Selecting Appliance and NVR Hardware
When it comes to video management and recording, the options for end users and integrators are predominantly three-fold: VMS with recording servers, VMS-based hardware appliances or NVRs. Typically, the latter devices are perceived as decreasing in flexibility as the solution becomes more hardware-specific, but is that still the case given recent increases in processing power?
In the past, most video management solutions could be defined as being either hardware- or software-based. Between the two options there was a gulf of difference in terms of features and functions, capacity and implementation. Crossover between the two was rare and many integrators found themselves in one camp or the other.
As with anything in technology, the pace of development is fast. Moore’s Law originally stated the capacity of processors would double every year. Despite being revised to two years in the mid-1970s, the reality is that processing power continues to rise and there is no reason to think that it won’t continue to do so, increasingly at a more accelerated pace.
This means assumptions made about software- and hardware-based video management and recording options five years ago, or even two years ago, will now be outdated and subsequently incorrect. Of course, as with any technology, there will still be devices and systems using outdated chipsets and processors in order to meet low price-points, so there are no blanket definitions that can be applied.
To further confuse things, the video management appliance has grown in popularity. These options fit between pure software and hardware alternatives. Where clearly perceivable differences existed between software and hardware choices, developments and advances have eroded the boundaries and for many integrators, a fresh view is needed to ascertain the best value propositions for specific applications.
Software-based systems have many significant benefits. Because software interfaces are more flexible, easier to adapt and are not limited by physical components with fixed abilities and capacities, the systems are quick to upgrade, simpler to interface with other solutions and often share commonality with regard to operating platforms. These benefits enhance the user experience as the system is usually customisable to allow users to get the features and functions they need from a system.
In a time when an increasing number of manufacturers are offering networked video, there is a need for video management solutions to interface with a wider range of devices including cameras and codecs, detectors, access control devices, plus a whole host of IoT sensors and peripheral building management systems.
Despite the best effort of various interoperability organisations, in order to benefit from the full functionality of the various connected devices, any management system must be able to implement a variety of drivers and active controls. Integration of these into a software-based model is fast and simple.
It is this interoperability that delivers an ability to create open systems. Because the software supplier is delivering direct integrations, it does allow a truly ‘open platform’ approach.
There is, of course, a downside to the benefits of software and that comes in the form of required infrastructure. For some applications an additional investment in robust servers and networking means that certain solutions are not financially viable, especially in smaller or low risk applications. Where existing corporate networks are running close to capacity, the IT department will do everything to keep away any bandwidth hungry services such as streaming of multiple video channels.
For many integrators, software-based solutions necessitate the specification of appropriate hardware, which in turn can complicate the installation process, especially if the customer has an incumbent IT supplier who has little or no knowledge of the system requirements for a security solution.
In such circumstances, a ‘black box’ solution is often seen as preferable. Such options come in the form of NVRs and increasingly as VMS appliances.
For many years, hardware-based solutions were the cornerstone of video management and recording, and for good reason. Dedicated hardware designed specifically to meet the demands of always-on security solutions can offer user-friendliness, enhanced resilience and ease of use and installation. Also, with hardware-based devices, there is a clear delineation between what is a part of a video system and what it not.
Consider connectivity. The majority of connections for a networked video solution will be the same regardless of whether it is an NVR, an appliance or a server running software-based VMS. Ethernet ports, display outputs and connectivity for peripheral devices are pretty much standard. However, a hardware device specifically designed for security can also include integral PoE switches, input and output connections, test circuitry, etc..
With regard to user-friendliness, dedicated security surveillance hardware delivers direct control over the video system. There is no risk of conflict with the operating system as this will be embedded, no potential issues with additional IT services running that might affect performance, and no peculiarities for simple device management. By adopting a singular approach, dedicated video surveillance hardware can deliver a more focused experience.
One reason that many integrators prefer hardware-based systems is that they present a clear definition of what is part of the surveillance system and what is not. If a video surveillance hardware device fails, then it is obvious the video system is at fault. It will either be the central hardware, the devices connected to it, or the various connections. Either way, the integrator knows it is their responsibility. However, if a software package fails, is it the software or the server it is running on, or the network the server is connected to, or any combination of the various elements?
