Where to put your web server if your users are in both US and China?

1 min read

Our literature altert app Stork has many users in US and China. Our server is located in US West (silicon valley), and it’s been fast. But recently we got more and more complaints from China users that it’s slow to load our website and sometimes the connection even failed.

We need to find a solution. There are a few options, from easy to difficult:

  1. We can migrate the server to a location where both China and US users can access reasonably fast. Still single server and the same domain name.
  2. We can have one server in mainland China (say Beijing) serving China users only, and the US server to serve the rest of world. We use the same domain name but make it “smart” so it redirects users to the appropriate server.
  3. Similar to #2, but we use two different domain names, one for China and the other for the rest of the world.

Obviously option 1 is most attrative – if we can find such a location. To this end, we created 8 test servers in different locations and test the connection from US West and China (Beijing). Here is the result:

Latency to 8 servers

And the same data in a table (numbers are latency in seconds):

Server LocationFrom US WestFrom Beijing
US West0.839.0
US West4.332.3
US West0.134.5
China (Beijing)737.14.4
Local0.10.1
Hong Kong3.40.7
China (Beijing) 22.00.1
Sigapore1.32601.7
Tokyo0.71104.0

We can see that the connection from China to most other countries is very slow, and connection from US to most other countried are fast except to China. So internet wise, China and US are practically disconnected!

The latency between China and US (West) is dependent on time.

The latency between China and US (West) is dependent on time. In the plot above, you can see that the latency is actually fairly small during the night time in China (note that the unit of the time in the above plot is UTC). The variation of latency suggests that we can’t measure the connection speed just at one time point.

But there is one exception which gives us hope. It’s Hong Kong! Its connection to Beijing is fast, and to US is reasonable. So it’s a good choice to choose Hong Kong. Actually we did this and the result is pretty good. Both our China users and users from other regions can access our website reliably and reasonably fast.

So if you have a significant number of users or customers in China, you may consider to host your web server in Hong Kong.


Receive email notification via email
Don't want to miss new papers in your field? Check out Stork we developed:

文献分析可以发SCI文章吗?

该文章可以通过该链接完整阅读(包括图片。这些图片有时微信禁止转载)。 我们印象当中,发文章需要做大量的实验和分析。实验对象可以是原子、病毒,可以是疾病、人群,也可以是恒星、宇宙。可是,如果分析的对象是文献本身,这样的研究有吗?能发文章吗? 小编虽然在科研领域20年,但是对这个问题还没有明确答案。于是这两天埋头钻研,结果脑洞大开,发现了一个原来从没听说过的领域:文献计量学,英文叫 Bibliometrics。借用百度百科的话,文献计量学是指用数学和统计学的方法,定量地分析一切知识载体的交叉科学。而这个领域竟然已存在100余年,不得不承认自己实在是孤陋寡闻。 于是,我们用“文献分析”的方法分析了“文献分析”类的文献(咦,怎么有点拗口?),看看有哪些意想不到的发现。我们用的关键词非常简单,就是 bibliometrics。数据来源是PubMed,分析工具是文献鸟的大分析。 01 — 上万篇论文用了“文献计量学” 首先,不仅有论文用了文献计量学的方法,而且有很多。在PubMed数据库里有一万余篇这类的论文。如果看趋势,我们发现在2005-2010增长速度较快,这几年稳定到每年750余篇。 02 — “文献计量学”论文可以发表在好期刊 那么,这上万篇的论文都发表在什么期刊?是灌水期刊吗?从上图的分布图来看,文献计量学的论文在各类期刊都有。在前25个期刊中,有Nature(254篇)、Science(74篇)、Lancet(41篇)、JAMA(32篇)这类顶尖文章(黄色标记),也有J Clin Epidemiol等这样的不错的期刊(绿色),当然也有影响力更低的期刊。 这么多Nature、Science等文章,自然引起了小编的怀疑和警惕。难道我通过分析文献就可以发表CNS文章?是不是太简单了?于是,小编点击进去,看看有哪些文章发表在这些顶级期刊。 果真不出所料,绝大部分在顶尖期刊的论文是评论、新闻、通信类型的。但是,通过仔细排查,我们还是找到了正规的、研究类的文章。下面举几个例子。 Lancet. 2019 Feb 9;393(10171):550-559. doi: 10.1016/S0140-6736(18)32995-7....
Xu Cui
14 sec read

Remote fNIRS

Everytime I met Dr Leanne Hirshfield, I am impressed by her energy and passion with the fNIRS technology. It was a pleasure listening to...
Xu Cui
4 min read

Google Chrome takes too much CPU

I have been noticing for a while that Chrome takes too much of my CPU. It constantly consumes about ~30% or even more CPU...
Xu Cui
32 sec read

Leave a Reply

Your email address will not be published. Required fields are marked *