HackerTrans
TopNewTrendsCommentsPastAskShowJobs

Jamesbeyond

no profile record

Submissions

Tell the trend of databases from a new release

doris.apache.org
3 points·by Jamesbeyond·3 years ago·1 comments

[untitled]

1 points·by Jamesbeyond·3 years ago·0 comments

[untitled]

1 points·by Jamesbeyond·3 years ago·0 comments

[untitled]

1 points·by Jamesbeyond·3 years ago·0 comments

A strong case for transitioning from ClickHouse to Apache Doris

old.reddit.com
2 points·by Jamesbeyond·3 years ago·1 comments

[untitled]

1 points·by Jamesbeyond·3 years ago·0 comments

We Increase Database Query Concurrency by 20 Times

medium.com
2 points·by Jamesbeyond·3 years ago·2 comments

comments

Jamesbeyond
·3 years ago·discuss
To list the key functionalities of this new database, I think they represent many trends in the current data world:

Native support for semi-structured data. Elastic scaling of computation resources. Auto synchronization of data from MySQL and Oracle. Tiered storage of hot and cold data. Storage-compute separation. Kubernetes deployment. CCR. High concurrency. Inverted index. Auto fine-tuning. Data lakehousing capabilities. Parallel execution. Efficient data updates. Resource isolation for multi-tenancy.
Jamesbeyond
·3 years ago·discuss
Yep, Chinese characters are terrifyingly complicated, but the Chinese language is more compact sentence-wise and text-wise. Actually, the Chinese were acutely aware of how inconvenient the heavily-stroked characters are. That's why they invented simplified Chinese.
Jamesbeyond
·3 years ago·discuss
[dead]
Jamesbeyond
·3 years ago·discuss
[dead]
Jamesbeyond
·3 years ago·discuss
Quick computation, no duplication, and no data loss, realized by the idempotent writing of Apache Doris and the two-stage commit mechanism of Apache Flink.
Jamesbeyond
·3 years ago·discuss
It takes quite some fine-tuning to generate a photo that is not creepy (this is so important...) and worth buying.
Jamesbeyond
·3 years ago·discuss
There are two common log analytic solutions within the industry: inverted index (Elasticsearch) and lightweight index / no index (Grafana Loki). We opt for the first one since the other one is just trading query speed for high throughput and low storage costs, while query speed is the biggest concern for people. Now there is a way to improve on the first option, and achieve two times faster query speed with only one fifth of the storage space.
Jamesbeyond
·3 years ago·discuss
TLDR: The transition was mainly for partial update of columns, lower storage costs, and lower maintenance costs.
Jamesbeyond
·3 years ago·discuss
Hi, engineers. PMC (project management committee) member of Apache Doris here. (If you have never heard of Apache Doris, it is an analytic database.) We recently summarize our efforts in improving query concurrency in this writeup: https://dev.to/apachedoris/30000-qps-per-node-how-we-increas... We've been trying to handle both high-throughput and high-concurrency queries on one single platform, and we are kinda proud of what we have achieved, but we also want some input from HN, like do you think 30,000 per node is competitive enough, and most importantly, do you see any room for improvement in these 9 methods?
Jamesbeyond
·3 years ago·discuss
Row storage, short-circuit, prepared statement, and row cache.