Concourse
ConcourseDB is a self-tuning database that makes it easy to quickly build reliable and scalable systems. Concourse dynamically adapts to any application and offers features like automatic indexing, version control, and distributed ACID transactions within a big data platform that manages itself, reduces costs and frees developers to focus on core business problems.
This is version 0.5.0 of Concourse.
Quickstart
Let's assume we have the an array of JSON objects that describe NBA players.
from concourse.concourse import *
data = [
{
"name": "Lebron James",
"age": 30,
"team": "Cleveland Cavaliers"
},
{
"name": "Kevin Durant",
"age": 26,
"team": "OKC Thunder"
},
{
"name": "Kobe Bryant",
"age": 36,
"team": "LA Lakers"
}
]You can use Concourse to quickly insert the data and do some quick analysis. Notice that we don't have to declare a schema, create any structure or configure any indexes.
concourse = Concourse.connect()
records = concourse.insert(data) # each object is added to a distinct record
lebron = records.pop()
durant = records.pop()
kobe = records.pop()You can read and modify individual attributes without loading the entire record.
concourse.get("age", kobe)
concourse.add("name", "KD", durant)
concourse.remove("jersey_number", 23, lebron)You can easily find records that match a criteria and select the desired since everything is automatically indexed.
concourse.select(criteria="team = Chicago Bulls")
concourse.select(keys=["name", "team"], criteria="age bw 22 29")You can even query data from the past without doing any extra work.
concourse.get("age", durant, "04/2009")
concourse.find("team = Chicago Bulls at 2011")
concourse.select(criteria="age > 25 and team != Chicago Bulls", timestamp="two years ago")It is very easy to analyze how data has changed over time and revert to previous states.
# Analyze how data has changed over time and revert to previous states
concourse.audit("team", lebron)
concourse.revert("jersey_number", kobe, "two years ago")And ACID transactions are available for important, cross record changes.
concourse.stage()
try:
concourse.set("current_team", "OKC Thunder", lebron)
concourse.set("current_team", "Cleveland Cavs", durant)
concourse.commit()
except TransactionException:
concourse.abort()You can find more examples in the examples directory. More information is also available in the Concourse Guide and API documentation.
Motivation
Concourse was built in response to the fact that the existing landscape of storage technologies are simply unfit to meet the requirements of developers today.
Whether you use SQL or NoSQL, the existing data storage solutions are incredibly time intensive to manage. Administrators and developers are constantly forced to reevaluate decisions, profile the system, and perform manual optimizations. Even systems that claim to address this problem only do so by trading off critical features like ACID transactions. These requirements slow down development and introduce operational complexity and unnecessary scaling costs.
Planning
You have an awesome idea for a product? Great! But before you start building, you've got to figure out how to translate that idea into a rigid database structure. And if you want to change a feature down the line, you've got to migrate your existing data. This slows down your ability to prototype, iterate and pivot.
Profiling
Got performance problems? Well, now you have to spend time finding the culprit. Did you index the data correctly? Are your queries properly structured? Do you have an external cache in place? Meanwhile, your competitors are busy gaining traction.
Scaling
You have more users and more data. Now the database needs more resources. Sure, the popular NoSQL solutions make this easier, but eventual consistency forces you to figure out what to do when the database can't be trusted.
Optimizing
Your database is mission critical, but as you can see, it needs a babysitter. This forces you to allocate an abundance of resources just to figure out which settings to tweak each time your app grows. Time and money wasted.
The Solution
With Concourse, we have re-imagined the concept of a database from the ground up. We've created a solution that is flexible, easy to use, provides strong consistency, and optimizes itself in real-time.
Automatic Indexing
You no longer need to plan queries in advance because Concourse automatically indexes all of your data while guaranteeing constant time writes that are super fast. Concourse fully supports ad-hoc range and predicate queries and automatically caches frequently requested data for optimal performance.
Version Control
Concourse automatically tracks changes to your data, just like Git does for source code. Of course this means that you can easily audit changes and revert to previous states without downtime; but it also means that you have the power to query data from the past. Version control in Concourse makes it possible to build applications that know what was known when and can analyze real-time changes over time.
ACID Transactions
Concourse supports truly distributed ACID transactions without restriction. And we use dynamic resource management with just-in-time locking to ensure that they deliver both the highest performance and strongest consistency. So no need to guess when your data will eventually become consistent. When distributed Concourse responds to a query, you can trust the results immediately.
Simple Data Model
Concourse's document-graph structure is lightweight and flexible–it supports any kind of data and very large scales. Data about each person, place or thing is stored in a record, which is simply a collection of key/value pairs. And you can create links among records to easily model all the relationships within your data.
Schemaless
Since Concourse makes very few assumptions about data, it integrates with your application seamlessly and never needs a translator (goodbye object-relational impedance mismatch)! You never have to declare structure up front–no schema, tables, or indexes–or specify value types because Concourse is smart enough to figure it out. Concourse dynamically adapts to your application so that you can focus on building value without having to drag the database along.
Search
Concourse supports rich full text search right out the box, so you don't need to deploy an external search server. Data is automatically indexed and searchable in real-time without ever diminishing write performance. In Concourse, you can always perform as-you-type searches that match full or partial terms.
Overview
System Requirements
- At least 256 MB of available memory
- Linux or OS X
- Java 1.7+
Versioning
Concourse will be maintained under the Semantic Versioning
guidelines such that release versions will be formatted as <major>.<minor>.<patch>
where
- breaking backward compatibility bumps the major,
- new additions while maintaining backward compatibility bumps the minor, and
- bug fixes or miscellaneous changes bumps the patch.
Additional Resources
Contributing
Read the contributing guidelines to learn how to get involved in the community. We value and welcome constructive contributions from anyone regardless of skill level :)
Mailing Lists
Credits
Author
- Jeff Nelson
License
Copyright © 2013-2015 Cinchapi, Inc.
Concourse is released under the Apache License, Version 2.0. For more information see LICENSE, which is included with this package.