Introduction to NuoDB – An Elastically Scalable Cloud Database


Traditional relational databases are built upon a synchronous, client-server architecture that is often limited in terms of scalability requirements that are posed by distributed computing systems. As a result, various sharding, caching, and replication techniques emerged to cope with these demands. On the other hand, NoSQL solutions have emerged on the ground of the CAP theorem. Data management systems like BigTable, HBase, MongoDB, Cassandra, and Dynamo offer different capabilities depending on how they balance consistency, availability, and partition tolerance. However, they gave up supporting SQL and ACID properties, which are critical in the relational database world.

NuoDB is a complete re-think of relational databases that is built on a new foundation; partial, on-demand replication. Under the hood, NuoDB is an asynchronous, decentralized, peer-to-peer database. It uses the concept of Atoms, these are objects that are being replicated . In NuoDB everything is an Atom; database, schema, sequence, table, index, records, blobs, data are all Atoms. NuoDB holds a patent on this peer-to-peer object replication.

NuoDB Architecture

NuoDB architecture has three layers: management layer, SQL layer and data layer. The management layer is comprised of an agent that manages the NuoDB processes running on a particular computer, it starts and stops them and it also collects statistics from the transaction and storage engines. Certain agents are configured to be a broker – brokers communicate with the client initially and then the broker introduces the client to the transaction engine. From then on the client can communicate directly with the transaction engines. NuoDB management layer also offers a command line and a web-based management tool to manage the databases. NuoDB also offer a command line loader for exporting and importing data.

At the  SQL layer NuoDB has transaction engines that provide access to a single database.The transaction engine parses, compiles, optimizes and executes the SQL statements on behalf of the clients.

At the data layer NuoDB has storage managers that provide persistence of the data. A storage manager uses key/value pairs to store the information but it can also use more sophisticated stores e.g. HDFS.

In case of a minimal configuration we can run every components (broker, transaction engine and storage manager) on the same machine. NuoDB can be easily scaled out and can be made redundant by adding multiple brokers, transaction engines and storage managers. In more complex scenarios we can run NuoDB in the AWS cloud or across multiple corporate datacenters providing geo-redundancy. Below is an example of a redundant architecture with two brokers, two transaction engines and two storage managers.NuoDBRedundantArchitecture

Getting Started

NuoDB is available on multiple platforms like Windows 32 and 64-bit, Linux 64-bit (Ubuntu, RHEL, Suse), Mac OSX 10.7, Solaris 11 (Inter 64-bit).  For this article we used a Ubuntu 12.04 LTS virtual machine.

First we need to start up the components as discussed above; the broker/agent, then from the command line management tool we can start up the transaction engine and the storage manager. We also need to configure the properties file to contain the settings for the domain.

$ vi ./etc/
# A flag specifying whether this agent should be run as a connection broker
broker = true

# The name used to identify the domain that this agent is a part of 
domain = stock

# The default administrative password, and the secret used by agents to
# setup and maintain the domain securely
domainPassword = stock
# Start agent
$ java -DpropertiesUrl=file:///home/notroot/nuodb/etc/ -jar jar/nuoagent.jar --verbose &>/tmp/stock.log &

# Start command line manager
$ java -jar jar/nuodbmanager.jar --broker localhost --password stock
nuodb [stock] > show domain summary

[broker] localhost/

## Create a new domain administrator user
nuodb [stock] > create domain administrator user istvan password istvan

## Start Storage Manager
nuodb [stock] > start process sm
Database: stock
Host: localhost
Process command-line options: --dba-user stock --dba-password stock
Archive directory: /home/notroot/nuodb/data
Initialize archive: true

Started: [SM] ubuntu/ [ pid = 3467 ] ACTIVE

## ps -ef | grep nuodb
## notroot   3467  3396  0 12:01 pts/0    00:00:00 /home/notroot/nuodb- --connect-key 7124934669079864995

## Start Transaction Engine
nuodb [stock/stock] > start process te
Host: localhost
Process command-line options: --dba-user stock --dba-password stock

Started: [TE] ubuntu/ [ pid = 3494 ] ACTIVE

## ps -ef| grep nuodb
## notroot   3494  3396  0 12:06 pts/0    00:00:00 /home/notroot/nuodb- --connect-key 8587006928753483386

Note, that we started the storage manager with initialize yes option. This is only for the first time, any subsequent startup shall use initialize no option, otherwise the data will be overwritten.