Benchmark recently spoke with a manufacturer, who had been called to site following an integrator’s inability to solve a fault. The IT department insisted the problem was not theirs and was placing pressure on the integrator to rectify the issue. After a week of troubleshooting, the IT department finally realised the issue had been created by configuration changes they had made. Such was the nature of the issue that the integrator would never have been able to identify it given their authorisation and permissions for the overall system.
The downside of hardware is that many devices, especially those at the budget end of the market, are limited by the capacity of the physical components used. While some of the limitations might be acceptable today, the pace of change as dictated by Moore’s Law does means that in one or two years the hardware might be unsuitable for newer devices and technologies. Upgrading hardware-based systems is not a simple task and often necessitates the replacement of core elements such as NVRs.
Additionally, any hardware element will have limitations in terms of how much data it can manage. With a server-based system using software, some degree of management can allow loads to be customised to suit the user’s needs. With hardware solutions, these options can be less flexible as the manufacturer usually imposes limitations to ensure the performance does not degrade if loads become heavy.
The middle ground
Most video surveillance manufacturers who produce NVRs are aware of the flexibility inherent in software, so they typically will offer some form of software-based VMS. Equally, many VMS suppliers recognise the popularity of a single-box solution. The ability to plug a device in and have it managing and recording surveillance footage without any installation or program configuration is appealing. As a result, the development of surveillance appliances has increased.
Appliances plough the middle ground between software-based VMS systems and NVRs. They are typically small form factor hardened servers which feature an embedded OS and VMS package, often pre-licensed and partially configured. They allow end users and integrators to enjoy the flexibility of a VMS, combined with the ease of use and faster installation which is typically associated with hardware-based NVRs.
Appliances do lack some of the flexibility of software-based alternatives, but this is due to the fact the hardware’s capacity is a limiting factor. Priced to be competitive with NVRs, the appliance servers are built to a cost which in turn can restrict scalability.
The right specification
As processing power has increased, many of the issues which impacted on the performance of NVRs and appliances have been eliminated. One example is the management of video streams. In the past it was not uncommon to see NVRs billed as HD recorders, but these were often unable to actually record HD compliant footage. For example, a 16 channel NVR with a recording rate of 200fps would only be capable of recording 8 HD streams. If the full capacity was used, the cameras could not record real-time, which is a requirement for HD.
Standardised resolutions such as HD and 4K UHD require specific resolutions and frame rates. If both elements are not correct the stream cannot be classified as HD or 4K UHD.
Where frame rates were delivered to ensure compliant footage, another limitation was often introduced: bit-rate.
If inadequate processors are specified, a compromise will need to be made somewhere. If it is on frame rate the video might not be standards-compliant, but if it is on bit-rate, the video quality may suffer.
Users are familiar with HD and 4K UHD video as it used widely in other systems as well as in their home lives. Therefore, their expectations are of a certain image quality and if that is not achieved they will ultimately be disappointed with their investment.
Benchmark tests have showed that to achieve a good level of detail without degradation, bit-rate for an HD stream needs to be between 4-10Mbps, dependent upon the level of activity in a scene. Therefore, if a restricted bit-rate is imposed, it may not be possible to ensure that image quality will match customer expectations. It is also important to over-estimate throughput as environmental conditions will obviously increase the load.
When considering resolution, frame rate and throughput, it is important to establish whether figures are for live viewing or recording. Often specifications include two sets of figures which differ greatly. As the latest processors find their way into more NVRs and appliances, such load-based issues should become less common. However, at the budget end of the market, it is likely that older processing options may be used for some time to come.
Advances in processing aren’t just enabling greater flexibility with regard to video quality. Where once the majority of NVRs were only compatible with devices from the given manufacturer, third party devices are now supported by many recorders. The options vary dependent on make and model. Some are still limited to devices from the manufacturer of the NVR, while others include compatibility with devices from a limited number of partner companies. Others will support the main functions of ONVIF, but a growing number of hardware manufacturers are embracing the open platform approach.
