Originally tags were implemented just as a label, not a key/value pair. I saw that a
number of other time series dbs use key/value pairs and decided to make the change. It can
also produce more efficient queries. For example, filtering against N labels would require
querying N partitions (in the original schema). Querying against one tag name with N
different values however would only require querying 1 partition.
On Jun 24, 2015, at 6:25 AM, Michael Burman
<miburman(a)redhat.com> wrote:
Hi,
I don't agree with changing the terms. Many of metrics competing products (such as
InfluxDB / OpenTSDB) use the same terminology, so it is easier for customers to understand
the differences and similarities between products. Using different terminology was one of
the difficulties in RHQ (like noted in the F2F meeting) and I don't think we should
repeat the same mistake.
- Micke
----- Original Message -----
From: "Gary Brown" <gbrown(a)redhat.com>
To: "Discussions around Hawkular development"
<hawkular-dev(a)lists.jboss.org>
Sent: Wednesday, June 24, 2015 11:45:56 AM
Subject: [Hawkular-dev] Tag concept in metrics and alerts
Hi
Both metrics and alerts use the term 'tag' - however metrics use a name/value
pair for a tag, and alerts uses it as a label. Wondering whether we need consistency?
When I hear the term tag, I usually envisage a label - e.g. as in git. So wondering
whether it would be clearer for metrics to rename tags to properties?
Possibly metrics needs the concept of a tag/label as well?
Regards
Gary
_______________________________________________
hawkular-dev mailing list
hawkular-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/hawkular-dev
_______________________________________________
hawkular-dev mailing list
hawkular-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/hawkular-dev