "test_legacy_vs_blobs_1_10000k", 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_legacy_1_10000k", 
+            "param": null, 
+            "stats": {
+                "mean": 9.607597470283508, 
+                "ld15iqr": 9.551655054092407, 
+                "iqr_outliers": 0, 
+                "outliers": "1;0", 
+                "q3": 9.663439989089966, 
+                "q1": 9.55175495147705, 
+                "iqr": 0.11168503761291504, 
+                "total": 38.43038988113403, 
+                "rounds": 4, 
+                "median": 9.563774466514587, 
+                "stddev_outliers": 1, 
+                "min": 9.551655054092407, 
+                "hd15iqr": 9.751185894012451, 
+                "ops": 0.10408429402803564, 
+                "stddev": 0.09638854573531173, 
+                "iterations": 1, 
+                "max": 9.751185894012451
+            }
+        }, 
+        {
+            "extra_info": {
+                "cpu_percent": 40.2, 
+                "doc": ""
+            }, 
+            "name": "test_blobs_1_10000k", 
+            "params": null, 
+            "options": {
+                "max_time": 1.0, 
+                "min_rounds": 5, 
+                "min_time": 5e-06, 
+                "warmup": false, 
+                "timer": "time", 
+                "disable_gc": false
+            }, 
+            "group": "test_legacy_vs_blobs_1_10000k", 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_blobs_1_10000k", 
+            "param": null, 
+            "stats": {
+                "mean": 2.179791212081909, 
+                "ld15iqr": 2.1572468280792236, 
+                "iqr_outliers": 0, 
+                "outliers": "2;0", 
+                "q3": 2.193374514579773, 
+                "q1": 2.1662079095840454, 
+                "iqr": 0.02716660499572754, 
+                "total": 8.719164848327637, 
+                "rounds": 4, 
+                "median": 2.1782270669937134, 
+                "stddev_outliers": 2, 
+                "min": 2.1572468280792236, 
+                "hd15iqr": 2.2054638862609863, 
+                "ops": 0.45875953369171735, 
+                "stddev": 0.01992429175453132, 
+                "iterations": 1, 
+                "max": 2.2054638862609863
+            }
+        }
+    ]
+}
\ No newline at end of file
diff --git a/scripts/legacy-vs-blobs/data/no-cache/1000_10k.json b/scripts/legacy-vs-blobs/data/no-cache/1000_10k.json
new file mode 100644
index 00000000..a585ad3a
--- /dev/null
+++ b/scripts/legacy-vs-blobs/data/no-cache/1000_10k.json
@@ -0,0 +1,111 @@
+{
+    "commit_info": {
+        "dirty": true, 
+        "branch": "8856", 
+        "author_time": "2017-09-01T09:50:11-03:00", 
+        "id": "365d9663cbbd524366b7dfe73ac5cccefeb357c8", 
+        "time": "2017-09-01T09:50:20-03:00", 
+        "project": "soledad"
+    }, 
+    "machine_info": {
+        "node": "pajeh", 
+        "processor": "", 
+        "machine": "x86_64", 
+        "system": "Linux", 
+        "python_build": [
+            "default", 
+            "Jan 19 2017 14:48:08"
+        ], 
+        "python_implementation": "CPython", 
+        "python_implementation_version": "2.7.13", 
+        "host": "pajeh", 
+        "cpu": {
+            "hardware": "unknown", 
+            "brand": "Intel(R) Core(TM) i7-4600U CPU @ 2.10GHz", 
+            "vendor_id": "GenuineIntel"
+        }, 
+        "release": "4.9.0-3-amd64", 
+        "python_compiler": "GCC 6.3.0 20170118", 
+        "python_version": "2.7.13"
+    }, 
+    "datetime": "2017-09-01T16:28:31.243069", 
+    "version": "3.1.1", 
+    "benchmarks": [
+        {
+            "stats": {
+                "ld15iqr": 38.49253582954407, 
+                "outliers": "2;0", 
+                "median": 52.51530361175537, 
+                "iterations": 1, 
+                "q3": 65.07869005203247, 
+                "iqr_outliers": 0, 
+                "rounds": 4, 
+                "ops": 0.018550090252706083, 
+                "total": 215.63237404823303, 
+                "mean": 53.90809351205826, 
+                "max": 72.10923099517822, 
+                "iqr": 22.341193079948425, 
+                "stddev_outliers": 2, 
+                "q1": 42.737496972084045, 
+                "min": 38.49253582954407, 
+                "hd15iqr": 72.10923099517822, 
+                "stddev": 14.537598498896823
+            }, 
