diff --git a/metrics/report/README.md b/metrics/report/README.md
new file mode 100644
index 00000000..0986db2d
--- /dev/null
+++ b/metrics/report/README.md
@@ -0,0 +1,73 @@
+* [cloud-native-setup metrics report generator](#cloud-native-setup-metrics-report-generator)
+   * [Data gathering](#data-gathering)
+   * [Report generation](#report-generation)
+   * [Debugging and development](#debugging-and-development)
+
+# cloud-native-setup metrics report generator
+
+The files within this directory can be used to generate a 'metrics report'
+for Kubernetes.
+
+The primary workflow consists of two stages:
+
+1) Run the provided report metrics data gathering scripts on the system(s) you wish
+to analyze.
+2) Run the provided report generation script to analyze the data and generate a
+report file.
+
+## Data gathering
+
+Data gathering is provided by the `grabdata.sh` script. When run, this script
+executes a set of tests from the `cloud-native-setup/metrics` directory. The JSON results files
+will be placed into the `cloud-native-setup/metrics/results` directory.
+
+Once the results are generated, create a suitably named subdirectory of
+`tests/metrics/results`, and move the JSON files into it.
+
+Repeat this process if you want to compare multiple sets of results. Note, the
+report generation scripts process all subfolders of `tests/metrics/results` when
+generating the report.
+
+You can restrict the subset of tests run by `grabdata.sh` via its commandline parameters:
+
+| Option | Description |
+| ------ | ----------- |
+| -a | Run all tests (default) |
+| -s | Run the scaling tests |
+| -h | Print this help |
+
+## Report generation
+
+Report generation is provided by the `makereport.sh` script. By default this script 
+processes all subdirectories of the `cloud-native-setup/metrics/results` directory to generate the report.
+To run in the default mode, execute the following:
+
+```sh
+$ ./makereport.sh
+```
+
+The report generation tool uses [Rmarkdown](https://github.com/rstudio/rmarkdown),
+[R](https://www.r-project.org/about.html) and [pandoc](https://pandoc.org/) to produce
+a PDF report. To avoid the need for all users to set up a working environment
+with all the necessary tooling, the `makereport.sh` script utilises a `Dockerfile` with
+the environment pre-defined in order to produce the report. Thus, you need to
+have Docker installed on your system in order to run the report generation.
+
+The resulting `metrics_report.pdf` is generated into the `output` subdir of the `report`
+directory.
+
+## Debugging and development
+
+To aid in script development and debugging, the `makereport.sh` script offers a debug
+facility via the `-d` command line option. Using this option will place you into a `bash`
+shell within the running `Dockerfile` image used to generate the report. From there you
+can examine the Docker image environment, and execute the generation scripts. E.g., to
+test the `scaling.R` script, you can execute:
+
+```bash
+$ makereport.sh -d
+# R
+> source('/inputdir/Env.R')
+> source('/scripts/scaling.R')
+```
+
diff --git a/metrics/report/grabdata.sh b/metrics/report/grabdata.sh
new file mode 100755
index 00000000..b67f8842
--- /dev/null
+++ b/metrics/report/grabdata.sh
@@ -0,0 +1,97 @@
+#!/bin/bash
+# Copyright (c) 2019 Intel Corporation
+#
+# SPDX-License-Identifier: Apache-2.0
+
+# Run a set of the metrics tests to gather data to be used with the report
+# generator. The general ideal is to have the tests configured to generate
+# useful, meaninful and repeatable (stable, with minimised variance) results.
+# If the tests have to be run more or longer to achieve that, then generally
+# that is fine - this test is not intended to be quick, it is intended to
+# be repeatable.
+
+# Note - no 'set -e' in this file - if one of the metrics tests fails
+# then we wish to continue to try the rest.
+# Finally at the end, in some situations, we explicitly exit with a
+# failure code if necessary.
+
+SCRIPT_DIR=$(dirname "$(readlink -f "$0")")
+source "${SCRIPT_DIR}/../lib/common.bash"
+RESULTS_DIR=${SCRIPT_DIR}/../results
+
+# By default we run all the tests
+RUN_ALL=1
+
+help() {
+	usage=$(cat << EOF
+Usage: $0 [-h] [options]
+   Description:
+        This script gathers a number of metrics for use in the
+        report generation script. Which tests are run can be
+        configured on the commandline. Specifically enabling
+        individual tests will disable the 'all' option, unless
+        'all' is also specified last.
+   Options:
+        -a,         Run all tests (default).
