VMS/VSaaS Recording Guide
This 23-page guide provides an in-depth explanation of VMS/VSaaS recording, including NVRs.
Inside we cover:
- Recording Management Locations
- Managing Recording vs Recording Storage
- Physical Recording Mediums
- Network Storage
- 2 Methods For Storage
- Types Of Recording
- File Formats
- Recording Databases
- VMS Recorder Hardware Capacity
- Storage Duration
- Retention Laws/Regulations
- Storage Capacity
- Calculating Required Capacity
- Frame Rate Impact
- Smart Codecs Impact
- Recording Scheduling
- Automatic Storage Deletion
- RAID / Redundancy
- Trickling
- Encrypted Recording Uncommon
This is part 2 of our new VMS / VSaaS course starting in the fall.
Overview
Video recording is a ubiquitous and critical aspect of VMS/VSaaS because many systems are not monitored live, and even those that are, typically have more cameras than humans could witness all events reliably.
Additionally, recorded video is used as evidence after an incident happens, to identify subjects, review activities, or share with others.
Managing Recording vs Recording Storage
While video is most commonly managed and recorded in the same physical device, managing recordings and storing video are 2 fundamentally different processes.
Storing video, in practice, is no different than saving files to a file system on a computer. However, managing video surveillance recording requires a process to index video files in a defined location for a set length of time, which is critical to a properly functioning system.
Using a single hardware device, an NVR, for recording and storage is simple and eliminates points of failure. However, this simplicity comes with limitations, and separating the 2 can provide scalability and flexibility.
Video Recording Locations
Video recording is the process of storing video and associating it with a database or directory of each video file that is stored. Video can be recorded across 4 fundamental options in video surveillance; NVR, VMS, Camera, and Cloud.
Camera
Recording in a camera can eliminate the need for a separate recorder but as many surveillance systems have more than 1 camera, having to record each camera individually is more complex than NVR/VMS/Cloud many-to-one architecture.
While most IP cameras support internal storage (e.g. microSD, SSD, Flash), some also support external file share storage for increased capacity. However, using internal storage higher cost per byte than NVR/VMS Recorders. Because of the cost and complexity, camera-based recording is not common.
NVR
Because they are an all-in-one hardware and software appliance, NVRs offer the lowest cost for most video surveillance applications. Multiple cameras are recorded by a single box, and video is most often stored on internal storage.
NVRs rarely support recording redundancy, and storage expansion is often limited to the number of hard drives that can be installed in the NVR. Moreover, NVRs are a single point of failure, which could result in multiple camera recordings lost because of a single broken/damaged/stolen NVR.
VMS
Because VMSes are, by common definition, software applications, they can typically be installed on commercial-off-the-shelf computers. This offers hardware flexibility to build your own system, based on preferred hardware components or manufacturers, which is not possible with NVRs. However, this flexibility also requires specialized knowledge to design for specific system requirements and is significantly more expensive than NVRs.
Many VMSes are modular, meaning they can be configured either as an all-in-one system (fundamentally the same as an NVR) or with separate system management servers and VMS recorder servers. However, some VMSes have maximum camera license limitations per system, which may limit expansion, similar to NVRs.
Additionally, while VMSes most commonly record to internal storage, support for network-based extended storage is widely offered.
Cloud
While cloud recording is the most scalable and flexible option, it needs consistent Internet connectivity to the on-premise devices. Cloud supports one camera to unlimited camera recording, with some VSaaS supporting storing video in the cloud and/or on-premise devices.
While cloud recording is effectively the inverse of NVR recording, offering expansion, flexibility, and redundancy, there are practical limitations to the number of cameras that can be recorded at a location, based on the Internet upload bandwidth.
While cloud recording is redundant in practice, it typically requires an expensive recurring subscription which cameras, NVRs, and most VMSes do not. However, using cloud recording eliminates or decreases the need for on-premise hardware, other than cameras.
Physical Recording Mediums
The most common physical recording mediums are Hard Disk Drives (HDD), Removable Memory Cards, and Solid State Drives (SSDs):
Hard Disk Drives
Hard Disk Drives, or Hard Drives/HDD, are the most widely used and lowest cost physical recording medium in video surveillance. However, using HDDs implies that one needs to stream from cameras to a device where the HDD is contained, as using HDDs inside of cameras is impractical, and virtually never done.
HDDs used in surveillance ranges from 1TB to 18TB, but IPVM statistics show that 4 to 8TB drives for video surveillance are most commonly used.
The average street price is $25/TB for surveillance-rated drives from 4TB up to 16TB which is ~10% higher than consumer-grade HDDs. However, in many cases, enterprise data center drives have similar performance ratings and longer warranties.