+            "name": "test_legacy_1000_10k", 
+            "param": null, 
+            "group": "test_legacy_vs_blobs_1000_10k", 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_legacy_1000_10k", 
+            "params": null, 
+            "extra_info": {
+                "doc": "", 
+                "cpu_percent": 26.9
+            }, 
+            "options": {
+                "warmup": false, 
+                "max_time": 1.0, 
+                "min_time": 5e-06, 
+                "min_rounds": 5, 
+                "timer": "time", 
+                "disable_gc": false
+            }
+        }, 
+        {
+            "stats": {
+                "ld15iqr": 176.17668509483337, 
+                "outliers": "2;0", 
+                "median": 192.90700149536133, 
+                "iterations": 1, 
+                "q3": 206.93408203125, 
+                "iqr_outliers": 0, 
+                "rounds": 4, 
+                "ops": 0.005138941398078599, 
+                "total": 778.3704249858856, 
+                "mean": 194.5926062464714, 
+                "max": 216.3797369003296, 
+                "iqr": 24.68295156955719, 
+                "stddev_outliers": 2, 
+                "q1": 182.2511304616928, 
+                "min": 176.17668509483337, 
+                "hd15iqr": 216.3797369003296, 
+                "stddev": 16.94586147142875
+            }, 
+            "name": "test_blobs_1000_10k", 
+            "param": null, 
+            "group": "test_legacy_vs_blobs_1000_10k", 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_blobs_1000_10k", 
+            "params": null, 
+            "extra_info": {
+                "doc": "", 
+                "cpu_percent": 10.9
+            }, 
+            "options": {
+                "warmup": false, 
+                "max_time": 1.0, 
+                "min_time": 5e-06, 
+                "min_rounds": 5, 
+                "timer": "time", 
+                "disable_gc": false
+            }
+        }
+    ]
+}
\ No newline at end of file
diff --git a/scripts/legacy-vs-blobs/data/no-cache/100_100k.json b/scripts/legacy-vs-blobs/data/no-cache/100_100k.json
new file mode 100644
index 00000000..2cbbe638
--- /dev/null
+++ b/scripts/legacy-vs-blobs/data/no-cache/100_100k.json
@@ -0,0 +1,111 @@
+{
+    "commit_info": {
+        "time": "2017-09-01T09:50:20-03:00", 
+        "branch": "8856", 
+        "id": "365d9663cbbd524366b7dfe73ac5cccefeb357c8", 
+        "project": "soledad", 
+        "dirty": true, 
+        "author_time": "2017-09-01T09:50:11-03:00"
+    }, 
+    "version": "3.1.1", 
+    "benchmarks": [
+        {
+            "stats": {
+                "stddev_outliers": 1, 
+                "ops": 0.08172728164268425, 
+                "stddev": 1.0083003646390902, 
+                "hd15iqr": 13.605333805084229, 
+                "total": 48.943264961242676, 
+                "iqr": 1.4744374752044678, 
+                "iqr_outliers": 0, 
+                "q1": 11.498597502708435, 
+                "q3": 12.973034977912903, 
+                "ld15iqr": 11.303098917007446, 
+                "min": 11.303098917007446, 
+                "iterations": 1, 
+                "max": 13.605333805084229, 
+                "outliers": "1;0", 
+                "rounds": 4, 
+                "median": 12.0174161195755, 
+                "mean": 12.235816240310669
+            }, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_legacy_100_100k", 
+            "options": {
+                "warmup": false, 
+                "min_time": 5e-06, 
+                "max_time": 1.0, 
+                "disable_gc": false, 
+                "min_rounds": 5, 
+                "timer": "time"
+            }, 
+            "param": null, 
+            "name": "test_legacy_100_100k", 
+            "group": "test_legacy_vs_blobs_100_100k", 
+            "extra_info": {
+                "doc": "", 
+                "cpu_percent": 25.2
+            }, 
+            "params": null
+        }, 
+        {
+            "stats": {
+                "stddev_outliers": 2, 
+                "ops": 0.052999047893223836, 
+                "stddev": 0.6994077398028641, 
+                "hd15iqr": 19.686198949813843, 
+                "total": 75.47305393218994, 
+                "iqr": 1.0998969078063965, 
+                "iqr_outliers": 0, 
+                "q1": 18.318315029144287, 
+                "q3": 19.418211936950684, 
