Initial update. Still a work in progress

Still working on this, wanted to get the basics up

README.md

@@ -9,7 +9,12 @@ cd image
 docker build -t px-bench .
 docker push ...
 ```
-* Edit `px-bench.yml` to set the ConfigMap `env` to set desired values. If necessary, update `image:` to reflect the image you built.
+* Create a namespace for your benchmarking, and set your context to it (or ensure that you are applying all YAML below to that namespace)
+* TBD: Create a wrapper script that will create the namespace and apply all YAML in order and with correct timing!
+* Copy an example `px-bench-env.yml` from the examples directory. There are various examples for different clound enviroments.
+* Edit `px-bench-env.yml` to set the ConfigMap `env` to set desired values. If necessary, update `image:` to reflect the image you built.
+* `kubectl apply -f px-bench.yml` to start apply the configuration settings.
+* Do NOT edit `px-bench-main.yml` (unless you are attempting to change the behavior of the benchmark!)
 * In order to consume most of the available RAM so it is not used for buffering, run `kubectl apply -f chewram.yml`.
 * Wait for `kubectl get pod -n chewram` for all the pods to show as `1/1 Running`.
 * `kubectl apply -f px-bench.yml` to start the run.

TheoryOfOps.md

@@ -0,0 +1,6 @@
+# Theory of Operations 
+
+## This document gives guidance on how to use px-bench and
+## the reasoning behind its architecture and default settings.
+
+PLACEHOLDER

examples/aws/px-bench-env.yml

@@ -0,0 +1,61 @@
+######
+# AWS Example. Valid for EKS or other K8s orchestrators hosted in AWS.
+#
+# See https://docs.aws.amazon.com/eks/latest/userguide/storage.html for links to various drivers available in EKS.
+#
+# EBS: Since Portworx would consume EBS volumes as Portworx Clouddrives, attempting "apples to apples" comparisons
+#      between AWS native storage and Portworx  should be done with EBS volumes that are the same size as the
+#      Portworx clouddrive used for the pool. 
+#      Installation of the EBS CSI is documented here: https://docs.aws.amazon.com/eks/latest/userguide/ebs-csi.html
+#
+#      The configmap below (see the Iteration Control section near the bottom) references an EBS storageclass 
+#      called "ebs-sc", update it to your desired AWS storageclass. This example will benchmark "ebs-sc" vs.
+#      the Portworx storageclass "px-csi-db"
+#
+---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: env
+######
+# Options that control the underlying fio benchmark utility. These equate to the same-named option in fio
+#
+data:
+  file_size: "10GiB"        # size of region of IO testing, will also be the size of the file
+  pvc_size: "150Gi"         # If the storageclass is backed by a larger disk from infrastructure,
+                            # such as a Portworx pool, then pvc_size should be the same size as
+                            # the backing storage if it's desired to be testing the same
+                            # size and performance storage (See TheoryOfOps.md)
+  runtime: "20"             # runtime in seconds. Start low the first time to ensure everything
+                            # is working as planned. Then increase to desired runtime.
+  iodepth: "128"
+  rwmixread: "60"           # only used for readwrite and randrw workloads
+  rwmixwrite: "40"          # should be 100 - rwmixread
+  ramp_time: "5"            # ramp time per test. As with runtime, keep low for quick tests
+                            # and then increase to desired ramp_time for later tests
+
+  # Less commonly changed fio options
+  ioengine: "libaio"
+  direct: "1"
+  fsync_on_close: "1"
+  numjobs: "1"
+  group_reporting: "1"
+
+  ######
+  # px-bench specific options. These are not fio options
+  mixed_workloads: "rw readwrite randrw"   # rw is a synonym for readwrite
+
+  ######
+  # Iteration Control: Each variable will be looped over to create the jobs to be run.
+  #                    Do not change the variable name. The variable must consist of one or more space
+  #                    delimited strings that are legal values for fio to use in a job file.
+  #
+  #                    For now, the only supported variables here are blocksize and randwrite
+  #
+  #                    Example: If you have 2 blocksize values, and 3 randwrite values, this will result in 6 jobfiles.
+
+  blocksize_list: "4k 16k 256k"            # We have not tested fio's support for different values for read and write ("bs=8k,32k")
+  readwrite_list: "read write readwrite"   # valid values are read, write, randread, readwrite (or rw), randrw.
+  storageclass_list: "ebs-sc px-csi-db"    # Existing storageclasses for use by the benchmarks
+  warmup_blocksize: 256k
+  warmup_loops: "1"

examples/generic/px-bench-env.yml

@@ -0,0 +1,48 @@