While many low-cost NVRs use standard consumer-grade HDDs, specialized surveillance hard drives are marketed for continuous writing of video, which is not a common concern outside of video surveillance.
Removable Memory Cards
Removable memory cards (e.g. SD, microSD, Compact Flash) are primarily used to store video in IP cameras and do not require a user/installer to disassemble the camera to install or replace. These can be used both for primary storage and/or redundancy, recording to the camera while the network is down and the recorder cannot be reached.
Removable cards are ~10x more expensive per byte than surveillance-grade hard drives, and support in NVR/VMSes is limited. Because of these factors, recording to removable cards is not common, as it is generally cheaper and easier to connect to a recorder that stores the video. Additionally, if a camera is damaged or stolen, the loss of recorded video is likely.
On the other hand, removable card recording is widely offered on IP surveillance cameras and recording on cameras is a growing trend marketed by a number of VSaaS emphasizing the elimination of NVRs.
Solid State Drives
Solid state drives (SSDs) are not commonly used for video surveillance recording in NVRs or VMSes but are increasing in network storage devices (SAN and NAS). SSDs require less power and have no moving parts, unlike HDDs. However, cost-per-byte is ~10x higher than surveillance-grade HDD, and commercially available sizes are generally limited to 4TB.
Very few camera manufacturers have support or use SSDs, only Verkada and a few others have. Moreover, 1TB SD cards marketed for video surveillance are available, which are much smaller, and replaceable.
While there are longevity concerns in video surveillance due to continuous rewrite cycles, modern SSD technology is used in enterprise environments with similar reliability and warranties as comparative HDDs.
Network Storage Expansion
Network-based storage devices are used to expand storage capacity (more hard drives) to an NVR or VMS, but generally do not directly manage the recording. This expansion enables a system to support more cameras, store longer recording times, or increase the resolution/quality of the current cameras.
There are 2 common options in video surveillance for adding network storage; NAS and SAN.
NAS storage is a low-cost option of network-attached storage for small-scale systems. Storage speed and redundancy are less common than Recorder or SAN-based systems. NAS devices are simple to set up, and they typically use Network File System (NFS) protocols, transferring video in a file-by-file process. This means they appear in a recorder's interface like a standard hard drive, but are slower than SAN or internal drives.
Storing video in a SAN offers large-scale expansion, supports flexible design architectures and increased data redundancy. A SAN is often connected over Fibre Channel, and commonly uses SCSI protocols to access deeper block-level data for quicker writing or reading of files than a NAS.
Adding secondary hardware for storage will typically significantly increase costs when compared to adding more storage in the primary recorder. Moreover, SANs typically require manufacturer-specific training for higher-level technicians or engineers to configure and support.
2 Methods For Recording
There are 2 common methods for recording in video surveillance; Record In Place and Buffer and Record.
Most systems use Record in Place, however, there are 2 challenges of video recording that can be addressed Buffer and Record. Because recorders may use high capacity, slow hard drives, fast playback or instant replay video may take 5 - 10 seconds to buffer before playing, which can be too long in high-security environments. Secondly, as VSaaS offerings use the Internet/Cloud for recording, during periods of high video bandwidth use, a system's Internet connection may not have the upload capabilities needed.
To address these issues, Buffer and Record systems use 2 storage locations; a buffering device and a primary recording device. This method is less common but used by one of the largest VMSes, Milestone. Moreover, it is growing in VSaaS architecture to decrease issues related to Internet upload bottlenecks.
Buffer and Record systems initially store the video in a temporary, small-capacity drive, typically containing less than 2 days' videos. Video is then moved to a secondary, primary storage location where it remains for a specified duration.
For example, this is used in cloud-hosted VSaaS, temporarily buffering video on-premise in a camera or gateway, and then backed up to cloud storage:
When using high-speed (10k+ RPM) HDDs for the temporary buffer location, Buffer and Record would offer moderate performance advantages compared to a slower speed (7.2k RPM) HDD used in Record in Place, but only if reviewing video while it is still on the faster speed drives. As such, this would primarily be useful in live monitored environments where instant replay or quick search/investigation of recording was common.
Record In Place is the more common and simple of the methods and is used in NVRS, most VMSes, and some VSaaSes. As the name indicates, Record In Place systems store video once and it remains in place for a specified duration.
This method requires fewer hard drives, a single recording database, and lower recorder hardware requirements because the recorder does not have to transfer video while writing new video.
Types Of Recording
Video is recorded using 3 fundamental strategies; continuous, on-motion/event, and boost on event/motion.