+                "ld15iqr": 18.07945704460144, 
+                "min": 18.07945704460144, 
+                "iterations": 1, 
+                "max": 19.686198949813843, 
+                "outliers": "2;0", 
+                "rounds": 4, 
+                "median": 18.85369896888733, 
+                "mean": 18.868263483047485
+            }, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_blobs_100_100k", 
+            "options": {
+                "warmup": false, 
+                "min_time": 5e-06, 
+                "max_time": 1.0, 
+                "disable_gc": false, 
+                "min_rounds": 5, 
+                "timer": "time"
+            }, 
+            "param": null, 
+            "name": "test_blobs_100_100k", 
+            "group": "test_legacy_vs_blobs_100_100k", 
+            "extra_info": {
+                "doc": "", 
+                "cpu_percent": 14.7
+            }, 
+            "params": null
+        }
+    ], 
+    "datetime": "2017-09-01T15:44:20.745800", 
+    "machine_info": {
+        "host": "pajeh", 
+        "python_compiler": "GCC 6.3.0 20170118", 
+        "python_build": [
+            "default", 
+            "Jan 19 2017 14:48:08"
+        ], 
+        "python_implementation": "CPython", 
+        "cpu": {
+            "hardware": "unknown", 
+            "vendor_id": "GenuineIntel", 
+            "brand": "Intel(R) Core(TM) i7-4600U CPU @ 2.10GHz"
+        }, 
+        "node": "pajeh", 
+        "release": "4.9.0-3-amd64", 
+        "processor": "", 
+        "python_implementation_version": "2.7.13", 
+        "python_version": "2.7.13", 
+        "system": "Linux", 
+        "machine": "x86_64"
+    }
+}
\ No newline at end of file
diff --git a/scripts/legacy-vs-blobs/data/no-cache/10_1000k.json b/scripts/legacy-vs-blobs/data/no-cache/10_1000k.json
new file mode 100644
index 00000000..778e5c63
--- /dev/null
+++ b/scripts/legacy-vs-blobs/data/no-cache/10_1000k.json
@@ -0,0 +1,111 @@
+{
+    "version": "3.1.1", 
+    "machine_info": {
+        "python_implementation": "CPython", 
+        "host": "pajeh", 
+        "python_build": [
+            "default", 
+            "Jan 19 2017 14:48:08"
+        ], 
+        "cpu": {
+            "hardware": "unknown", 
+            "vendor_id": "GenuineIntel", 
+            "brand": "Intel(R) Core(TM) i7-4600U CPU @ 2.10GHz"
+        }, 
+        "release": "4.9.0-3-amd64", 
+        "python_version": "2.7.13", 
+        "machine": "x86_64", 
+        "python_implementation_version": "2.7.13", 
+        "processor": "", 
+        "node": "pajeh", 
+        "system": "Linux", 
+        "python_compiler": "GCC 6.3.0 20170118"
+    }, 
+    "commit_info": {
+        "id": "365d9663cbbd524366b7dfe73ac5cccefeb357c8", 
+        "time": "2017-09-01T09:50:20-03:00", 
+        "project": "soledad", 
+        "dirty": true, 
+        "branch": "8856", 
+        "author_time": "2017-09-01T09:50:11-03:00"
+    }, 
+    "datetime": "2017-09-01T15:37:48.278845", 
+    "benchmarks": [
+        {
+            "param": null, 
+            "extra_info": {
+                "cpu_percent": 27.9, 
+                "doc": ""
+            }, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_legacy_10_1000k", 
+            "group": "test_legacy_vs_blobs_10_1000k", 
+            "params": null, 
+            "name": "test_legacy_10_1000k", 
+            "stats": {
+                "mean": 8.937849223613739, 
+                "min": 8.711518049240112, 
+                "outliers": "2;0", 
+                "total": 35.751396894454956, 
+                "iqr_outliers": 0, 
+                "iterations": 1, 
+                "ops": 0.11188374014612, 
+                "rounds": 4, 
+                "stddev": 0.20914882546041347, 
+                "hd15iqr": 9.213748931884766, 
+                "median": 8.913064956665039, 
+                "stddev_outliers": 2, 
+                "max": 9.213748931884766, 
+                "q1": 8.794075012207031, 
+                "ld15iqr": 8.711518049240112, 
+                "q3": 9.081623435020447, 
+                "iqr": 0.2875484228134155
+            }, 
+            "options": {
+                "disable_gc": false, 
+                "min_time": 5e-06, 
+                "min_rounds": 5, 
+                "timer": "time", 
+                "max_time": 1.0, 
+                "warmup": false
+            }
+        }, 
+        {
+            "param": null, 
