Add uma stats to help evaluate the impact of changes to the quota allocation logic.

storage/common/storage_histograms.cc

Index: storage/common/storage_histograms.cc
diff --git a/storage/common/storage_histograms.cc b/storage/common/storage_histograms.cc
new file mode 100644
index 0000000000000000000000000000000000000000..adb99124fc37f696fcbd83bcda8087d547431d54
--- /dev/null
+++ b/storage/common/storage_histograms.cc
@@ -0,0 +1,36 @@
+// Copyright 2017 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "storage/common/storage_histograms.h"
+
+#include "base/metrics/histogram_macros.h"
+
+namespace storage {
+
+namespace {
+
+base::HistogramBase* GetBytesCountHistogram(const std::string& name) {
+  const int kMin = 1;
+  const int kMax = 10000000;

[jsbell, 2017/05/04 19:43:37]

Hrm, actually... does this mean the histogram maxes out at 10GB read?

Since this is recorded for the lifetime of the Chrome process, that actually
seems like it might be a bit low. I suppose we'll find out?

[jsbell, 2017/05/04 19:48:00]

Never mind, it's per-operation. That should be plenty. :)

It does mean interpreting it will require caution, e.g. 100 reads of 1k and 50
writes of 2k would be a 1:1 ratio but might show up as 2:1 depending on the
bucketing.

+  const int kNumBuckets = 50;
+  return base::Histogram::FactoryGet(
+      name, kMin, kMax, kNumBuckets,
+      base::Histogram::kUmaTargetedHistogramFlag);
+}
+
+}  // anonomous namespace
+
+void RecordBytesWritten(const char* label, int amount) {
+  std::string name("Storage.BytesWritten.");
+  name.append(label);
+  GetBytesCountHistogram(name)->Add(amount);
+}
+
+void RecordBytesRead(const char* label, int amount) {
+  std::string name("Storage.BytesRead.");
+  name.append(label);
+  GetBytesCountHistogram(name)->Add(amount);
+}
+
+}  // namespace storage