Skip navigation

Tag Archives: database

Mark Hurd: ‘I felt Oracle was in a position to do something nobody else could do’

Mark Hurd, former president of HP in Silicon Valley, is excited about his prospects at Oracle. Because of the software giant’s deep reach into the business ecology, they are in a position to leverage their virtual monopoly to create very useful and innovative technologies for all involved. It all depends. Some companies in similar positions get lazy and just let their momentum take over the market as long as they can… we shall see.

Mark Hurd: ‘I felt Oracle was in a position to do something nobody else could do’

By 7:00PM BST 26 May 2012

Mark Hurd is mounting his own Silicon Valley comeback.

In 2010, his friend and tennis partner, Oracle chief executive Larry Ellison, gave him a job and blasted the Hewlett Packard board that had ousted him amid sexual-harassment allegations for having made “the worst personnel decision since the idiots on the Apple board fired Steve Jobs”.

A Jobs-style messianic return to HP is out of the question and the comparison between the two men bears no further scrutiny, but sitting in Oracle’s preternaturally bland Moorgate offices, Hurd, 55, and now co-president of Oracle, knows he has at least landed on his feet. (read more)

The most insightful comment Mark Hurd had to make in his interview was his generational concept of customer service… “I’m old…” he begins, and relates that his generation is used to crappy customer service. Not getting a response or being stuck at a call center on hold for fifteen minutes is the status quo that nobody in his age group would even flinch at. Not so with his daughter’s generation: they want answers immediately, and will take their business and searches where they can get that response time. An allusion was made in this respect to mobile computing platforms and this demand for immediate information access from future, younger generations of consumers.

Open Source DataBase Company 10Gen Raises $42 Million Round

NoSQL Strikes again. Looks like this new, non-relational database technology has come onto the technology market strongly for its features and inherent advantages. It’s good to see some product diversification on the software market. It keeps everyone on the alert and makes for incentive-based efforts to innovate and improve what’s already out there in an effort to compete.

Open Source Database Company 10Gen Raises $42 Million Round

10gen, creator of the open source MongoDB database software, this morning said it has raised $42 million in new capital in a venture round led byNew Enterprise Associates. Also participating in the deal are existing investors Sequoia Capital, Flybridge Capital and Union Square Ventures. The company has now raised more than $73 million in venture capital since inception. (read more)

Not bad for a plucky “little” open source project…

Why Does NoSQL Matter?

NoSQL is an alternative to the standard Relational Database Management System (RDBMS). It apparently performs very fast and efficiently with querying large data sets (also referred to as “big data”); it apparently is not intended as a replacement for SQL databases, but poses a specialized solution for typical database/data store problems in the enterprise-wide scope of an IT organization.

This article can be found online at the following location:
http://www.databasejournal.com/news/mongodb-nosql.html

MongoDB – Why Does NoSQL Matter?
November 7, 2011

In recent years, the drumbeat of vendors proclaiming the ascendancy of NoSQL has become increasingly loud. One of the NoSQL vendors that is seeing business results from its NoSQL solution is 10gen, which is the lead commercial sponsor behind the open source MongoDB NoSQL database.

“We’re seeing the NoSQL space really taking off now and it’s being used in a significant way by a lot of people, including a lot of large enterprises,” Dwight Merriman, CEO and co-founder of 10gen, told InternetNews.com. “So big Internet companies like Craigslist and Shutterfly as well as big companies like SAP, Telefonica and LexisNexis are using it.”

Competition from the traditional database vendors also helps to validate the NoSQL space. At the OpenWorld Conference last month, Oracle announced its own NoSQL solution.

From a competitive perspective, Merriman sees MongoDB as the most popular NoSQL database in use today. His metrics for popular are somewhat indirect, though, and include MongoDB job postings and Google search popularity.

In terms of where NoSQL is fitting into enterprises, Merriman isn’t necessarily seeing NoSQL as a replacement for SQL databases but rather as a new tier of database technology usage. He noted that most Fortune 500 enterprises have an Oracle type relational database in their organizations already. The relational database is used for Online Transaction Processing (OLTP). Most big enterprises also have some form of data warehouse and a business reporting and intelligence database.

“Basically every large enterprise in the world has those two buckets for sure and what we’re seeing are enterprises adding a third bucket, which is a NoSQL basis,” Merriman said. “So on a forward basis, enterprises will have three classes of databases instead of two.”

When it comes to new Internet companies, Merriman expect that those companies will choose only one of three types of databases. The chosen database will likely be the one that maps closest to the problem the company is trying to solve.

Merriman noted that to use MongoDB, an organization must write new code.

“It doesn’t use SQL so if your old code assumes a relational data model so you have to write new code,” Merriman said.

For those that are looking to migrate to MongoDB, Merriman said enterprise developers need to re-write the code that talks to the database. He noted that there is not a lot of tooling for migration, since it’s not easy to automate.

“There are all the normal tools you’d expect in the database for import, export and monitoring of data,” Merriman said. “They’re just not specifically designed for migration.”

MongoDB 2.x

MongoDB 2.0 was recently released providing new concurrency features to the open source NoSQL database.