+            "extra_info": {
+                "cpu_percent": 31.7, 
+                "doc": ""
+            }, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_blobs_10_1000k", 
+            "group": "test_legacy_vs_blobs_10_1000k", 
+            "params": null, 
+            "name": "test_blobs_10_1000k", 
+            "stats": {
+                "mean": 3.1771076917648315, 
+                "min": 3.077280044555664, 
+                "outliers": "1;0", 
+                "total": 12.708430767059326, 
+                "iqr_outliers": 0, 
+                "iterations": 1, 
+                "ops": 0.3147516851858793, 
+                "rounds": 4, 
+                "stddev": 0.0741566234351459, 
+                "hd15iqr": 3.2503108978271484, 
+                "median": 3.190419912338257, 
+                "stddev_outliers": 1, 
+                "max": 3.2503108978271484, 
+                "q1": 3.1237324476242065, 
+                "ld15iqr": 3.077280044555664, 
+                "q3": 3.2304829359054565, 
+                "iqr": 0.10675048828125
+            }, 
+            "options": {
+                "disable_gc": false, 
+                "min_time": 5e-06, 
+                "min_rounds": 5, 
+                "timer": "time", 
+                "max_time": 1.0, 
+                "warmup": false
+            }
+        }
+    ]
+}
\ No newline at end of file
diff --git a/scripts/legacy-vs-blobs/data/no-cache/1_10000k.json b/scripts/legacy-vs-blobs/data/no-cache/1_10000k.json
new file mode 100644
index 00000000..0697cd6b
--- /dev/null
+++ b/scripts/legacy-vs-blobs/data/no-cache/1_10000k.json
@@ -0,0 +1,111 @@
+{
+    "benchmarks": [
+        {
+            "options": {
+                "disable_gc": false, 
+                "timer": "time", 
+                "min_rounds": 5, 
+                "min_time": 5e-06, 
+                "max_time": 1.0, 
+                "warmup": false
+            }, 
+            "params": null, 
+            "extra_info": {
+                "cpu_percent": 25.8, 
+                "doc": ""
+            }, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_legacy_1_10000k", 
+            "group": "test_legacy_vs_blobs_1_10000k", 
+            "name": "test_legacy_1_10000k", 
+            "param": null, 
+            "stats": {
+                "hd15iqr": 9.522547006607056, 
+                "stddev": 0.04062225700829493, 
+                "ops": 0.10556240321945518, 
+                "iterations": 1, 
+                "min": 9.423872947692871, 
+                "total": 37.89227867126465, 
+                "stddev_outliers": 2, 
+                "max": 9.522547006607056, 
+                "q1": 9.445196866989136, 
+                "mean": 9.473069667816162, 
+                "iqr_outliers": 0, 
+                "rounds": 4, 
+                "iqr": 0.055745601654052734, 
+                "median": 9.47292935848236, 
+                "q3": 9.500942468643188, 
+                "ld15iqr": 9.423872947692871, 
+                "outliers": "2;0"
+            }
+        }, 
+        {
+            "options": {
+                "disable_gc": false, 
+                "timer": "time", 
+                "min_rounds": 5, 
+                "min_time": 5e-06, 
+                "max_time": 1.0, 
+                "warmup": false
+            }, 
+            "params": null, 
+            "extra_info": {
+                "cpu_percent": 33.3, 
+                "doc": ""
+            }, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_blobs_1_10000k", 
+            "group": "test_legacy_vs_blobs_1_10000k", 
+            "name": "test_blobs_1_10000k", 
+            "param": null, 
+            "stats": {
+                "hd15iqr": 3.758697032928467, 
+                "stddev": 0.619090853876154, 
+                "ops": 0.3292143381447483, 
+                "iterations": 1, 
+                "min": 2.344238042831421, 
+                "total": 12.150139093399048, 
+                "stddev_outliers": 2, 
+                "max": 3.758697032928467, 
+                "q1": 2.5475854873657227, 
+                "mean": 3.037534773349762, 
+                "iqr_outliers": 0, 
+                "rounds": 4, 
+                "iqr": 0.9798985719680786, 
+                "median": 3.02360200881958, 
+                "q3": 3.5274840593338013, 
+                "ld15iqr": 2.344238042831421, 
+                "outliers": "2;0"
+            }
+        }
+    ], 