+        -h,         Print this help.
+        -s,         Run the scaling tests.
+EOF
+)
+	echo "$usage"
+}
+
+# Set up the initial state
+init() {
+	metrics_onetime_init
+
+	local OPTIND
+	while getopts "ahs" opt;do
+		case ${opt} in
+		a)
+		    RUN_ALL=1
+		    ;;
+		h)
+		    help
+		    exit 0;
+		    ;;
+		s)
+		    RUN_SCALING=1
+		    RUN_ALL=
+		    ;;
+		?)
+		    # parse failure
+		    help
+		    die "Failed to parse arguments"
+		    ;;
+		esac
+	done
+	shift $((OPTIND-1))
+}
+
+run_scaling() {
+	echo "Running scaling tests"
+
+	(cd scaling; ./k8s_scale.sh)
+}
+
+# Execute metrics scripts
+run() {
+	pushd "$SCRIPT_DIR/.."
+
+	if [ -n "$RUN_ALL" ] || [ -n "$RUN_SCALING" ]; then
+		run_scaling
+	fi
+
+	popd
+}
+
+finish() {
+	echo "Now please create a suitably descriptively named subdirectory in"
+	echo "$RESULTS_DIR and copy the .json results files into it before running"
+	echo "this script again."
+}
+
+init "$@"
+run
+finish
+
diff --git a/metrics/report/makereport.sh b/metrics/report/makereport.sh
new file mode 100755
index 00000000..70900799
--- /dev/null
+++ b/metrics/report/makereport.sh
@@ -0,0 +1,72 @@
+#!/bin/bash
+# Copyright (c) 2019 Intel Corporation
+#
+# SPDX-License-Identifier: Apache-2.0
+
+# Take the data found in subdirectories of the metrics 'results' directory,
+# and turn them into a PDF report. Use a Dockerfile containing all the tooling
+# and scripts we need to do that.
+
+set -e
+
+SCRIPT_PATH=$(dirname "$(readlink -f "$0")")
+source "${SCRIPT_PATH}/../lib/common.bash"
+
+IMAGE="${IMAGE:-metrics-report}"
+DOCKERFILE="${SCRIPT_PATH}/report_dockerfile/Dockerfile"
+
+HOSTINPUTDIR="${SCRIPT_PATH}/../results"
+RENVFILE="${HOSTINPUTDIR}/Env.R"
+HOSTOUTPUTDIR="${SCRIPT_PATH}/output"
+
+GUESTINPUTDIR="/inputdir/"
+GUESTOUTPUTDIR="/outputdir/"
+
+setup() {
+	echo "Checking subdirectories"
+	check_subdir="$(ls -dx ${HOSTINPUTDIR}/*/ 2> /dev/null | wc -l)"
+	if [ $check_subdir -eq 0 ]; then
+		die "No subdirs in [${HOSTINPUTDIR}] to read results from."
+	fi
+
+	echo "Checking Dockerfile"
+	check_dockerfiles_images "$IMAGE" "$DOCKERFILE"
+
+	mkdir -p "$HOSTOUTPUTDIR" && true
+
+	echo "inputdir=\"${GUESTINPUTDIR}\"" > ${RENVFILE}
+	echo "outputdir=\"${GUESTOUTPUTDIR}\"" >> ${RENVFILE}
+
+	# A bit of a hack to get an R syntax'd list of dirs to process
+	# Also, need it as not host-side dir path - so short relative names
+	resultdirs="$(cd ${HOSTINPUTDIR}; ls -dx */)"
+	resultdirslist=$(echo ${resultdirs} | sed 's/ \+/", "/g')
+	echo "resultdirs=c(" >> ${RENVFILE}
+	echo "	\"${resultdirslist}\"" >> ${RENVFILE}
+	echo ")" >> ${RENVFILE}
+}
+
+run() {
+	docker run -ti --rm -v ${HOSTINPUTDIR}:${GUESTINPUTDIR} -v ${HOSTOUTPUTDIR}:${GUESTOUTPUTDIR} ${IMAGE} ${extra_command}
+	ls -la ${HOSTOUTPUTDIR}/*
+}
+
+main() {
+
+	local OPTIND
+	while getopts "d" opt;do
+		case ${opt} in
+		d)
+			# In debug mode, run a shell instead of the default report generation
+			extra_command="bash"
+			;;
+		esac
+	done
+	shift $((OPTIND-1))
+
+	setup
+	run
+}
+
+main "$@"
+
diff --git a/metrics/report/report_dockerfile/Dockerfile b/metrics/report/report_dockerfile/Dockerfile
new file mode 100755
index 00000000..bf1c1717
--- /dev/null
+++ b/metrics/report/report_dockerfile/Dockerfile
@@ -0,0 +1,40 @@
+# Copyright (c) 2018-2019 Intel Corporation
+#
+# SPDX-License-Identifier: Apache-2.0
+
+# Set up an Ubuntu image with the components needed to generate a
+# metrics report. That includes:
+#  - R
+#  - The R 'tidyverse'
+#  - pandoc
+#  - The report generation R files and helper scripts
+
+# Start with the base rocker tidyverse.
