By Lee Caswell and Steve Surfaro
Practical tips to reduce video surveillance infrastructure costs
Designing storage systems by using camera simulators and employing new storage options
Two models for storage: With NVRs/servers and without (using serverless storage)
Image courtesy Lee Caswell, Pivot3
According to market research firm Frost & Sullivan, servers and storage infrastructure make up a substantial amount of video surveillance installation costs. Physical security integrators can take advantage of the right tools and solutions to minimize these substantial costs in order to make bids more competitive and improve reseller margins. Specifically, camera simulator tools reduce guesswork by providing realistic expectations of capacity and bandwidth needs, and clustered NVR storage can reduce storage over-provisioning and eliminate the need for standalone NVR server hardware.
Conventional surveillance infrastructure relies on traditional servers to manage a fixed number of incoming video feeds, and then store these streams on disk storage that is directly attached to each server. The bandwidth and capacity required for each camera are calculated upfront, and fixed storage and server capacity are purchased to match the load. This can become a lengthy process since each camera’s resolution, frame rate, compression and motion detection settings must be individually calculated. This fixed architecture is replicated as many times as the camera load requires, which can lead to chronic “over-provisioning” of storage that may never be used.
Actual mileage may vary
With any video-based surveillance system design, bandwidth estimation and network load are important considerations. Compression rates for MPEG4 and H.264 can vary based on actual motion observed in the field. H.264 represents a substantial bandwidth reduction over MPEG-4 compression (often up to 40 percent). Similarly, motion detection schemes may deliver quite different bit rates and capacity needs depending on the installation particulars. A less technical but common reason for variances in planned infrastructure is simply that camera changes are made during field installs to meet user recording specifications. Swapping in higher-resolution cameras to make a customer happy or beefing up frame rates to meet real-world needs can play havoc with carefully calculated server and storage requirements.
Using a camera bandwidth estimator, such as the Axis system design DVD, and calculating for different frame rates are appropriate efforts for design estimation.
Clustered storage reduces risk of being wrong
Unlike direct attached storage, clustered storage is shared by many NVR servers and can be dynamically changed without affecting the network or camera infrastructure. The aggregate bandwidth and capacity of the storage infrastructure are available and automatically load-balanced for the cameras. The risk of configuring individual RAIDs is minimized and the need to “over-provision” each RAID is reduced. If additional capacity is required to meet new camera or retention requirements, it is easily added either logically or physically with clustered storage.
This can result in important cost savings. For example, let’s take an installation requiring 200 TB of capacity. When using 10 fixed capacity 20 TB RAIDs, a prudent installer may over-provision each RAID by 20 percent or 4 TB to accommodate worst-case compression rates and unanticipated field issues. With clustered storage, the installer could factor in a 5 percent override of 10 TB for the entire installation and allocate the extra storage as needed. The 30 TB capacity reduction is a roughly $30K saving that can be used to provide a more competitive bid or improve reseller margins.
|Comparison ||DAS ||Clustered NVR SAN ||Savings |
|Overprovision factor ||4 TB/RAID ||10 TB/SAN || |
|Safety over-provisioning ||40 TB ||10 TB || |
|Cost at $1/GB ||$40K ||$10K ||$30K |
Server virtualization can eliminate server infrastructure
The latest innovation in clustered storage architectures is to run NVR applications on virtual machines running in the clustered server appliances. This application of server virtualization technology allows resellers to eliminate the cost, power, cooling and rack space of the dedicated NVR server infrastructure hardware.
For example if five dedicated external NVRs were to be deployed, then the installer could eliminate these five NVRs by running the applications on the storage hardware already deployed. Here’s a breakdown on the cost comparison:
|Comparison ||DAS ||Clustered NVR SAN ||Savings |
|Dedicated NVR servers ||5 ||0 || |
|Server cost @ $3K/server ||$15K ||$0 ||$15K |
|Server power & cooling ||5KW ||2.XKW || 44% |
|Rack space @2U/server ||10U ||-0- ||10U |
In today’s cost-sensitive environment, those integrators that manage the infrastructure costs of video surveillance bids will provide more competitive bids and retain higher margins. Using a camera simulator allows integrators to run performance tests with the actual NVR application, determine maximum performance, and more accurately estimate retention times. The use of clustered storage can also help by reducing over-provisioning and simplifying re-allocation of capacity and bandwidth in the field. The latest implementation of clustered storage also eliminates dedicated NVRs.
About the authors: Lee Caswell, is founder and chief marketing officer at storage technology firm Pivot3; Steve Surfaro is strategic channel manager and security industry liaison at IP video company Axis Communications.