SAM SIG: Gracenote:
A Case Study in Massively Scalable B2B2C System Architecture
Gracenote: A Case Study in Massively Scalable B2B2C System Architecture
Gracenote is the home of CDDB. When you insert a CD into your player, there is nothing on the CD indicating its disc name or track titles or artists. There's only the length of each track to identify it. But it turns out that's a virtual fingerprint -- almost unique -- and Gracenote has created a database and recognition technology that allows them to automatically identify virtually any CD ever made. Their technology is used in almost every CD and MP3 player application, including Apple's iTunes, as well as in many consumer electronic devices. Since the initial debut of their CD recognition service in 1996, then known as CDDB, Gracenote has expanded its technology into new forms of media recognition, including music search engines and audio file recognition.
Gracenote has built a system that can scale easily to serve a massive number of users (essentially without practical limit), though their current design doesn't allow for their dataset to scale without restriction.
How did they do that? One of their simplifying assumptions is that their dataset, consisting mostly of information of limited dimensions (there are only so many albums and movies in the world), will fit comfortably into the space afforded by each individual server of a given type (there are many types, however). Data is cleverly segmented to fit in this way, rather than arbitrarily divided as in a cluster. While they're a B2B company, what they do is mostly direct end user interaction -- most of their traffic is through partners, but initiated by end users. Thus B2B2C.
About the Presenter
Steve Scherf, Co-Founder and VP Architecture/Service, Gracenote
Steve Scherf is a Bay Area native with 23 years of Unix experience. He graduated from UC Santa Cruz in 1988 with a Bachelor's degree in Math and Biology. Not seeing a future in either discipline, he spent his last year in college honing his programming skills in order to move into software development.
Since that time he has worked for companies such as Altos Computer Systems, Acer America and Stratus Computers as a Unix kernel developer, delving into such areas as filesystems, networking, memory management, I/O subsystems, OS-level RAID, SCSI devices, embedded software, and, most notably, fault tolerant systems. He holds several issued and pending patents related to I/O subsystems and music recognition.
In 1996, along with partner Ti Kan, Steve developed the seminal Internet CD recognition service known as CDDB. This came at the cusp of the Internet boom and the dawn of the digital music revolution, and CDDB flourished. By 1998 the CDDB service was in high demand, and the partners founded CDDB Inc. to allow it to further grow and evolve. The company was subsequently purchased by Escient Inc. and eventually spun off to become Gracenote.
As employee #1 at Gracenote, Steve now serves as co-founder and chief architect for the company. In this capacity, he oversees the development of all of Gracenote's new technologies, especially those involving client-server interaction. In addition, he personally manages the small team of systems engineers that develop and maintain all Gracenote services. Above all, he still keeps his hands dirty, having architected Gracenote's current service and developed the lion's share of the code himself.
Cubberley Community Center
4000 Middlefield Road, Room H-1
Palo Alto, CA
6:30 - 7:00 p.m. Registration/Networking/Refreshments/Pizza
7:00 - 9:00 p.m. Presentations
$15 at the door for non-SDForum members
No charge for SDForum members
No registration required
--> More on the Software Architecture & Modeling....