+# We would have used the 'verse' base, that already has some of the docs processing
+# installed, but I could not figure out how to add in the extra bits we needed to
+# the lite tex version is uses.
+FROM rocker/tidyverse
+
+# Version of the Dockerfile
+LABEL DOCKERFILE_VERSION="1.0"
+
+# Without this some of the package installs stop to try and ask questions...
+ENV DEBIAN_FRONTEND=noninteractive
+
+# Install the extra doc processing parts we need for our Rmarkdown PDF flow.
+RUN apt-get update -qq && \
+  apt-get install -y \
+    texlive-latex-base \
+    texlive-fonts-recommended \
+    latex-xcolor
+
+# Install the extra R packages we need.
+RUN install2.r --error --deps TRUE \
+	gridExtra \
+	ggpubr
+
+# Pull in our actual worker scripts
+COPY . /scripts
+
+# By default generate the report
+CMD ["/scripts/genreport.sh"]
diff --git a/metrics/report/report_dockerfile/genreport.sh b/metrics/report/report_dockerfile/genreport.sh
new file mode 100755
index 00000000..c27379b1
--- /dev/null
+++ b/metrics/report/report_dockerfile/genreport.sh
@@ -0,0 +1,14 @@
+#!/bin/bash
+# Copyright (c) 2018-2019 Intel Corporation
+#
+# SPDX-License-Identifier: Apache-2.0
+
+REPORTNAME="metrics_report.pdf"
+
+cd scripts
+
+Rscript --slave -e "library(knitr);knit('metrics_report.Rmd')"
+Rscript --slave -e "library(knitr);pandoc('metrics_report.md', format='latex')"
+
+cp /scripts/${REPORTNAME} /outputdir
+echo "The report, named ${REPORTNAME}, can be found in the output directory"
diff --git a/metrics/report/report_dockerfile/metrics_report.Rmd b/metrics/report/report_dockerfile/metrics_report.Rmd
new file mode 100644
index 00000000..950bbc1e
--- /dev/null
+++ b/metrics/report/report_dockerfile/metrics_report.Rmd
@@ -0,0 +1,40 @@
+---
+# Copyright (c) 2018-2019 Intel Corporation
+#
+# SPDX-License-Identifier: Apache-2.0
+#
+title: "Kubernetes metrics report"
+author: "Auto generated"
+date: "`r format(Sys.time(), '%d %B, %Y')`"
+output:
+  pdf_document:
+urlcolor: blue
+---
+
+```{r setup, include=FALSE}
+#Set these opts to get pdf images which fit into beamer slides better
+opts_chunk$set(dev = 'pdf')
+# Pick up any env set by the invoking script, such as the root dir of the
+# results data tree
+source("/inputdir/Env.R")
+```
+\pagebreak
+
+# Introduction
+This report compares the metrics between multiple sets of data generated from
+the [cloud-native-setup report generation scripts](https://github.com/clearlinux/cloud-native-setup/metrics/report/README.md).
+
+This report was generated using the data from the **`r resultdirs`** results directories.
+
+\pagebreak
+
+# Runtime scaling
+This [test](https://github.com/clearlinux/cloud-native-setup/metrics/scaling/k8s_scale.sh)
+measures the system memory 'free' reduction, CPU idle % and pod boot time as it launches more
+and more idle `busybox` pods on a single node Kubernetes cluster.
+
+> Note: CPU % is measured as a system whole - 100% represents *all* CPUs on the node.
+
+```{r, echo=FALSE, fig.cap="K8S scaling"}
+source('scaling.R')
+```
diff --git a/metrics/report/report_dockerfile/scaling.R b/metrics/report/report_dockerfile/scaling.R
new file mode 100755
index 00000000..d93a2111
--- /dev/null
+++ b/metrics/report/report_dockerfile/scaling.R
@@ -0,0 +1,198 @@
+#!/usr/bin/env Rscript
+# Copyright (c) 2018-2019 Intel Corporation
+#
+# SPDX-License-Identifier: Apache-2.0
+
+# Show system memory reduction, and hence container 'density', by analysing the
+# scaling footprint data results and the 'system free' memory.
