This article originally appeared in the September 2020 issue of Security Business magazine. When sharing, don’t forget to mention @SecBusinessMag on Twitter and Security Business magazine on LinkedIn.
When a crime occurs, law enforcement agencies are in a race against the clock to gather, review, and analyze every piece of evidence available. Today, video surveillance systems using high definition cameras and intelligent video analytics ensure security personnel have eyes everywhere, all the time. If operating properly, surveillance systems make it possible, via archived video footage, for criminal courts to see what happened, even if no person was physically present to witness the crime.
Here, however, lies the problem. When video footage is submitted as key evidence, the question of validity largely depends on the quality of the video. If surveillance footage is forensic-grade and definitively corroborates a prosecutor’s hypothesis, a conviction is likely. If video footage is low-quality or contains recording gaps, the footage may be considered inconclusive as evidence in a court of law, defeating the purpose of investing in video surveillance.
The causes of low-quality surveillance video range from improper system design to intrinsic deficiencies of the video surveillance system. System design can be fixed by correcting the design errors, which can be anywhere from using incorrect switches to overloading a system; however, the only way to correct intrinsic deficiencies of a video surveillance system is to switch to one with an architecture optimized for video traffic.
Potential Shortcomings of NVR Architecture
Network Video Recorder (NVR) systems are the most popular storage option for surveillance deployments with small numbers of cameras; however, they were originally developed for IT data processing. Like all data processing equipment, both the processing unit and memory are packaged in a single chassis to reduce cost. When applied to video surveillance, NVRs use disk drives to store large quantities of video data and appeal to security customers because of their abundance and affordability.
When a surveillance system has fewer than 75 cameras with low bandwidth, NVRs normally work quite well. They improve functionality by over-provisioning processing power and storage to compensate for any inefficiencies that arise from differences between IT data and video streams.
One of the chief problems with NVRs, however, is their lack of scalability, due to architecture and packaging. Most video surveillance systems evolve over time, with migrations to better cameras, increased area of coverage, increased retention time, and the addition of AI. All these changes require scalability. It should come as no surprise that, for surveillance deployments supporting enterprises that demand scalability, NVRs habitually fall short.
It is important to keep in mind that NVRs employ an architecture engineered to process IT data; however, video data streams have very different characteristics. IT data, for example, is often processed in short, sporadic bursts of activity, whereas surveillance systems generate video data that must be written in a continuous stream. If IT data storage is optimized for 70% read and 30% write workloads, yet video surveillance storage requires 95% write and 5% read, it is no wonder traditional NVR storage architecture can produce issues for surveillance systems, such as dropped frames and gaps in recording.
The bottom line is that when disk drives cannot write video data fast enough to keep up with the incoming data stream because of disk drive latency, then recording gaps are created. To address these problems, the security marketplace needs to embrace a new architecture, uniquely designed for video data.
Surveillance-Defined Architecture
An architecture optimized for video surveillance needs to be highly scalable and integrated with technologies that minimize drive latency and maximize data reliability. The defining characteristic of these solutions is that they scale computing and storage separately – enabling end-users to scale-out computing power by adding more server blades, if and when more CPU power is needed. It also enables users to scale-up storage capacity expansion chassis when more storage space is needed.
This independent scaling design allows storage architecture to grow as a surveillance system inevitably expands. One architecture that meets these goals is an IPSAN (Internet Protocol Storage Area Network)-based architecture, wherein storage devices are connected to processing units using IPSAN. When more storage is needed because of an increase in retention time, higher definition cameras, or the addition of more cameras, another storage unit can be added to the system by daisy chain.
With this surveillance-defined architecture like the one described above, an end-user can: Scale CPU and storage independently, paying only for what they need, which generates cost-savings; and benefit from a custom-built IPSAN storage array that is built for surveillance and thereby able to record thousands of video streams through a single processing unit with unparalleled efficiency. This helps to reduce the cost per stream and ensure no loss of video frames.
As mentioned, disk drive latency is the main cause of recording gaps. To reduce disk drive latency, a storage system in this architecture needs to incorporate data management technologies that can reduce disk head movements, which is the main cause of latency. With properly designed algorithms, it is possible to use ordinary disk drives to record video streams at up to 48Gbps.
RAID technologies are typically used in all video surveillance systems to guard against the loss of data. If data is corrupted, a data rebuild can be activated to recover the lost data; however, this process takes a lot of time, and the recovery time increases exponentially with the size of the storage. In the case of data loss due to multiple drive failures, even RAID 6 will not be able to recover the lost data. To enhance data reliability and integrity, storage systems should incorporate AI technologies that can determine the life expectancy of each drive, in order to proactively prevent data loss from multi-drive failures.
Thanks to surveillance-optimized technologies and features, these robust storage systems enhance recording throughput, data integrity and system reliability.
When integrators look to select a video surveillance solution for an enterprise customer’s surveillance system, be sure to choose a technology partner that engineers products with surveillance in mind. If an existing storage solution using NVR architecture isn’t producing the desired results, consider upgrading to an IPSAN based video surveillance system – one that allows a video surveillance system to grow with the customer’s ever-changing surveillance needs, while offering cutting edge data reliability and integrity.
Yu Hao Lin is COO of Rasilient Systems. Request more info about the company at www.securityinfowatch.com/10484633.