+    "version": "3.1.1", 
+    "commit_info": {
+        "author_time": "2017-09-01T14:16:21-03:00", 
+        "project": "soledad", 
+        "id": "7a5698ad8c1457460d95cbd6dd237fb96e218d43", 
+        "time": "2017-09-01T14:16:26-03:00", 
+        "dirty": true, 
+        "branch": "8856"
+    }, 
+    "machine_info": {
+        "host": "pajeh", 
+        "python_version": "2.7.13", 
+        "python_compiler": "GCC 6.3.0 20170118", 
+        "system": "Linux", 
+        "python_implementation": "CPython", 
+        "python_implementation_version": "2.7.13", 
+        "python_build": [
+            "default", 
+            "Jan 19 2017 14:48:08"
+        ], 
+        "machine": "x86_64", 
+        "release": "4.9.0-3-amd64", 
+        "processor": "", 
+        "node": "pajeh", 
+        "cpu": {
+            "hardware": "unknown", 
+            "brand": "Intel(R) Core(TM) i7-4600U CPU @ 2.10GHz", 
+            "vendor_id": "GenuineIntel"
+        }
+    }, 
+    "datetime": "2017-09-01T22:20:37.591520"
+}
\ No newline at end of file
diff --git a/scripts/legacy-vs-blobs/data/persistent/1000_10k.json b/scripts/legacy-vs-blobs/data/persistent/1000_10k.json
new file mode 100644
index 00000000..909e6f21
--- /dev/null
+++ b/scripts/legacy-vs-blobs/data/persistent/1000_10k.json
@@ -0,0 +1,73 @@
+{
+    "version": "3.1.1", 
+    "machine_info": {
+        "python_version": "2.7.13", 
+        "processor": "", 
+        "node": "pajeh", 
+        "python_compiler": "GCC 6.3.0 20170118", 
+        "python_implementation_version": "2.7.13", 
+        "python_implementation": "CPython", 
+        "machine": "x86_64", 
+        "system": "Linux", 
+        "host": "pajeh", 
+        "cpu": {
+            "hardware": "unknown", 
+            "vendor_id": "GenuineIntel", 
+            "brand": "Intel(R) Core(TM) i7-4600U CPU @ 2.10GHz"
+        }, 
+        "python_build": [
+            "default", 
+            "Jan 19 2017 14:48:08"
+        ], 
+        "release": "4.9.0-3-amd64"
+    }, 
+    "commit_info": {
+        "id": "1e132568857e6c2d4fbbbf886d3b7ab2009c3c9e", 
+        "time": "2017-09-02T02:38:57-03:00", 
+        "dirty": true, 
+        "branch": "8856", 
+        "author_time": "2017-09-02T02:38:57-03:00", 
+        "project": "soledad"
+    }, 
+    "benchmarks": [
+        {
+            "param": null, 
+            "stats": {
+                "ld15iqr": 50.158271074295044, 
+                "total": 290.0146920681, 
+                "iterations": 1, 
+                "stddev": 19.375312299153435, 
+                "stddev_outliers": 2, 
+                "mean": 72.503673017025, 
+                "q1": 56.99843108654022, 
+                "rounds": 4, 
+                "q3": 88.00891494750977, 
+                "iqr": 31.010483860969543, 
+                "max": 93.9663679599762, 
+                "iqr_outliers": 0, 
+                "hd15iqr": 93.9663679599762, 
+                "median": 72.94502651691437, 
+                "ops": 0.013792404693279255, 
+                "min": 50.158271074295044, 
+                "outliers": "2;0"
+            }, 
+            "params": null, 
+            "name": "test_blobs_1000_10k", 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_blobs_1000_10k", 
+            "group": "test_legacy_vs_blobs_1000_10k", 
+            "options": {
+                "timer": "time", 
+                "disable_gc": false, 
+                "min_rounds": 5, 
+                "min_time": 5e-06, 
+                "warmup": false, 
+                "max_time": 1.0
+            }, 
+            "extra_info": {
+                "cpu_percent": 17.4, 
+                "doc": ""
+            }
+        }
+    ], 
+    "datetime": "2017-09-02T05:51:50.400601"
+}
\ No newline at end of file
diff --git a/scripts/legacy-vs-blobs/data/persistent/100_100k.json b/scripts/legacy-vs-blobs/data/persistent/100_100k.json
new file mode 100644
index 00000000..48f0e077
--- /dev/null
+++ b/scripts/legacy-vs-blobs/data/persistent/100_100k.json
@@ -0,0 +1,73 @@
+{
+    "machine_info": {
+        "host": "pajeh", 
+        "python_compiler": "GCC 6.3.0 20170118", 
+        "python_build": [
+            "default", 
+            "Jan 19 2017 14:48:08"
+        ], 
+        "python_implementation": "CPython", 
+        "processor": "", 
+        "cpu": {
+            "hardware": "unknown", 
+            "vendor_id": "GenuineIntel", 
+            "brand": "Intel(R) Core(TM) i7-4600U CPU @ 2.10GHz"
