Time-series databases with horizontal scaling for metric monitoring?

Question

I'm researching about time-series databases for metrics monitoring that scale horizontally easily.

I read about Prometheus comparison with other databases (as of 2014). It seems that uses less space that the rest of the options, but it does not scale horizontally. This is relevant also, but is from 2014 and I would like to hear about new databases also.

Could anyone compare time-series database alternatives?

I'm aware of this question but is from 2014 and new alternatives should be taken into account in this question

score 3 · Answer 1 · answered Feb 20 '16 at 00:20

InfluxDB is a single Go binary with no external dependencies that supports clustering for horizontal scaling of storage and throughput. A single node c4.8xlarge machine can process over 500k points per second, and with appropriate schemas it is possible to persist over 1 million values per second on a much smaller box (c4.2xlarge). Customers are managing over 500GB in a single node, although recent storage engine improvements have driven the size on disk down to ~10% of prior requirements. InfluxDB only requires 2-3 bytes per numeric metric recorded.

In addition, InfluxDB is a member of the TICK stack, a unified time-series platform for the collection, storage, management, altering, and visualization of time-series metrics for the IoT, devops, and real-time monitoring use cases. Each member of the TICK stack supports multiple protocols and data sources/sinks, but when combined together the ease of use and simplicity of operation are very compelling.

I run support for InfluxData, the company behind the OSS projects.

I read in the Prometheus link that I posted before that InfluxDB as of 2014 used to use a lot of memory; it's nice to see that that was fixed. — Martin Aparicio Pons, Feb 20 '16 at 14:13

score 1 · Answer 2 · answered Feb 17 '16 at 13:22

If you're looking at a solution to store metrics with high scalability, it may worth looking at cyanite = a plugin for graphite to store data on Cassandra: https://www.youtube.com/watch?v=htp-HiT5I1Q,

Github repo: https://github.com/brutasse/graphite-cyanite

score 1 · Answer 3 · edited Jun 20 '18 at 16:31

1

Axibase Time Series Database can be scaled up by adding storage nodes to the cluster without the need to shard and index data manually.

ATSD stores data quite efficiently too: 4+ bytes per time:value tuple, depending on regularity and value data type/variance.

Disclosure: I work for company developing ATSD.

edited Jun 20 '18 at 16:31

kghamilton

23
7

answered Feb 17 '16 at 15:08

Sergei Rodionov

4,079
6
27
44

Time-series databases with horizontal scaling for metric monitoring?

3 Answers3