+
+suppressMessages(suppressWarnings(library(ggplot2)))	# ability to plot nicely.
+							# So we can plot multiple graphs
+library(gridExtra)					# together.
+suppressMessages(suppressWarnings(library(ggpubr)))	# for ggtexttable.
+suppressMessages(library(jsonlite))			# to load the data.
+suppressMessages(library(scales))			# For de-science notation of axis
+
+testnames=c(
+	"k8s-scaling.*"
+)
+
+data=c()
+stats=c()
+rstats=c()
+rstats_names=c()
+cstats=c()
+cstats_names=c()
+
+skip_points=0	# Shall we draw the points as well as lines on the graphs.
+
+# FIXME GRAHAM - bomb if there are no source dirs?!
+
+for (currentdir in resultdirs) {
+	count=1
+	dirstats=c()
+	for (testname in testnames) {
+		matchdir=paste(inputdir, currentdir, sep="")
+		matchfile=paste(testname, '\\.json', sep="")
+		files=list.files(matchdir, pattern=matchfile)
+		if ( length(files) == 0 ) {
+			#warning(paste("Pattern [", matchdir, "/", matchfile, "] matched nothing"))
+		}
+		for (ffound in files) {
+			fname=paste(inputdir, currentdir, ffound, sep="")
+			if ( !file.exists(fname)) {
+				warning(paste("Skipping non-existent file: ", fname))
+				next
+			}
+
+			# Derive the name from the test result dirname
+			datasetname=basename(currentdir)
+
+			# Import the data
+			fdata=fromJSON(fname)
+			# De-nest the test name specific data
+			shortname=substr(ffound, 1, nchar(ffound)-nchar(".json"))
+			fdata=fdata[[shortname]]
+
+			testname=datasetname
+
+			cdata=data.frame(avail_gb=as.numeric(fdata$BootResults$mem_free$Result)/(1024*1024))
+			cdata=cbind(cdata, cpu_idle=as.numeric(fdata$BootResults$cpu_idle$Result))
+			# convert ms to seconds
+			cdata=cbind(cdata, boot_time=as.numeric(fdata$BootResults$launch_time$Result)/1000)
+			# FIXME - we should seq from 0 index
+			if (length(cdata[, "avail_gb"]) > 20) {
+				skip_points=1
+			}
+
+			cdata=cbind(cdata, count=seq_len(length(cdata[, "avail_gb"])))
+			cdata=cbind(cdata, testname=rep(testname, length(cdata[, "avail_gb"]) ))
+			cdata=cbind(cdata, dataset=rep(datasetname, length(cdata[, "avail_gb"]) ))
+
+			# Gather our statistics
+			# '-1' containers, as the first entry should be a data capture of before
+			# the first container was run.
+			# FIXME - once the test starts to store a stats baseline in slot 0, then
+			# we should re-enable the '-1'
+			#sdata=data.frame(num_containers=length(cdata[, "avail_gb"])-1)
+			sdata=data.frame(num_containers=length(cdata[, "avail_gb"]))
+			# Work out memory reduction by subtracting last (most consumed) from
+			# first (which should be 0-containers)
+			sdata=cbind(sdata, mem_consumed= cdata[, "avail_gb"][1] -
+				cdata[, "avail_gb"][length(cdata[, "avail_gb"])])
+			sdata=cbind(sdata, cpu_consumed= cdata[, "cpu_idle"][1] -
+				cdata[, "cpu_idle"][length(cdata[, "cpu_idle"])])
+			sdata=cbind(sdata, boot_time=cdata[, "boot_time"][length(cdata[, "boot_time"])])
+			sdata=cbind(sdata, avg_gb_per_c=sdata$mem_consumed / sdata$num_containers)
+			sdata=cbind(sdata, runtime=testname)
+
+			# Store away as a single set
+			data=rbind(data, cdata)
+			stats=rbind(stats, sdata)
+
+			ms = c(
+				"Test"=testname,
+				"n"=sdata$num_containers,
+				"size"=round((sdata$mem_consumed), 3),
+				"gb/n"=round(sdata$avg_gb_per_c, digits=4),
+				"n/Gb"= round(1 / sdata$avg_gb_per_c, digits=2)
+			)
+
+			cs = c(
+				"Test"=testname,
+				"n"=sdata$num_containers,
+				"cpu"=round(sdata$cpu_consumed, digits=3),
+				"cpu/n"=round((sdata$cpu_consumed / sdata$num_containers), digits=4)
+			)
+
+			rstats=rbind(rstats, ms)
+			cstats=rbind(cstats, cs)
+			count = count + 1
+		}
+	}
+}
+
+#FIXME - if we end up with no data here, we should probably abort cleanly, or we
+# end up spewing errors for trying to access empty arrays etc.
