The popularity of the “CSI” series of television shows is creating havoc for surveillance directors. When security directors see “CSI” actors instantly retrieve year-old video evidence, zoom in on far-away objects and see video from many different angles, they question why they can’t have the same capabilities they see on TV in the systems they use in real life. What they are finding out is that while some of the show’s capabilities are futuristic, many of the capabilities could be realized today if traditional security systems were re-engineered to take advantage of new camera, software and storage technologies.
Three rapidly evolving components exist in today’s digital surveillance systems that are helping to align possibilities with expectations: high-resolution megapixel cameras, video management software, and high-definition video storage. Integrators and end users alike appreciate the value of a high-quality network camera that captures images in varying conditions, such as indoors / outdoors, or high / low lighting. These same people take a lot of care in selecting the right video management software that provides monitoring, recording and event management functions.
As for the storage component, many system designers get well down the road on large-scale surveillance systems before they realize that “CSI”-type video expectations can lead to huge storage capacity and performance requirements. These needs can overwhelm project costs if careful attention is not paid up front to appropriate storage system design. Here are some pointers on how to think about design considerations as you plan large surveillance installations. User Expectations
Tip #1 is to accept that users expect to have “CSI”-type video capabilities. The genie is out of the bottle; people want what they see on TV. This points to a clear need for digital systems over analog. Users want to search their video quickly to find the needle in the haystack, and this simply can’t be done with a mountain of tapes. The digital video management systems available today are very advanced in their capabilities for cataloging the incoming video from hundreds of cameras, searching recorded images according to specified criteria, and even preserving the chain of custody so that the video is admissible as evidence in court.
Security experts have developed a strong preference for megapixel IP cameras because they provide better camera coverage and better picture detail. In addition, digital pan / tilt / zoom allows the user to move around a large image using a smaller viewing window. So, if we think of that “CSI” murder scene, the investigators can zoom in on multiple areas of the images—the tire tread marks in the dirt, the pattern of the blood splatters, the license plate on the car down the street, and so on. You don’t have to decide in advance what to zoom in on for recording.
Once the images are captured, there seems to be a boundless desire for endless retention periods. As a result, massively scalable storage systems are now a de facto requirement to support such long retention periods. Sometimes in real life, retention requirements are driven by law. For example, a juvenile penitentiary is required to store video for a number of years, sometimes spanning the full length of time a person is incarcerated. To ensure that video is available for the long term, care must be taken in how it is stored.
Tip #2 is recognize that video datastreams stress all elements of a system. As a data type, video has distinctly different properties than, say, elements in a database or document. Packets of these traditional business data types tend to be relatively small in size, move across a network in occasional bursts, and have little impact on other network traffic. If an error occurs during transmission, the transaction can usually be restarted with few negative effects.
In contrast, digital video data flows in an “always on” sequential stream. Due to the enormous size of this type of data and its consumption of bandwidth, transmitting video over a packet-switched network can quickly degrade the performance of the entire network. But it’s not just capacity that’s being chewed up. Large sequential data require fat pipes throughout the system. The streaming (isochronous) nature of video introduces time sensitivity that has unique characteristics. The real-time nature of video means that failures and maintenance to all system elements must be sustained dynamically without affecting the never-ending incoming video data streams.
Data compression and buffering offset some of the problems with digital video data, but they don’t solve everything. Other essential elements include a robust network and a high-performance storage management system.
Tip #3 is that general purpose storage systems are not designed for video. We have already discussed the need for massively scalable storage systems. Locally attached storage systems are barely sufficient for even a few days’ worth of data. A single megapixel digital camera can generate a terabyte of data in just a few days. With video surveillance systems utilizing dozens or even hundreds of cameras, a storage area network is your only option, but even here you must be selective.
General purpose storage systems may have a large enough capacity to hold sufficient video, but the performance can be disappointing. Such systems simply aren’t designed for video. Rather, they are designed to handle reference and archived data (like our business database) that is not accessed very frequently.
In addition, high-performance storage systems have poor characteristics for video use. They maximize random IO performance over sequential metrics. They rely on expensive custom components such as Fibre Channel switches. Once configured, they are not easy to change or expand. And, they require specialized IT skills to manage. Then, too, the cost of general purpose storage systems can be staggering. For example, a simple direct-attached storage system costs about $2 per gigabyte for storage capacity. When you move to the level of an enterprise storage area network, the cost jumps to about $8 per gigabyte. Even worse, a user must pay for storage capacity up front, perhaps years before it will ever be needed (if at all).
Tip #4 is that clustered storage is specifically designed for video workloads and user needs. The city of Long Beach has selected the Pivot3 High-Definition storage solution for its citywide video surveillance storage needs. The footage captured by video surveillance cameras is managed and stored on the Pivot3 scalable IP-based systems. This allows the public safety department to save significant financial resources by only buying the storage capacity it needs and grow it later, based on its requirements.
“We have built one of the nation’s first and largest wireless camera system infrastructures, using the 4.9 GHz spectrum for all of the cameras,” said Lieutenant Steve Ditmars with the Long Beach, CA Police Department. “The system has been embraced by patrol officers in the area and has been used in several cases.”
“Although our camera and wireless system were designed with expansion in mind, we realized only later the large scaling required from a storage perspective. That is why we chose Pivot3. We now have a storage system that can meet our current and future needs at a lower cost. The system was implemented quickly, and, because of its ‘pay-as-you-grow’ model, it costs us only a fraction of traditional storage area networks solutions.”
The Pivot3 High-Definition Storage Cluster addresses all of the issues of IO bandwidth, storage capacity, availability, reliability, management and cost, making it the ideal storage solution for video surveillance. The Pivot3 architecture provides direct, parallel access from the video servers to a series of inexpensive networked nodes called Databanks. This creates a high-performance storage cluster where the video data is distributed, protected and accessed in parallel across multiple databanks connected via common gigabit Ethernet.
This virtual architecture allows performance and capacity scaling that are absolutely critical for video storage. The cost of this system is less than half that of a traditional enterprise storage system at under $3 per gigabyte of capacity. The Pivot3 storage cluster is based on commodity hardware to minimize acquisition and operating costs. Also, it is easy to manage, as security personnel can manage it without the need of specialized IT skills.
In the end, it comes down to selecting the right tools to handle the data-intensive needs of video. Similar to the way the desktop market has evolved with different systems now available for gamers, business users and home consumers, the video market is now large enough and has specific requirements that will lead to the deployment of custom solutions that meet its unique needs.
Lee Caswell is the founder and chief marketing officer for Pivot3. He is an experienced technology executive with more than 20 years of marketing and management experience in the storage, networking, and digital video markets.