 |
 |
 |
 |
Algorithmic Development on top of Marvell/Galileo Network Switch Core
The Customer
The customer provides integration of Hardware, DSP, and Software solutions to
vendors around the globe.
The Product
The product is a real-time, distributed multimedia surveillance system targeted
at homeland security and general Digital-CCTV (Closed-Circuit TV).
It allows multiple users to record, playback, and analyze in real-time
rich-media (video, audio, and events) in a secured and redundant manner.
Each box of the system performs several independent processes like video and
audio acquisition (from digital and analog sources), compression, digital
switching between sources, digital recording, synchronized network transmission
(Streaming), and VMD (Video Motion Detection) for unattended intrusion detection.
All these functions are performed on a highly integrated system comprising 11
DSP's, around an ETX Pentium-III host running GNU/Linux, in a 2Ux19 box.
The product supports 8 video and 8 audio input channels, and provides
MPEG4/H.263 video compression of Full-D1 (4CIF) resolution, at rates of up to
30FPS (Frames Per Second).
It uses RTSP and RTP for Networking Streaming, with rates of 128Kbps to 4Mbps
per channel. The maximum achievable network throughput per box is of 48Mbps with
extremely low latency per stream (50-250msec)
The Challenge
As the solution is targeted at providing perimeter security (sometimes even
spanning hundreds of kilometers), the ability to interconnect tens to hundreds
of boxes using Ethernet in a "ring topology", was imperative.
While several solutions (e.g. STP) exist to overcome the "closed-loop" issues
arising from this configuration, none provide good performance, as all are based
on "cutting the ring", which effectively converts it into a "Bus".
The "ring topology" was a requirement not only for reducing cabling costs, and
to provide alternative routes in case of damaged links (no single point of
failure), but also to be able to provide optimal exploitation of the network
bandwidth.
Ximpo Group's Embedded and Networking experts were hired to perform the
analysis, and to design a tangible solution.
Ximpo Group's experts designed an algorithm which combined user-space,
kernel-level, and hardware-control of the Marvell/Galileo network switch core
available on the board.
This algorithm not only fulfilled all the requirements, but also enabled
fail-over treatment an order-of-magnitude faster than with any other comparable
solutions.
The work concluded with knowledge transfer of the solution, and escorting the
integration of the solution into the customer's management system until
market-deployment.
The product is already in the market, exceeding sales expectations,
and highly acclaimed for some of its unique features.
More information about the product available at:
<<back
|
|
 |
 |
 |
|
|