+        }, 
+        "node": "pajeh", 
+        "system": "Linux", 
+        "release": "4.9.0-3-amd64", 
+        "python_version": "2.7.13", 
+        "machine": "x86_64", 
+        "python_implementation_version": "2.7.13"
+    }, 
+    "commit_info": {
+        "branch": "8856", 
+        "author_time": "2017-09-02T02:38:57-03:00", 
+        "time": "2017-09-02T02:38:57-03:00", 
+        "id": "1e132568857e6c2d4fbbbf886d3b7ab2009c3c9e", 
+        "project": "soledad", 
+        "dirty": true
+    }, 
+    "version": "3.1.1", 
+    "benchmarks": [
+        {
+            "group": "test_legacy_vs_blobs_100_100k", 
+            "params": null, 
+            "name": "test_blobs_100_100k", 
+            "stats": {
+                "total": 20.738121271133423, 
+                "stddev_outliers": 2, 
+                "stddev": 0.9161744731162039, 
+                "hd15iqr": 6.378293991088867, 
+                "rounds": 4, 
+                "max": 6.378293991088867, 
+                "median": 5.054657578468323, 
+                "iqr": 1.3696125745773315, 
+                "q3": 5.8693366050720215, 
+                "q1": 4.49972403049469, 
+                "mean": 5.184530317783356, 
+                "ops": 0.1928815029916827, 
+                "iqr_outliers": 0, 
+                "min": 4.25051212310791, 
+                "iterations": 1, 
+                "ld15iqr": 4.25051212310791, 
+                "outliers": "2;0"
+            }, 
+            "options": {
+                "timer": "time", 
+                "disable_gc": false, 
+                "warmup": false, 
+                "max_time": 1.0, 
+                "min_rounds": 5, 
+                "min_time": 5e-06
+            }, 
+            "param": null, 
+            "extra_info": {
+                "cpu_percent": 24.7, 
+                "doc": ""
+            }, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_blobs_100_100k"
+        }
+    ], 
+    "datetime": "2017-09-02T05:56:28.497384"
+}
\ No newline at end of file
diff --git a/scripts/legacy-vs-blobs/data/persistent/10_1000k.json b/scripts/legacy-vs-blobs/data/persistent/10_1000k.json
new file mode 100644
index 00000000..529cca8c
--- /dev/null
+++ b/scripts/legacy-vs-blobs/data/persistent/10_1000k.json
@@ -0,0 +1,73 @@
+{
+    "machine_info": {
+        "python_compiler": "GCC 6.3.0 20170118", 
+        "python_implementation_version": "2.7.13", 
+        "node": "pajeh", 
+        "machine": "x86_64", 
+        "host": "pajeh", 
+        "cpu": {
+            "hardware": "unknown", 
+            "vendor_id": "GenuineIntel", 
+            "brand": "Intel(R) Core(TM) i7-4600U CPU @ 2.10GHz"
+        }, 
+        "system": "Linux", 
+        "python_build": [
+            "default", 
+            "Jan 19 2017 14:48:08"
+        ], 
+        "release": "4.9.0-3-amd64", 
+        "python_version": "2.7.13", 
+        "processor": "", 
+        "python_implementation": "CPython"
+    }, 
+    "commit_info": {
+        "id": "1e132568857e6c2d4fbbbf886d3b7ab2009c3c9e", 
+        "time": "2017-09-02T02:38:57-03:00", 
+        "dirty": true, 
+        "branch": "8856", 
+        "author_time": "2017-09-02T02:38:57-03:00", 
+        "project": "soledad"
+    }, 
+    "version": "3.1.1", 
+    "benchmarks": [
+        {
+            "stats": {
+                "ops": 0.4919825789326136, 
+                "stddev": 0.1042862860397157, 
+                "mean": 2.032592296600342, 
+                "stddev_outliers": 2, 
+                "q3": 2.112483024597168, 
+                "q1": 1.9527015686035156, 
+                "rounds": 4, 
+                "max": 2.1607320308685303, 
+                "hd15iqr": 2.1607320308685303, 
+                "iqr_outliers": 0, 
+                "ld15iqr": 1.91776704788208, 
+                "iterations": 1, 
+                "median": 2.0259350538253784, 
+                "iqr": 0.15978145599365234, 
+                "total": 8.130369186401367, 
+                "outliers": "2;0", 
+                "min": 1.91776704788208
+            }, 
+            "param": null, 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_blobs_10_1000k", 
+            "params": null, 
+            "name": "test_blobs_10_1000k", 
+            "group": "test_legacy_vs_blobs_10_1000k", 
+            "options": {
+                "disable_gc": false, 
+                "min_rounds": 5, 
+                "timer": "time", 
