No real data, no any impact, no systematic test Post-doc

今天来了个德国不就是意大利的Post-doc来visiting作报告，据说在香港大学也呆过，Martin把他问的到后来一句都答不上来：

Martin: Did you do cpu/memory profiling? What black box test did you do?

P: What  black box do you mean?

Martin: When will your system break?

P:break, what do you mean?

Martin: yes, every system will break, in which circumstance for any reason?

P: maybe when the data set are too large, I don't know, they are all in memory, I don't have that large data. maybe I can test with big data.

Martin: What impact do your system have? with your spatial and text algorithm. you said you are going to follow the direct in next 3 years.

P: I struggle to get real data, but there is no such data, they are not public. I hope I can have real data to apply on my system in future.

Martin: you can have terabytes of data from twitter, it's easy.

P: but their data can not directly be used, 

Martin: so you mean you need clean data for you...Hannes, what do you think of this problem?（听口气已经忍无可忍了，难道你想别人把现成的clean data摆在你面前吗？）

P: oh, it's possible he is right because bla bla(a nice guy,没有落井下石hit a person when he is down, Romulo在一旁偷笑)

Martin:  do you try PostgreSQL since you use R-tree, they have long~~ history r-tree?

P: no,

Martin: why don't you use postgresql to test?

P:  er...I just chose to do stand-alone, I only think of use stand-alone.

Martin: did you read xxx's paper in 2002, he has the same problem, 

P: I don't read the paper...

听得我在旁边真是替他捏汗捉急啊，博士后，做了几年没有真正数据测试过的没有产生实际任何价值的所谓模型系统，对同行多年的成果没有任何对比研究，还打算未来一抹黑走到底，真可怕。

总结：

1. 要对系统进行深入性能分析。CPU，MEMORY，IO.

2. 测试系统的崩溃点即极限及黑盒测试。

3. 一定要做有实际价值的工作！

以此为鉴！