Often video appliances take a more open approach, using device drivers created for the software-only packages. However, often the information regarding the full extent of support for third party devices – and indeed whether all functions of a device will be available – is not immediately available. That said, manufacturers who fully embrace the open platform approach will provide comprehensive lists of compatible cameras, codecs and other products.
With regard to ONVIF, it is worth remembering that previous specifications allowed statements of compatibility when some, but not all, criteria were met. As such, there can be connectivity issues. ONVIF’s newer Profile S is more robust.
Today’s NVRs and appliances are significantly more powerful than the devices of just a few years ago, and in terms of video performance and quality, a careful specification will ensure that high quality real time streams can be managed. Additionally, for those who wish to use best-of-breed technologies, a more open approach makes this possible. There is no longer a need to compromise on image quality or compatibility if selecting a hardware-based system.
Added intelligence with GPUs
One of the more significant benefits to be realised from increases in processing capacity has been the ability to enhance the intelligent services offered by NVRs and appliances. These include enhanced search and scenario creation functions, video and audio analytics, event management features and even the inclusion of business intelligence functionalities and integrations with business systems.
The increased use of GPU processors will see additional features and functionalities become ever more widespread. In the past NVRs and appliances relied on CPUs (central processing units) rather than GPUs (graphics processing units). The CPU makes use of a few cores that process a few software threads simultaneously. A GPU makes use of hundreds of cores to manage thousands of software threads simultaneously. As a result, GPUs can accelerate processing by up to 100 times over traditional CPUs.
This huge acceleration in processing speeds and increased data handling capacity enables captured data to be used effectively. For example, searches can use a greater amount of criteria and the increased speed allows searches across all recorded footage rather than specific time windows or video streams. Date and time information for searching is predominantly used to limit the amount of data being processed. With GPUs this is no longer a concern. The upside is that suspects may have visited the site on previous occasions and this can be included in searches.
Obviously, greater search parameters will result in data matches across numerous cameras on varied dates. Therefore, an NVR or appliance should be able to create evidence-based scenarios, allowing all associated data – video, audio, alarms and status information – to be grouped together, making it easier to manage or export when dealing with incidents.
Just as GPUs can enhance search and video management, they also allow more effective use of video and audio analytics. Machine learning eliminates a degree of the required set-up and configuration associated with optimising IVA. Changes in environmental conditions, even down to fine details like low sun in winter months, can impact on the accuracy of visual content analysis. The computational power of GPUs makes learning such conditions a more achievable task, all without impacting on performance.
Video is a great enabler and as such represent a highly beneficial tool with regards to site management. Depot managers can assess traffic flow in and out of a site, verify which loading and unloading areas are free, ensure that health and safety policies are correctly implemented, and capture status details about deliveries and collections via video streams. In a warehouse, the video surveillance system can offer security along with tracking of assets, code scanning, replenishment reporting, etc..
If something can be visually verified by a human, then video can deliver the same degree of verification to local or remote systems. It can also collect detailed information that might be laborious or not financially viable for staff to collect. Integration with systems designed for specific vertical sectors enables added value to be realised.
A final point is that NVRs and appliances which utilise GPUs can also enable hardware acceleration to boost performance when multiple high resolution streams are being managed.
Moore’s Law assures us that within 12-18 months the potential processing power in NVRs and appliances will be double what it is today. The same is true if we look back to 12-18 months ago. Devices launched then will be left behind by today’s products. Keeping up with how manufacturers have exploited that additional power shows hardware-based systems are no longer subject to the limitations of the past.
It is vital for integrators to remember that while the capacity and performance of chipsets has increased, it is not necessarily true that all new products make use of the latest components.
Older products still in production may not be upgraded. Devices aimed at the low cost market might retain the use of older processors. Stocks of older components might still be in the supply chain.
Integrators and end users should ensure they are specifying and receiving the latest generation systems, because if they are not, in less than 12 months the systems which have been installed could be several generations behind the curve.