PagesPeed Insights-总阻塞时间
嘿,所以我看到了与PagesPeed Insights的狂野矛盾。 我对每个核心网络至关重要的特定都了解很多。 但是,也许我错过了一些东西。
9/10我得到95+ ...但是有时由于TBT(有时是LCP),我的分数只会储备。
这是我过去的得分14次。 (这是移动)
99、99、99、100、97、99、81(650ms)! ,100
我正在使用React。
每当发生这种情况时,都会有一个灰色的建议(避免长长的主线任务): (用示例
替换网站名称)
…js/main.39443e3e.js(example.com)
1,530 ms
322 ms
…js/main.39443e3e.js(example.com)
1,852 ms
288 ms
https://example.com
1,028 ms
207 ms
…js/main.39443e3e.js(example.com)
// 2,140 ms
204 ms
https://example.com
780 ms
72 ms
…js/main.39443e3e.js(example.com)
2,344 ms
53 ms
https://example.com
977 ms
51 ms
https://example.com
是否有人有任何想法?它只是片状的洞察力吗?我需要发送链接,以便客户可以自己看到分数...但是我不喜欢1/10似乎没有理由发送分数不好。
Hey so I'm seeing wild inconsistencies with PageSpeed insights.
I know quite a bit about the specific on each core web vital.
However, maybe I'm missing something.
9/10 I'm getting 95+ ... but sometimes my score will just TANK due to TBT (and sometimes LCP).
Here's my past scores ran 14 times. (this is mobile)
99, 99, 99, 100, 97, 99, 81(650ms)!!!!!!!!!, 99, 99, 98, 100, 98, 99, 94(240ms tbt)!!, 100
I'm using react.
Whenever this happens there's a greyed suggestion (Avoid long main-thread tasks):
(Replaced site name with example
for privacy)
…js/main.39443e3e.js(example.com)
1,530 ms
322 ms
…js/main.39443e3e.js(example.com)
1,852 ms
288 ms
https://example.com
1,028 ms
207 ms
…js/main.39443e3e.js(example.com)
// 2,140 ms
204 ms
https://example.com
780 ms
72 ms
…js/main.39443e3e.js(example.com)
2,344 ms
53 ms
https://example.com
977 ms
51 ms
https://example.com
Does anyone have any ideas what's going on here? Is it just PageSpeed Insights being flaky? I need to send a link so clients can see the score for themselves... but I don't like 1/10 sending a bad score for seemingly no reason.
如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。

绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱,如果其他用户或者作者回复了您的评论,将不能在第一时间通知您!
发布评论
评论(1)
对于性能测试,机器和网络中发生的任何内容都可以为您的测量添加噪音。
为了进行性能测试,我使用了一些套件。我有一个在AWS上,在网络上隔离,盒子上没有其他运行。如果我使用共享VM,我会得到钉。它必须是一个专用的VM,其RAM过多,并且所有10次运行都将在彼此的5%以内。
For performance testing, whatever is going on in the machine and network can add noise to your measurement.
For performance testing, I have a few suites that I use. I have one that is on AWS, isolated on the network with nothing else running on the box. If I use a shared VM, I get spikes. It has to be a dedicated VM with an excessive amount of ram, and all 10 runs will be within 5% of each other.