Then we can connect to the database using nuosql client – the first argument is the name of the database (stock), and we need to specify the database admin username/password. After login we can set the schema with use command to stock.:

$ bin/nuosql stock --user stock --password stock
SQL> use stock
SQL> show
	autocommit state is on
	semicolon completion is required
	current schema is STOCK
SQL> show tables

	No tables found in schema STOCK

SQL> create table Stock
   > (
   >    Id             Integer not NULL generated always as identity primary key,
   >    StockDate      Date,
   >    StockOpen      Decimal(8,2),
   >    StockHigh      Decimal(8,2),
   >    StockLow       Decimal(8,2),
   >    StockClose     Decimal(8,2),
   >    StockVolume    Integer,
   >    StockAdjClose  Decimal(8,2)
   > );
SQL> show tables

	Tables in schema STOCK


We can then load the data stored in csv file format into the database table. The CSV file – google.csv for stock information – was downloaded from

$ bin/nuoloader --schema stock --user stock --password stock --import "/home/notroot/nuodb/samples/stock/google.csv",skip --to "insert into Stock values(default,?,?,?,?,?,?,?)" stock &> /tmp/nuoloader.log

Imported 2163 rows, failed 0 rows, size 101897 bytes from /home/notroot/nuodb/sa

Then we can login again using nuosql and run a regular SQL query to retrieve the top 10 stock values and the corresponding date (ordered by adj close value):

notroot@ubuntu:~/nuodb$ bin/nuosql stock --user stock --password stockSQL> use stock
SQL> select count(*) from stock;

SQL> select StockDate, StockOpen,StockClose, StockVolume, StockAdjClose from stock order by StockAdjClose desc limit 10;

 ---------- ---------- ----------- ------------ -------------- 

 2013-03-05   828.93     838.60      4044100        838.60     
 2013-03-11   831.69     834.82      1594700        834.82     
 2013-03-07   834.06     832.60      2052700        832.60     
 2013-03-08   834.50     831.52      2911900        831.52     
 2013-03-06   841.03     831.38      2873000        831.38     
 2013-03-12   830.71     827.61      2008300        827.61     
 2013-03-13   827.90     825.31      1641300        825.31     
 2013-03-14   826.99     821.54      1651200        821.54     
 2013-03-04   805.30     821.50      2775600        821.50     
 2013-03-20   816.83     814.71      1463800        814.71

Java Client – JDBC for NuoDB

NuoDB supports various programming languages for client applications such as Java, .NET, PHP, Ruby and Node.js. In this section we demonstrate that NuoDB supports JDBC in the same way that it is available for traditional relational databases. The Java program needs to add nuodbjdbc.jar to its classpath.

Below is an example Java code ( to retrieve the highest stock value ever (ordered by adj close) and the related date:

$ cat
import java.sql.Connection;
import java.sql.DriverManager;
import java.sql.PreparedStatement;
import java.sql.ResultSet;
import java.sql.SQLException;
import java.sql.Statement;
import java.util.Properties;

public class StockDB {

    /** The driver class provided by NimbusDB. */
    public static final String DRIVER_CLASS =

    /** The base URL for connecting to a local database server. */
    public static final String DATABASE_URL =

    // the established connection to a local server
    private final Connection dbConnection;

     * Creates an instance of DB and connects to a local server,
     * as the given user, to work with the given named database
     * @param user the user name for the connection
     * @param password the password for the given user
     * @param dbName the name of the database at the server to use
    public StockDB(String user, String password, String dbName)
	throws SQLException
	Properties properties = new Properties();
	properties.put("user", user);
	properties.put("password", password);
	properties.put("schema", "stock");

	dbConnection =
	    DriverManager.getConnection(DATABASE_URL + dbName, properties);

    /** Closes the connection to the server. */
    public void close() throws SQLException {

     * Gets the name for the given id, or null if no name exists.
     * @param an identifier
     * @return the name associate with the identifier, or null
    public String getDateAndAdjClose() throws SQLException {
	Statement stmt = dbConnection.createStatement();
	ResultSet rs = stmt.
	    executeQuery("select stockdate, stockadjclose from stock order by stockadjclose desc limit 1");
	try {
	    if (
		return rs.getString(1) + ", " + rs.getString(2);
	    return null;
	} finally {

    /** Main-line for this example. */
    public static void main(String [] args) throws Exception {

	StockDB stockDB = new StockDB("stock", "stock", "stock");
	System.out.println("Date and AdjClose: "  + stockDB.getDateAndAdjClose());



Then we can run the Java program as follows:

notroot@ubuntu:~/nuodb/samples/java$ javac 
notroot@ubuntu:~/nuodb/samples/java$ java -classpath .:../../jar/nuodbjdbc.jar StockDB
Date and AdjClose: 2013-03-05, 838.60