Mysql compatible federqred query engine to Elasticsearch, Mongo,
Google Datastore, Cassandra, File-based backend sources, including join.
This proxy and query engine hosts a mysql protocol listener,
which then rewrites sql queries to native (elasticsearch, mongo, cassandra).
It works by implementing a full relational algebra layer
to run sql queries and poly-fill missing features
from underlying sources. So, a backend key-value storage such as cassandra
can now have complete WHERE clause support as well as aggregate functions etc.
Most similar to prestodb but in Golang, and focused on easy to add custom data sources.
- Distributed run queries across multiple servers
- Hackable Sources Very easy to add a new Source for your custom data, files, json, csv, storage.
- Hackable Functions Add custom go functions to extend the sql language.
- Joins Get join functionality between heterogeneous sources.
- Frontends currently only MySql protocol is supported but RethinkDB (for real-time api) is planned, and are pluggable.
- Backends Elasticsearch, Google-Datastore, Mongo currently implemented. Csv, Json files, and custom formats (protobuf) are in progress.
- NOT Production ready. Currently supporting a few non-critical use-cases (ad-hoc queries, support tool) in production.
This example imports a couple hours worth of historical data from https://www.githubarchive.org/ into a local elasticsearch server for example. Requires nats.io and etcd server running local as well. Docker setup coming soon.
cd tools/importgithub
# assuming elasticsearch on localhost elase --host=myeshost
go build && ./importgithub
# using dataux.conf from root of this project
go build
./dataux --config=dataux.conf
# now that dataux is running use mysql-client to connect
mysql -h 127.0.0.1 -P 4000now run some queries
use datauxtest;
show tables;
describe github_watch;
select cardinality(`actor`) AS users_who_watched, min(`repository.id`) as oldest_repo from github_watch;
SELECT actor, `repository.name`, `repository.stargazers_count`, `repository.language`
FROM github_watch where `repository.language` = "Go";
select actor, repository.name from github_watch where repository.stargazers_count BETWEEN "1000" AND 1100;
SELECT actor, repository.organization AS org
FROM github_watch
WHERE repository.created_at BETWEEN "2008-10-21T17:20:37Z" AND "2008-10-21T19:20:37Z";
select actor, repository.name from github_watch where repository.name IN ("node", "docker","d3","myicons", "bootstrap") limit 100;
select cardinality(`actor`) AS users_who_watched, count(*) as ct, min(`repository.id`) as oldest_repo
FROM github_watch
WHERE repository.description LIKE "database";
# to add other data sources see dataux.conf example
- Sources
- Cassandra
- Big-Query
- Big-Table
- Json-Files
- Writes
- write propogation: inbound insert/update gets written multiple places.
- write lambda functions: allow arbitray functions to get nats.io pub/sub of write events.
| Mongo | SQL Query |
|---|---|
show collections |
show tables; |
| na, -- runtime inspection | describe mytable; |
db.accounts.find({},{created:{"$gte":"1/1/2016"}}).count(); |
select count(*) from accounts WHERE created > "1/1/2016"; |
db.article.find({"year":{"$in": [2013,2014,2015] }},{}} |
select * from article WHERE year IN (2015,2014,2013); |
db.article.find({"created":{"$gte": new Date('Aug 01, 2011'), "$lte": new Date('Aug 03, 2013') },{title:1,count:1,author:1}} |
`SELECT title, count, author FROM article WHERE created BETWEEN todate("2011-08-01") AND todate("2013-08-03") |
db.article.find({"title":{"Pattern":"^list","Options":"i"}},{title:1,count:1}) |
SELECT title, count AS ct FROM article WHERE title like \"list%\" |
| na, not avail in mongo (polyfill in dataux) | SELECT avg(CHAR_LENGTH(CAST(title, \"AS\", \"CHAR\"))) AS title_avg FROM article; |
| need to document ... | select min(year), max(year), avg(year), sum(year) from table WHERE exists(a); |
| ES API | SQL Query |
|---|---|
| Aliases | show tables; |
| Mapping | describe mytable; |
| hits.total for filter | select count(*) from table WHERE exists(a); |
| aggs min, max, avg, sum | select min(year), max(year), avg(year), sum(year) from table WHERE exists(a); |
| filter: terms | select * from table WHERE year IN (2015,2014,2013); |
| filter: gte, range | select * from table WHERE year BETWEEN 2012 AND 2014 |
Hacking
- see tools/importgithub for tool to import 2 days of github data for examples above.
# to run queries below, run test data import
go get -u github.com/dataux/dataux
cd $GOPATH/src/github.com/dataux/dataux/tools/importgithub
go build
./importgithub ## will import ~200k plus docs from Github archive
- Data-Accessability Making it easier to query, access, share, and use data. Protocol shifting (for accessibility). Sharing/Replication between db types.
- Scalability/Sharding Implement sharding, connection sharing
| Name | Scaling | Ease Of Access (sql, etc) | Comments |
|---|---|---|---|
| Vitess | Y | for scaling (sharding), very mature | |
| twemproxy | Y | for scaling memcache | |
| Couchbase N1QL | Y | Y | sql interface to couchbase k/v (and full-text-index) |
| prestodb | Y | query front end to multiple backends, distributed | |
| cratedb | Y | Y | all-in-one db, not a proxy, sql to es |
| codis | Y | for scaling redis | |
| MariaDB MaxScale | Y | for scaling mysql/mariadb (sharding) mature | |
| Netflix Dynomite | Y | not really sql, just multi-store k/v | |
| redishappy | Y | for scaling redis, haproxy | |
| mixer | Y | simple mysql sharding |
We use more and more databases, flatfiles, message queues, etc. For db's the primary reader/writer is fine but secondary readers such as investigating ad-hoc issues means we might be accessing and learning many different query languages.
Credit to mixer, derived mysql connection pieces from it (which was forked from vitess).
In Internet architectures, data systems are typically categorized into source-of-truth systems that serve as primary stores for the user-generated writes, and derived data stores or indexes which serve reads and other complex queries. The data in these secondary stores is often derived from the primary data through custom transformations, sometimes involving complex processing driven by business logic. Similarly data in caching tiers is derived from reads against the primary data store, but needs to get invalidated or refreshed when the primary data gets mutated. A fundamental requirement emerging from these kinds of data architectures is the need to reliably capture, flow and process primary data changes.
from Databus