+
+# Set up the text table headers
+colnames(rstats)=c("Test", "n", "Tot_Gb", "avg_Gb", "n_per_Gb")
+colnames(cstats)=c("Test", "n", "Tot_CPU", "avg_CPU")
+
+# Build us a text table of numerical results
+mem_stats_plot = suppressWarnings(ggtexttable(data.frame(rstats),
+	theme=ttheme(base_size=10),
+	rows=NULL
+	))
+
+# plot how samples varioed over  'time'
+mem_line_plot <- ggplot() +
+	geom_line( data=data, aes(count, avail_gb, colour=testname, group=dataset), alpha=0.2) +
+	geom_smooth( data=data, aes(count, avail_gb, colour=testname, group=dataset), se=FALSE, method="loess", size=0.3) +
+	xlab("Pods") +
+	ylab("System Avail (Gb)") +
+	scale_y_continuous(labels=comma) +
+	ggtitle("System Memory free") +
+	#ylim(0, NA) + # For big machines, better to not 0-index
+	theme(axis.text.x=element_text(angle=90))
+
+# If we only have relatively few samples, add points to the plot. Otherwise, skip as
+# the plot becomes far too noisy
+if ( skip_points == 0 ) {
+	mem_line_plot = mem_line_plot + geom_point( data=data, aes(count, avail_gb, colour=testname, group=dataset), alpha=0.3)
+}
+
+cpu_stats_plot = suppressWarnings(ggtexttable(data.frame(cstats),
+	theme=ttheme(base_size=10),
+	rows=NULL
+	))
+
+# plot how samples varioed over  'time'
+cpu_line_plot <- ggplot() +
+	geom_line( data=data, aes(count, cpu_idle, colour=testname, group=dataset), alpha=0.2) +
+	geom_smooth( data=data, aes(count, cpu_idle, colour=testname, group=dataset), se=FALSE, method="loess", size=0.3) +
+	xlab("Pods") +
+	ylab("System CPU Idle (%)") +
+	ggtitle("System CPU usage") +
+	#ylim(0, NA) + # For big machines, better to not 0-index
+	theme(axis.text.x=element_text(angle=90))
+
+if ( skip_points == 0 ) {
+	cpu_line_plot = cpu_line_plot + geom_point( data=data, aes(count, cpu_idle, colour=testname, group=dataset), alpha=0.3)
+}
+
+# Show how boot time changed
+boot_line_plot <- ggplot() +
+	geom_line( data=data, aes(count, boot_time, colour=testname, group=dataset), alpha=0.2) +
+	geom_smooth( data=data, aes(count, boot_time, colour=testname, group=dataset), se=FALSE, method="loess", size=0.3) +
+	xlab("pods") +
+	ylab("Boot time (s)") +
+	ggtitle("Pod boot time") +
+	#ylim(0, NA) + # For big machines, better to not 0-index
+	theme(axis.text.x=element_text(angle=90))
+
+if ( skip_points == 0 ) {
+	cpu_line_plot = cpu_line_plot + geom_point( data=data, aes(count, boot_time, colour=testname, group=dataset), alpha=0.3)
+}
+
+mem_text <- paste("Footprint density statistics")
+mem_text.p <- ggparagraph(text=mem_text, face="italic", size="10", color="black")
+
+cpu_text <- paste("System CPU consumption statistics")
+cpu_text.p <- ggparagraph(text=cpu_text, face="italic", size="10", color="black")
+
+# See https://www.r-bloggers.com/ggplot2-easy-way-to-mix-multiple-graphs-on-the-same-page/ for
+# excellent examples 
+master_plot = grid.arrange(
+	mem_line_plot,
+	cpu_line_plot,
+	mem_stats_plot,
+	cpu_stats_plot,
+	mem_text.p,
+	cpu_text.p,
+	boot_line_plot,
+	zeroGrob(),
+	nrow=4,
+	ncol=2,
+        heights=c(1, 0.8, 0.1, 1) )
+