homin

4 ms·

1.
by homin 7y ago
In theory yes, but none of our implementations specifically prevent reallocations as we were not targeting this use-case. It should be easy enough to add though.
2.
by homin 7y ago
HDR is great for the use-case when you can bound your range beforehand and merging is not a requirement, but those were also the reasons we needed to develop DDSketch.
3.
by homin 7y ago
Yes, the sketch can quickly answer questions like "How many requests were faster than 100ms?" 100ms maps to a bucket index, and we would just sum up all the buckets with indices at most that.
4.
by homin 7y ago
T-digest is definitely the best rank-accuracy sketch for higher percentiles in terms of size. However, it was much slower than GK, and we found that by increasing the rank-accuracy of GK (taking the penalty of a larger sketch-size), the res
5.
by homin 7y ago
Most previous sketches used what they called "rank accuracy". So if your inputs were [2^1, 2^2, ...., 2^1000]. The actual p95 is 2^950, and a 0.01 rank-accurate sketch would be allowed to give you any value between 2^940 to 2^960
6.
by homin 7y ago
Author here. We wanted to be able to graph p99, p99.9 metrics with arbitrary ranges, and found the existing solutions were not accurate enough for our needs. Happy to answer any questions. Code here: https://github.com/DataD
7.
Beyond Kolmogorov-Smirnov(datadoghq.com)
16 points by homin 9y ago|0 comments
8.
Piecewise regression without the pain (open-source code)(datadoghq.com)
22 points by homin 9y ago|0 comments
9.
by homin 9y ago
He's famously opinionated: http://www.wisdom.weizmann.ac.il/~oded/on-awards.html