Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
cdavid
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
14 ms
·
121.
▲
by
cdavid
4y ago
I agree certain orders of magnitude are harder to guess accurately. My claim for 1k qps being conservative is mostly based on * my current company, but I can't share more precise numbers. So I guess not very convincing :) * however, at
122.
▲
by
cdavid
4y ago
you're right about TB vs PB ofc :) But then keep in mind the assumptions were super conservative: * 1qps is likely off by at least half order magnitude * 1 byte per even is obviously off by several order of magnitudes, let's say j
123.
▲
by
cdavid
4y ago
> Am I doing the math wrong? Your mistake is to assume you need to look at per stay. You should look at per query if you want to improve your search. My math is based on the following: - 1000 qps - 1 query has on average 100 results -
124.
▲
by
cdavid
4y ago
Even using your numbers, if you want to keep say only 3 months of data, we're talking about 1-2 PB already. Being able to query those data across different segments, aggregate into different dimensions is already quite far beyond what
125.
▲
by
cdavid
4y ago
Search interactions data is one of the most valuable data in marketplaces. I never worked at Airbnb, but I worked at companies smaller than airbnb where improving ranking had many millions $ / year impact on revenue. > And even then
126.
▲
by
cdavid
4y ago
They most likely do. Think about e.g. ranking in their search/recommendation engine. To be able to train ranking ML models, you would need to at least track the views, clicks, purchases, etc. done through their platform. For each searc
127.
▲
by
cdavid
4y ago
Most people who get married the "western style", do it w/ fake priest, etc. It has no recognized religious value. Other people do it the "shinto way" or the "buddhist way".
128.
▲
by
cdavid
4y ago
wedding as well. I don't have numbers, but most people who got married in Japan I know, including myself, did it.
129.
▲
by
cdavid
5y ago
relative tolerance is not always enough, especially near 0. This classic article gives a nice overview of absolute, relative, ulp-based and other methods for comparison of float numbers: https://randomascii.wordpress.com/201
130.
▲
by
cdavid
5y ago
This is not a very good article, and I would not recommend it beyond simple use cases. The problem is that there is no right way, it depends on the usecase and the magnitude of numbers you're comparing. See e.g. https://bitb
131.
▲
by
cdavid
5y ago
They also radically transformed the landscape around semiconductors in the 60ies/70ies. E.g. a big section of the famed "high output management" from Andy Grove is about how Intel had to reinvent itself under the pressure of
132.
▲
by
cdavid
5y ago
Depends on the Kanjis. I used to work for a company doing Japanese handwriting recognition, and we would sometimes get request "can you support kanjis not in unicode" and "lost" contracts because we could not. This is es
133.
▲
by
cdavid
5y ago
Indeed, this is what you should do. Proceed a bit carefully if you don't have a good relationship yet with your manager: by good relationship, I mean knowing them enough to figure out how much you trust them. You would also benefit fro
134.
▲
by
cdavid
5y ago
As a manager of managers, I can confirm this is a fairly accurate of my job, in some aspects at least. The lack of concrete results is maybe the most frustrating as a former engineer (it is like your compilation cycles takes months if not q
135.
▲
by
cdavid
5y ago
Analog stuff (tb 303, sh101, etc.) in early acid house was used because it was cheap at that time, mostly. M1 came later than early acid house or techno after all Same reason FM was actually used a lot in electronic music as well. You can h
136.
▲
by
cdavid
5y ago
> The biggest problem is that managers can easily fall into a trap where they have everyone in the same place so they can ask individuals questions that only really matter to them. That can easily happen. On the other hand, where would y
137.
▲
by
cdavid
5y ago
the notion of ownership in stock ownership is tenuous at best though: https://www.johnkay.com/2015/11/11/is-it-meaningful-to-talk-... .
138.
▲
by
cdavid
5y ago
I think there is another powerful force that explains why so many managers are perceived as bad: in most companies, career growth for managers means moving up, and moving up is at best uncorrelated with managing your team well and with resp
139.
▲
by
cdavid
5y ago
Fourier transform, more exactly DFT, is a special case of matrix multiplication, when seen as a linear operator. The fact that you can do operations faster than the usual N*2 matrix / vector operation is due to the specific form of the
140.
▲
by
cdavid
5y ago
One of the main reason there is one party in power is because of the American occupation. While the American wanted to get rid of the the most "rightist" factions of the elite for obvious reasons, they quickly change to fight comm
141.
▲
by
cdavid
5y ago
The health system is fairly inhuman though, especially if you can't speak Japanese fluently
142.
▲
by
cdavid
5y ago
[disclaimer: while I have some small experience putting things in DC, including big GPU servers, I have never been anywhere near that scale, certainly not storage] 10k $ is for a server with no hard drive. W/ 12 Gb disks, and with enou
143.
▲
by
cdavid
5y ago
mathematics have historically been a focus of scientific training in France for a very long time. I don't think it is representative of a more general trend TBH.
144.
▲
by
cdavid
5y ago
For engineers who become managers, manager tools is my first recommendation as well. It can be cringy from time to time, but it really focuses on things we are not used to think much as engineers. And it gives a lot of practical tips one ca
145.
▲
by
cdavid
6y ago
I guess we can argue on what he meant to convey by this sentence, but would you agree this is not especially nuanced ? That one example was particularly bad, but I find it representative of the man. I fail to see the nuance or the depth.
146.
▲
by
cdavid
6y ago
I guess as a disclaimer I should say I am more of a progressive, and not from an American tradition. But in the same generic "conservative / learning libertarian" tradition as Sowell, I vastly prefer people like Tyler Cowen a
147.
▲
by
cdavid
6y ago
Yeah, I guess if your baseline is political commentator, almost anything would be an improvement. But I fail to understand how supposedly nuanced he is. To me, he looks like a caricature of the American libertarian/conservative flavor.
148.
▲
by
cdavid
6y ago
I don't get what make him popular around certain crowds in HN. He is very polemical (I mean he was arguing that Obama would lead the US to live under Sharia law), is not a particularly good scholar, nor very insightful. Certainly, noth
149.
▲
by
cdavid
6y ago
Just looking at your second link, the odds to hospitalized w/ a severe obesity (> 35 BMI) is 2.5 higher than people w/ BMI in healthy 20-25 range Now, according to https://www.cdc.gov/coronavirus/2019-ncov&
150.
▲
by
cdavid
6y ago
Obesity does not really explain order of magnitude fewer cases, and obesity is much less of a factor than age for mortality. I can't find the numbers now, but it is better to be younger and quite fat than much older and fit. Japan does
More ›