For the roadmap moving forward, Merriman said there is continuing work to further improve concurrency in the database. Additionally there is a new aggregation framework in the works that will make operations where users are aggregating and merging statistics easier and faster than the current model. MongoDB developers are also working on full-text search for a future release of MongoDB.

Sean Michael Kerner is a senior editor at InternetNews.com, the news service of Internet.com, the network for technology professionals.

Something tells me we’ll be seeing more of NoSQL as large and influential organizations such as Oracle and Google continue to utilize it as one of three different database solutions within their technical architecture. I suspect that most situations will not replace existing RDBMS structure, but instead initiate the database structure of newer projects and applications.

The Many Roles of Storage on an Oracle Database

There are three different types of storage used by Oracle databases.

  1. Data Files
  2. Control Files
  3. Redo Logs

Each are recommended to be redundant and also to have separate physical storage drives and controllers to maximize the likeliness of recovery of data in the event of a hardware failure.

Data Files

These are the physical representation of the data written from the database engine to the disk storage drive. The smallest unit of storage on a data file is a data block, which is usually 8 kb in size. This can be variably set, but the recommendation is to choose an Oracle data block to be a size that is a multiple of the host operating system’s block size.

Data files are represented on the database as a tablespace. A tablespace is a logical grouping of data files that store database information. In a tablespace, four major types of information is stored:

  • Data (information from tables)
  • Indexes
  • Rollback/Undo
  • Temporary

The acronym D.I.R.T makes this easy to remember!

Control Files

These contain information about the database name, when it was created, and the full path of the data files represented in the database. There are exactly three control files in a database and each is identical in content to one another for redundancy.

Redo Logs

A redo log contains one of two different types of commands that are issued against the database:

  1. DML: Data Manipulation Language, or commands such as INSERT, UPDATE and DELETE which change the data in database tables.
  2. DDL: Data Definition Language, or commands such as CREATE and ALTER. These commands create and change database objects within the database.

Redo logs are usually multiplexed across different physical drives and controllers for reduncancy. There are three different groups that are rotated as each group of storage gets filled.

When all three groups are filled, one of two different things happen depending on the mode set for the database.

Archive log mode forces the database engine to copy the contents of each redo log group to a separate location (archive log file) before the database overwrites the contents of the redo log group.

Non-Archive mode just skips the copy step and automatically overwrites the old redo log file group.

That’s it from a storage perspective, which is one of the three major components of an Oracle Database (the other two being: MEMORY and PROCESSES).

Oracle Storage Types

The three major roles of physical storage for an Oracle database

Oracle SGA Architecture

Stumbled across a great four part video series on You Tube with a lecture on Oracle database architecture, and specifically a discussion on the Oracle SGA. For anyone interested in understanding the internals and the pure definition of an Oracle database, this lecture series is worth the time to watch. If you follow his work on the board, you can create your own cool diagram of Oracle database architecture.

Oracle SGA Architecture Lecture

Oracle Server Architecture

High Level Diagram/Definition of an Oracle Server

Basic Oracle Concepts

It is useful to know the architectural details of the Oracle Relational Database Management System (RDBMS) such as its concurrency model, how it manages in its memory and disk, etc.

Database vs. Instance

One common point of confusion with Oracle databases is the distinction between databases and instances.

A database is a collection of data files found on disk or some sort of storage device.

An instance is a set of Oracle processes (also known as background processes) are what operate on the database files, such as storing and fetching data from the database.

Schemas

A collection of objects under one database user (owner). A database user can own one schema, and that same user can still have access to different, multiple schemas.

Two important schemas belong to the SYS and SYSTEM users. SYS schemas contain the data dictionary for the database. The data dictionary is a collection of read-only (query only) database tables and views containing metadata related to the objects in the database. Metadata includes definitions for all of a schema’s objects.

Incicdentally, the SYS user has the highest privileges in an Oracle database and is the equivallent to an “Administrator” or “root” account on the operating system.

Tablespaces

A schema objects that requires physical storage must belong to tablespace, which is a logical group of functionally related schema objects.

Data Blocks

A data block is the smallest unit of data used by ORacle during its I/O operations. Each operating system has a block size. However, Oracle works exclusively in units of Oracle data blocks… not operating system blocks.

Note that one data block may contain more than one row of a table.

Most discussions about performance measurement in Oracle almost always includes the concept of data blocks.

What Makes Oracle Different

Come to understand two concepts of Oracle: its locking mechanism and its multiversion read consistency model. These two characteristics set it apart from other RDBMS’s. Facts about Oracle’s locking behavior:

Oracle uses row-level locking. It makes sure only one transaction can write or modify a piece of data at any given time. Read operations (query or select) do not initiate locking behavior.

Overhead for locking is really low. The lock status is stored within the data block of the row, and not through some lock row status table.

Even if you modify every row of a very large table, Oracle will put a row lock on each row.

Oracle also keeps track of how to change back any blocks of data that are changed. This is stored in an undo (or rollback) segment. This undo segment is marked with a internal timestamp or SCN (system change number) so that if for example, you are querying data while it is being updated, Oracle will look through the undo segments to know what the data looked like at the time your query initiated. This is called multiversion read consistency.

Follow

Get every new post delivered to your Inbox.