+                "warmup": false, 
+                "min_time": 5e-06, 
+                "max_time": 1.0
+            }, 
+            "extra_info": {
+                "cpu_percent": 36.3, 
+                "doc": ""
+            }
+        }
+    ], 
+    "datetime": "2017-09-02T05:57:48.679847"
+}
\ No newline at end of file
diff --git a/scripts/legacy-vs-blobs/data/persistent/1_10000k.json b/scripts/legacy-vs-blobs/data/persistent/1_10000k.json
new file mode 100644
index 00000000..30d362ea
--- /dev/null
+++ b/scripts/legacy-vs-blobs/data/persistent/1_10000k.json
@@ -0,0 +1,73 @@
+{
+    "machine_info": {
+        "python_compiler": "GCC 6.3.0 20170118", 
+        "host": "pajeh", 
+        "python_build": [
+            "default", 
+            "Jan 19 2017 14:48:08"
+        ], 
+        "cpu": {
+            "hardware": "unknown", 
+            "vendor_id": "GenuineIntel", 
+            "brand": "Intel(R) Core(TM) i7-4600U CPU @ 2.10GHz"
+        }, 
+        "node": "pajeh", 
+        "system": "Linux", 
+        "release": "4.9.0-3-amd64", 
+        "python_implementation": "CPython", 
+        "processor": "", 
+        "python_implementation_version": "2.7.13", 
+        "python_version": "2.7.13", 
+        "machine": "x86_64"
+    }, 
+    "commit_info": {
+        "branch": "8856", 
+        "author_time": "2017-09-02T02:38:57-03:00", 
+        "dirty": true, 
+        "time": "2017-09-02T02:38:57-03:00", 
+        "id": "1e132568857e6c2d4fbbbf886d3b7ab2009c3c9e", 
+        "project": "soledad"
+    }, 
+    "version": "3.1.1", 
+    "benchmarks": [
+        {
+            "stats": {
+                "max": 2.1960339546203613, 
+                "median": 2.1957188844680786, 
+                "stddev_outliers": 1, 
+                "q3": 2.19601047039032, 
+                "iqr_outliers": 0, 
+                "q1": 2.194662928581238, 
+                "ld15iqr": 2.1938750743865967, 
+                "ops": 0.4555109930218146, 
+                "stddev": 0.001009686487541645, 
+                "hd15iqr": 2.1960339546203613, 
+                "rounds": 4, 
+                "iterations": 1, 
+                "total": 8.781346797943115, 
+                "mean": 2.195336699485779, 
+                "iqr": 0.0013475418090820312, 
+                "outliers": "1;0", 
+                "min": 2.1938750743865967
+            }, 
+            "options": {
+                "warmup": false, 
+                "max_time": 1.0, 
+                "timer": "time", 
+                "min_rounds": 5, 
+                "min_time": 5e-06, 
+                "disable_gc": false
+            }, 
+            "group": "test_legacy_vs_blobs_1_10000k", 
+            "params": null, 
+            "name": "test_blobs_1_10000k", 
+            "fullname": "tests/benchmarks/test_legacy_vs_blobs.py::test_blobs_1_10000k", 
+            "param": null, 
+            "extra_info": {
+                "cpu_percent": 40.2, 
+                "doc": ""
+            }
+        }
+    ], 
+    "datetime": "2017-09-02T06:00:45.837579"
+}
\ No newline at end of file
diff --git a/scripts/legacy-vs-blobs/legacy-vs-blobs.png b/scripts/legacy-vs-blobs/legacy-vs-blobs.png
new file mode 100644
index 00000000..6de5bab8
--- /dev/null
+++ b/scripts/legacy-vs-blobs/legacy-vs-blobs.png
diff --git a/scripts/legacy-vs-blobs/legacy-vs-blobs.py b/scripts/legacy-vs-blobs/legacy-vs-blobs.py
new file mode 100755
index 00000000..c10ac269
--- /dev/null
+++ b/scripts/legacy-vs-blobs/legacy-vs-blobs.py
@@ -0,0 +1,124 @@
+#!/usr/bin/env python
+
+# Plot bars comparing different implementations of mail pipeline.
+#
+# This script can be improved to account for arbitrary number of data sets, but
+# it is not doing it right now.
+
+import json
+import numpy as np
+import matplotlib.pyplot as plt
+
+# make a prettier graph
+from mpltools import style
+style.use('ggplot')
+
+OUTPUT_FILENAME = 'legacy-vs-blobs.png'
+
+# each value below will generate one bar for each for each (amount, size) pair.
+# The script expects to find files in ./data/SET/ for each set of
+# implementations.
+#
+# The baseline values will be the legacy results in ./data/no-cache/.
+
+graphs = [
+    '1_10000k',
+    '10_1000k',
+    '100_100k',
+    '1000_10k',
+]
+
+
+# the JSON structure returned by the following function is ugly, but the
+# original JSONs are even uglier, so this is here just to make the life of the
+# script easier.
