ParaScale Video-On-Demand Solutions

With the Web2.0 wave, video on demand (VOD) is fast gaining popularity. Video on demand requires the storage of numerous very large (10s to 100s of MB) files. With demand growing rapidly, most VOD service providers are struggling with storage issues related to massive capacity and throughput growth, even as VOD business models require a high-quality customer viewing experience.

The Challenge

Storing and accessing continuous media is a fundamental technology for all media distribution systems, including VOD and interactive TV. Unlike individual television studios that produce media, entertainment networks that distribute rich media for concurrent viewing by a large number of people, VOD allows service providers to offer an endless array of speciality content to consumers who are willing to pay to watch what they want, and when they want to.

To ensure profits, companies need to deliver video files seamlessly through the Internet or to cable channels, on demand, cost effectively and with optimum network bandwidth utilization. From the perspective of data center infrastructure, scalability is an issue as the read bandwidth is complicated by the wide distribution of read frequency. Where some videos are viewed hundreds of times per day, others are not viewed for hundreds of days. Things get worse as the videos in each camp change as they as they age and their promotion ebbs and flows.

The VOD repository also grows inexorably - as new releases are added to the current inventory. Along with infrastructure costs, planning for repository growth and managing read bandwidth are the primary data storage challenges. Video-on- demand can be highly bandwidth intensive eventually posing network bandwidth challenges. The inefficient provisioning or delivery of Video-on-Demand services can force unnecessary and expensive bandwidth and server upgrades.

The ParaScale Solution

ParaScale delivers an ideal solution for video-on-demand deployments.

First, ParaScale delivers massive read bandwidth (over 50-100 MBps with a common dual-core server with about 2G of RAM) on each storage node (SN) with any fraction of this bandwidth allocated to any single video. If a large number of people want to watch the same video, ParaScale replicates that video across all SNs so that network bandwidth of full ParaScale Cloud Storage (PCS) can be dedicated to that single video. In practice, some videos are widely replicated, a few others are replicated across only 20% of the SNs, and the vast majority is replicated only twice to protect against server or disk failure.

Second, ParaScale recognizes the importance of uninterrupted service for video-on-demand applications and delivers non-stop viewing service to its customers. Administrators can take individual servers off-line and conduct repairs or upgrades while the remaining servers continue operations. Also, to accommodate new videos or increased viewership, ParaScale deployment can grow incrementally without any interruption in service.

Third, ParaScale leads to reduced infrastructure overheads as it leverages commodity servers, switches, and disks efficiently to provide seamless user experience. Server densities can be high with the availability of cool-running hardware from several vendors. As ParaScale Storage Node load is low (simply delivering files to the network), each SN can support a large number of direct-attached disks.