Telemetry traffic settings description

25 views
Skip to first unread message

sarath.s...@huawei.com

unread,
Jun 4, 2018, 5:55:40 AM6/4/18
to Chrome benchmarking
Hi,

There is some confusion about NETWORK_CONFIGS in telemetry traffic settings.
In https://cs.chromium.org/chromium/src/third_party/catapult/telemetry/telemetry/page/traffic_setting.py?type=cs&q=NetworkConditionsSelector&sq=package:chromium&g=0&l=10 , bandwidth is given as kbps.
_Configs = collections.namedtuple( '_Configs', ('download_bandwidth_kbps,' 'upload_bandwidth_kbps,' 'round_trip_latency_ms'))

For eg., REGULAR_4G: _Configs(4 * 1024 * 1024 / 8, 3 * 1024 * 1024 / 8, 20),
Can we understand this "4 * 1024 * 1024 / 8" as 524288 Bytes per second.?

Thank you,
Sarath

Ned

unread,
Jun 4, 2018, 6:29:47 PM6/4/18
to sarath.s...@huawei.com, ca...@google.com, Chrome benchmarking
Hmhh, it's unclear what version of networkconditionsSelector.js we were using.

+Andrey Kosyakov what is the source of truth file that describes network throttling setting that devtool use today?

--
You received this message because you are subscribed to the Google Groups "Chrome benchmarking" group.
To unsubscribe from this group and stop receiving emails from it, send an email to benchmarking-d...@chromium.org.
To post to this group, send email to benchmar...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/benchmarking-dev/96675a77-0069-4905-862d-099f28d8415d%40chromium.org.

Andrey Kosyakov

unread,
Jun 4, 2018, 7:17:25 PM6/4/18
to Ned Nguyen, sarath.s...@huawei.com, benchmar...@chromium.org
On Mon, Jun 4, 2018 at 3:29 PM Ned <nedn...@google.com> wrote:
Hmhh, it's unclear what version of networkconditionsSelector.js we were using.

+Andrey Kosyakov what is the source of truth file that describes network throttling setting that devtool use today?

Ned

unread,
Jun 4, 2018, 7:25:54 PM6/4/18
to Andrey Kosyakov, sarath.s...@huawei.com, benchmar...@chromium.org
Hi Andrey, I meant to ask for the file which is similar to the old src/third_party/WebKit/Source/devtools/front_end/components/NetworkConditionsSelector.js

sarath.s...@huawei.com

unread,
Jun 26, 2018, 4:14:24 AM6/26/18
to Chrome benchmarking, ca...@google.com, sarath.s...@huawei.com, nedn...@google.com
It is very important. When running pageload benchmark with telemetry, tsproxy sets the network configuration w.r.t kilo bits per second.

Log for REGULAR_4G: _Configs(4 * 1024 * 1024 / 8, 3 * 1024 * 1024 / 8, 20):
(INFO) 2018-06-25 12:14:58,941 ts_proxy_server._IssueCommand:93  Issuing command to ts_proxy_server: set rtt 20
(INFO) 2018-06-25 12:14:58,941 ts_proxy_server._IssueCommand:93  Issuing command to ts_proxy_server: set inkbps 524288 // ==> (524.288 mbps)
(INFO) 2018-06-25 12:14:58,941 ts_proxy_server._IssueCommand:93  Issuing command to ts_proxy_server: set outkbps 393216

So it is observed that pageload times depends on RTT only when using any configs from https://cs.chromium.org/chromium/src/third_party/catapult/telemetry/telemetry/page/traffic_setting.py?l=27 , because any increase in bandwidth after some value (observed  10 mbps), does not affect the pageload time, but only RTT affects.

We need to correct this in catapult.

BR,
Sarath

Ned

unread,
Jun 26, 2018, 7:08:30 AM6/26/18
to sarath.s...@huawei.com, pme...@google.com, Chrome benchmarking, ca...@google.com
+Patrick Meenan can you confirm tthat ts_proxy use bits per second?

Patrick Meenan

unread,
Jun 26, 2018, 8:37:18 AM6/26/18
to Ned Nguyen, sarath.s...@huawei.com, benchmar...@chromium.org, Andrey Kosyakov
Yes, it is in kilo bits per second: Documentation, Code.


Ned

unread,
Jun 26, 2018, 2:04:06 PM6/26/18
to Patrick Meenan, sarath.s...@huawei.com, benchmar...@chromium.org, Andrey Kosyakov
Thanks folks. I send out https://chromium-review.googlesource.com/c/catapult/+/1115363 to update the traffic settings

sarath.s...@huawei.com

unread,
Jun 27, 2018, 4:22:28 AM6/27/18
to Chrome benchmarking, pme...@google.com, sarath.s...@huawei.com, ca...@google.com, nedn...@google.com
I just noticed the patch. Here is the new bandwidth values in the patch, but we need double check this.

    GPRS: _Configs(50 * 1024, 20 * 1024, 500), ==> 50 mbps
    REGULAR_2G: _Configs(250 * 1024, 50 * 1024, 300),
    GOOD_2G: _Configs(450 * 1024, 150 * 1024, 150),
    REGULAR_3G: _Configs(750 * 1024, 250 * 1024, 100),
    GOOD_3G: _Configs(1.5 * 1024 * 1024, 750 * 1024, 40),
    REGULAR_4G: _Configs(4 * 1024 * 1024, 3 * 1024 * 1024, 20),
    DSL: _Configs(2 * 1024 * 1024, 1 * 1024 * 1024, 5),
    WIFI: _Configs(30 * 1024 * 1024, 15 * 1024 * 1024, 2), ==> 30 Giga bits per second

Instead, in all configurations, we need to remove the multiplication with 1024.
Correct values should be:
    GPRS: _Configs(50, 20, 500), ==> 50 kbps
    WIFI: _Configs(30 * 1024, 15 * 1024, 2), ==> 30 mbps


Eg:
{title: "WiFi", value: {throughput: 30 * 1024 * 1024 / 8, latency: 2}}

throughputInKbps = (30 * 1024 * 1024 / 8) / (1024 / 8)
That becomes (30 * 1024) Kbps.

BR,
Sarath

s.sin...@gmail.com

unread,
Jun 27, 2018, 5:09:22 AM6/27/18
to Chrome benchmarking, pme...@google.com, sarath.s...@huawei.com, ca...@google.com, nedn...@google.com
Corrected the traffic settings in https://chromium-review.googlesource.com/c/catapult/+/1116553

BR,
Sarath
Reply all
Reply to author
Forward
0 new messages