With the continued migration from analog to IP that is taking place in the video surveillance market, one of the most important factors that organizations must consider is the demands network cameras will place on their existing IT infrastructure. What, if any, upgrades will have to be made to handle the increased bandwidth and storage requirements? The evolution of high-resolution imaging technology has dramatically improved the quality of video evidence, but not without a cost. A significant number of businesses simply do not have the IT infrastructure in place to handle the data generated by today’s advanced surveillance solutions.
In traditional IT workloads, such as the handling of emails or customer resource management (CRM) solutions, data is generally written to a disk or some other storage mechanism about 20 percent of the time and read approximately 80 percent of the time. Video surveillance data, on the other hand, works completely opposite. Data is written to a disk or storage device nearly 99 percent of the time and read around one percent of the time. Instead of having hundreds or thousands of workstations reading data as you would in a normal IT environment, thousands of cameras are now generating video data and only a small number of workstations are accessing that data.
While there is some variability depending on image resolution, frame rate, compression, scene activity and other factors, as a general rule, an IP camera will generate anywhere from one to 10 megabits-per-second (Mbps) of data per camera on a regular basis. This massive amount of highly variable data has to be ingested and written to discs nearly simultaneously. Video surveillance, especially larger deployments, places high demands on the bandwidth and storage requirements of an IT network. Additionally, most IT storage solutions are optimized for read-centric or read-intensive environments, which make them less than ideal for surveillance applications.
As a result, many organizations tend to over design their systems to handle worst-case scenarios — all cameras transmitting at 10 Mbps, for example — which leads to potential overprovisioning and an unnecessary increase in costs. Traditional methods of replicating and backing up data are also impractical and inefficient when it comes protecting stored video. It’s not surprising that one of the most common reasons for the failure of network video systems is the inability of server and storage systems to keep up with the massive ingestion of data, which results in the loss of incoming video. Other reasons for failure include the breakdown of system components, such as servers, disks and controllers that affect the ability to view live video or retrieve recorded footage.
Pitfalls of Traditional Architectures
To alleviate these concerns, many end users have opted to leverage direct attached storage (DAS) on their network. These solutions are similar to VCR and DVR dedicated storage architectures, but have their drawbacks:
- A single point of failure that could lead to lost video and system downtime. If the server, NVR or attached storage device goes down, end users run the risk of having their video network fail and their recorded footage lost.
- Stranded video “islands” that could prevent the user from retrieving recorded video in the event of a failure.
- Inability to scale a system as video needs change.
Another popular way that end users have sought to address the burdens of video on the network is through the use of network attached storage (NAS) or storage area network (SAN) solutions. In theory, this type of architecture is an ideal fit for the surveillance user given the large amounts of storage capacity that can be shared across cameras and video management servers. However, these systems also have their drawbacks:
- Many SAN and NAS systems have been designed for general IT environments and are not optimized for the demands of video.
- There are added costs involved in purchasing and maintaining these complex and separate VMS application servers.
A Virtualized Solution
The best way to keep these issues from becoming a problem, as well as keep costs at a minimum, is through virtualization. Using off-the-shelf servers and simple Ethernet connectivity, providers of hyper-converged infrastructure (HCI) have combined server and storage functions into a single appliance. HCI infrastructure can be applied to video surveillance storage applications to overcome the deficiencies commonly associated with direct attached or network attached storage solutions.
Because the storage pool is virtualized in the appliance (or multiple appliances across an array), issues surrounding scalability are no longer a concern as individual appliances can be added as needed to meet the growing surveillance needs of an organization (i.e. evolving surveillance needs, additional cameras or intelligent devices.) This scalability gives surveillance users the ability to make the necessary modifications to bandwidth requirements, which commonly change during the evolution of any video installation project. So, adding a camera here or there when a dead spot is found or increasing resolution can be accomplished easily.
Beyond the pooling of storage, failover protection prevents video data from being lost. While RAID or “erasure coding” is used to protect data within a single server or disk, HCI appliances use that same concept to protect data across all of the appliances in an array. This not only shields the end user from a disk drive failure, but from the failure of an appliance. This type of protection is not that common in the hyper-converged world, but being able to do RAID across appliances is a unique differentiator in today’s market.
HCI platforms also eliminate the need for dedicated VMS servers, which can be simply integrated within storage capabilities using the VMware ESXi hypervisor. That physical server can now be removed from the configuration entirely because it is no longer needed. The VMS doesn’t know it is running in a virtual machine and should that machine go down, it enables the client to recover that machine along with all their accompanying software licenses by rebooting it on another machine in the same array. This VM failover process typically takes about five minutes, whereas in a DAS or SAN environment, it could take anywhere from 24 to 48 hours — resulting in loss of camera recording or access to recorded video.
Considering the costs savings and the ability to prevent loss of critical data, virtualization is the best choice for organizations that need a robust IT solution to support their video surveillance network.