+#
+# We want to have something like:
+#
+#   data[variation][graph][implementation] = <stats>
+#
+# Where:
+#
+#   - variation is one data set under ./data (i.e. no-cache, cache, persistent,
+#     etc).
+#   - graph is one of the values in graphs variable above.
+#   - implementation is either legacy or blobs (we just need legacy for the
+#     no-cache variation, as that is the one we are using as baseline.
+
+def get_data():
+    folders = ['cache', 'no-cache', 'persistent']
+    data = {}
+    for folder in folders:
+        data[folder] = {}
+        for graph in graphs:
+            with open('data/%s/%s.json' % (folder, graph)) as f:
+                d = json.loads(f.read())
+                benchmarks = d['benchmarks']
+            data[folder][graph] = {}
+            for t in ['blobs', 'legacy']:
+                result = filter(lambda b: t in b['name'], benchmarks)
+                if result:
+                    result = result.pop()
+                    data[folder][graph][t] = result['stats']
+    return data
+
+
+def plot_data(data):
+
+    N = 4
+
+    # this is our baseline (i.e. legacy / legacy)
+    absolutes = (1, 1, 1, 1)
+
+    ind = np.arange(N)  # the x locations for the groups
+    width = 0.20        # the width of the bars
+
+    fig, ax = plt.subplots()
+    rects1 = ax.bar(ind, absolutes, width)
+
+    # for each graph, calculate the ratios
+    ratios = {'no-cache': [], 'cache': [], 'persistent': []}
+    for graph in graphs:
+        legacy = data['no-cache'][graph]['legacy']['mean']
+
+        # calculate ratios for no-cache / legacy
+        ratio = data['no-cache'][graph]['blobs']['mean'] / legacy
+        ratios['no-cache'].append(ratio)
+
+        # calculate ratios for cache / legacy
+        ratio = data['cache'][graph]['blobs']['mean'] / legacy
+        ratios['cache'].append(ratio)
+
+        # calculate ratios for persistent / legacy
+        ratio = data['persistent'][graph]['blobs']['mean'] / legacy
+        ratios['persistent'].append(ratio)
+
+    # create the boxes with the ratios
+    nocache = tuple(ratios['no-cache'])
+    rects2 = ax.bar(ind + width, nocache, width)
+
+    cache = tuple(ratios['cache'])
+    rects3 = ax.bar(ind + (2 * width), cache, width)
+
+    persistent = tuple(ratios['persistent'])
+    rects4 = ax.bar(ind + (3 * width), persistent, width)
+
+    # add some text for labels, title and axes ticks
+    ax.set_ylabel('Normalized execution time')
+    ax.set_title('Legacy vs Blobs mail pipeline')
+    ax.set_xticks(ind + (1.5 * width))
+    ax.set_xticklabels(tuple(map(lambda name: name.replace('_', ' '), graphs)))
+
+    ax.legend(
+        (rects1[0], rects2[0], rects3[0], rects4[0]),
+        ('legacy', 'blobs', 'blobs + session cache',
+         'blobs + session cache + persistent http'))
+    # ax.grid()
+
+    plt.savefig(OUTPUT_FILENAME)
+    # plt.show()
+
+
+if __name__ == '__main__':
+    data = get_data()
+    plot_data(data)
diff --git a/scripts/legacy-vs-blobs/makefile b/scripts/legacy-vs-blobs/makefile
new file mode 100644
index 00000000..649e181c
--- /dev/null
+++ b/scripts/legacy-vs-blobs/makefile
@@ -0,0 +1,14 @@
+VIRTUALENV_NAME = plot
+VIRTUALENV_ROOT = ~/.virtualenvs/$(VIRTUALENV_NAME)
+
+all: graph
+
+env:
+	if [ ! -d $(VIRTUALENV_ROOT) ]; then \
+	  mkdir -p $$(dirname $(VIRTUALENV_ROOT)); \
+	  virtualenv $(VIRTUALENV_ROOT); \
+	  $(VIRTUALENV_ROOT)/bin/pip install requirements.pip; \
+	fi
+
+graph: env
+	$(VIRTUALENV_ROOT)/bin/python ./legacy-vs-blobs.py
diff --git a/scripts/legacy-vs-blobs/requirements.pip b/scripts/legacy-vs-blobs/requirements.pip
new file mode 100644
index 00000000..db5d81e0
--- /dev/null
+++ b/scripts/legacy-vs-blobs/requirements.pip
@@ -0,0 +1,2 @@
+matplotlib
+numpy