+---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: env
+
+######
+# Options that control the underlying fio benchmark utility. These equate to the same-named option in fio
+#
+data:
+  file_size: "10GiB"        # size of region of IO testing, will also be the size of the file
+  pvc_size: "50Gi"          # If the storageclass is backed by a larger disk from infrastructure,
+                            # such as a Portworx pool, then pvc_size should be the same size as
+                            # the backing storage if it's desired to be testing the same
+                            # size and performance storage (See TheoryOfOps.md)
+  runtime: "20"             # runtime in seconds. Start low the first time to ensure everything
+                            # is working as planned. Then increase to desired runtime.
+  iodepth: "128"
+  rwmixread: "60"           # only used for readwrite and randrw workloads
+  rwmixwrite: "40"          # should be 100 - rwmixread
+  ramp_time: "5"            # ramp time per test. As with runtime, keep low for quick tests
+                            # and then increase to desired ramp_time for later tests
+
+  # Less commonly changed fio options
+  ioengine: "libaio"
+  direct: "1"
+  fsync_on_close: "1"
+  numjobs: "1"
+  group_reporting: "1"
+
+  ######
+  # px-bench specific options. These are not fio options
+  mixed_workloads: "rw readwrite randrw"   # rw is a synonym for readwrite
+
+  ######
+  # Iteration Control: Each variable will be looped over to create the jobs to be run.
+  # Do not change the variable name. The variable must consist of one or more space
+  # delimited strings that are legal values for fio to use in a job file.
+  #
+  # For now, the only supported variables here are blocksize and randwrite
+  #
+  # Example: If you have 2 blocksize values, and 3 randwrite values, this will result in 6 jobfiles.
+
+  blocksize_list: "4k 16k 256k"                 # We have not tested fio's support for different values for read and write ("bs=8k,32k")
+  readwrite_list: "read write readwrite"        # valid values are read, write, randread, readwrite (or rw), randrw.
+  storageclass_list: "storageclass-1 storageclass-2"    # Existing storageclasses for use by the benchmarks
+  warmup_blocksize: 256k
+  warmup_loops: "1"

px-bench.yml → px-bench-main.yml

@@ -1,53 +1,8 @@
-apiVersion: v1
-kind: Namespace
-metadata:
-  name: px-bench
----
-apiVersion: v1
-kind: ConfigMap
-metadata:
-  name: env
-  namespace: px-bench
-data:
-  file_size: "10GiB"             # size of region of IO testing, will also be the size of the file
-  pvc_size: "50Gi"
-  runtime: "20"             # runtime in seconds
-  iodepth: "128"
-  rwmixread: "60"           # only used for readwrite and randrw workloads
-  rwmixwrite: "40"
-  ramp_time: "5"            # ramp time per test
-
-  # Less commonly changed options
-  ioengine: "libaio"
-  direct: "1"
-  fsync_on_close: "1"
-  numjobs: "1"
-  group_reporting: "1"
-
-  ######
-  # px-bench specific options. These are not fio options
-  mixed_workloads: "rw readwrite randrw"   # rw is a synonym for readwrite
-
-  ######
-  # Iteration Control: Each variable will be looped over to create the jobs to be run.
-  # Do not change the variable name. The variable must consist of one or more space
-  # delimited strings that are legal values for fio to use in a job file.
-  #
-  # For now, the only supported variables here are blocksize and randwrite
-  #
-  # Example: If you have 2 blocksize values, and 3 randwrite values, this will result in 6 jobfiles.
-
-  blocksize_list: "4k 16k 256k"                 # We have not tested fio's support for different values for read and write ("bs=8k,32k")
-  readwrite_list: "read write readwrite"        # valid values are read, write, randread, readwrite (or rw), randrw.
-  storageclass_list: "px-db px-replicated"    # Existing storageclasses for use by the benchmarks
-  warmup_blocksize: 256k
-  warmup_loops: "1"
 ---
 apiVersion: v1
 kind: ConfigMap
 metadata:
   name: files
-  namespace: px-bench
 data:
   run.sh: |-
     #!/usr/bin/bash
@@ -110,7 +65,6 @@ data:
     kind: Job
     metadata:
       name: fio
-      namespace: px-bench
     spec:
       template:
         metadata:
@@ -202,13 +156,11 @@ apiVersion: v1
 kind: ServiceAccount
 metadata:
   name: px-bench
-  namespace: px-bench
 ---
 apiVersion: rbac.authorization.k8s.io/v1
 kind: Role
 metadata:
   name: px-bench
-  namespace: px-bench
 rules:
   - apiGroups: [""]
     resources:
@@ -229,7 +181,6 @@ apiVersion: rbac.authorization.k8s.io/v1
 kind: RoleBinding
 metadata:
   name: px-bench
-  namespace: px-bench
 subjects:
   - kind: ServiceAccount
     name: px-bench
@@ -242,7 +193,6 @@ apiVersion: batch/v1
 kind: Job
 metadata:
   name: run
-  namespace: px-bench
 spec:
   template:
     metadata: