diff options
| author | ziadh@chromium.org <ziadh@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98> | 2010-07-30 23:47:44 +0000 |
|---|---|---|
| committer | ziadh@chromium.org <ziadh@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98> | 2010-07-30 23:47:44 +0000 |
| commit | fcb7a34a7c030355e21186336a8cfa2ecfd1f378 (patch) | |
| tree | 513f6d44a4e0d29e3cceffb39f37e86b242517e0 /chrome/browser/browser_main.cc | |
| parent | 8b438c0d681001900e4a74cfc744e19a578476f8 (diff) | |
| download | chromium_src-fcb7a34a7c030355e21186336a8cfa2ecfd1f378.zip chromium_src-fcb7a34a7c030355e21186336a8cfa2ecfd1f378.tar.gz chromium_src-fcb7a34a7c030355e21186336a8cfa2ecfd1f378.tar.bz2 | |
Reduce magnitude of non-default group probabilities for 2 A/B tests
I reduced the non-default group probabilities for "ConnCountImpact" and "IdleSktToImpact" by an order of magnitude from [20%, 25%] to [2%, 2.5%] respectively.
BUG=50802
r=jar
Review URL: http://codereview.chromium.org/3056037
git-svn-id: svn://svn.chromium.org/chrome/trunk/src@54411 0039d316-1c4b-4281-b951-d872f2087c98
Diffstat (limited to 'chrome/browser/browser_main.cc')
| -rw-r--r-- | chrome/browser/browser_main.cc | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/chrome/browser/browser_main.cc b/chrome/browser/browser_main.cc index 277cac1..f0976b9 100644 --- a/chrome/browser/browser_main.cc +++ b/chrome/browser/browser_main.cc @@ -175,8 +175,8 @@ void BrowserMainParts::EarlyInitialization() { // on browsing. Too large a value might cause us to run into SYN flood detection // mechanisms. void BrowserMainParts::ConnectionFieldTrial() { - const FieldTrial::Probability kConnectDivisor = 100; - const FieldTrial::Probability kConnectProbability = 20; // 20% probability + const FieldTrial::Probability kConnectDivisor = 1000; + const FieldTrial::Probability kConnectProbability = 20; // 2% probability scoped_refptr<FieldTrial> connect_trial = new FieldTrial("ConnCountImpact", kConnectDivisor); @@ -219,8 +219,8 @@ void BrowserMainParts::ConnectionFieldTrial() { // result in more ERR_CONNECT_RESETs, requiring one RTT to receive the RST // packet and possibly another RTT to re-establish the connection. void BrowserMainParts::SocketTimeoutFieldTrial() { - const FieldTrial::Probability kIdleSocketTimeoutDivisor = 100; - // 25% probability + const FieldTrial::Probability kIdleSocketTimeoutDivisor = 1000; + // 2.5% probability const FieldTrial::Probability kSocketTimeoutProbability = 25; scoped_refptr<FieldTrial> socket_timeout_trial = |