Continuous recording means the video is always being recorded, even when there is no activity on camera. This is most common in high-security installs, corrections, and locations where money/transactions take place (e.g. banking, casinos, hospitality).
It ensures that everything that happens is recorded, but can require significantly more storage capacity than event/motion recording. This will increase the cost of hard drives and may also require additional servers.
Motion or event recording means the video is only recorded when the camera detects motion/analytic object detection) or is triggered by an event (e.g. input monitor, manual trigger).
This can require significantly lower storage capacity than continuous recording, however, runs the risk of missing a critical incident if it is not detected.
Additionally, in environments with constant motion, the storage savings by using motion recording will be minimized.
Boost recording is a combination of continuous and motion recording. Low frame rate and/or resolution video is recorded continuously, and when there is motion detected, the system records an increased frame rate and/or resolution video. Using boost recording ensures that events are not missed, even if only recorded at low quality.
Prior to the wide availability of H.264/H.265 and Smart Codecs, this was used to decrease storage requirements when using MPEG4 or Motion JPEG. However, H.264/H.265 and Smart Codecs effectively decrease the storage requirements in the same manner, with significantly less complexity.
File Formats
Video is typically stored using 2 fundamental formats; Open File Formats (e.g. ".mpg") or VMS File Formats.
VMS File Formats are more common in high-end VMSes, while Open File Formats are typically used in low-cost NVRs, VSaaS, and camera-based recording.
VMS File Formats are a wrapper that contains a standard format video (e.g. H.264) and metadata, which includes watermarks, bounding boxes, motion details, bookmarks, etc. As such, VMS File Formats require specialized software to decode/view.
Open File Formats contain video, no metadata, but do not require special software to view. Open File Formats generally do not offer encryption or watermarking, typically used in high security/large enterprise environments.
Recording Databases
When a video is recorded, the file is typically indexed using a recording database, which points viewing software to the correct storage directory.
Searching through a database is significantly quicker than a standard file directory, which is critical for investigations. Additionally, video metadata (e.g. motion, object classification, etc) is often stored in the same database, which is used for advanced video searching (e.g. people, vehicle, area of interest, etc).
VMSes commonly use Microsoft SQL, SQLite, and NoSQL database management software, which offer free and/or open-source versions. The database software is typically pre-packaged with the main VMS installation application but may require installation prior to installing the recorder software.
VMS Recorder Hardware Capacity
Specifying the right sized PC/server for a VMS Recorder is critically important. A VMS Recorder that is underpowered will drop video frames during recording and struggle to stream video to clients trying to review recordings. A VMS Recorder that is overpowered may be prohibitively expensive for an application.
VMS Recorder performance is most dependent on CPU and RAM, directly impacting how many cameras may be processed for viewing and recording.
Exact requirements vary depending on the specific VMS used, but the following is a consensus across manufacturer specifications:
- Small systems, ~1-16 cameras: Dual-core processors, such as Intel Core i5, 4GB RAM
- Mid-size systems, ~16-64 cameras: Quad-core processor, such as Core i7 or i9, 8GB Ram
- Large systems, ~64+ cameras: 6-core processors such as Xeon 3rd/4th Generation Scalable Processors or Core i9, 16 GB RAM
Some processor and RAM configuration choices may be restricted by form factors in pre-built systems. For instance, Core i3 processors are generally only found in desktop PCs, while Xeons may be found in high-end workstations and servers.
NVRs are designed and certified that a specific 'box' supports a maximum number of cameras (e.g., 4, 8, 16, etc.). Of course, this simplicity comes with limited flexibility for expansion. Cloud recording is effectively unlimited by definition, and camera recording is 1-to-1.
Storage Duration
Because some customers have minimum required days of recording, and others maximum allowed days, proper configuration of how long to record video is a necessity.
However, cameras and many DVR / NVR appliances only support a limited number of hard drives internally (1 and 2 being common for smaller systems). This can constrain storage duration, especially given the low cost of those appliances and the extra cost of expanding beyond them.
30 days of storage is most commonly offered, due to the decreased cost of storage over the last 10 years and the increased efficiency of storing cameras with smart codecs.
In some environments or countries, there are regulations and requirements for the maximum number of days recorded, or a minimum number of days due to regulations (e.g. food safety, health care, corrections).
Although surveillance video retention is often constricted by storage costs, written laws or regulations sometimes dictate a minimum or maximum period video must be kept for. However, there are no global or regional standards, so research is required in countries, states, and local jurisdictions.
In the United States, retention laws are very fragmented, with minimum/maximum periods of retention ranging significantly (i.e. 7 days to 90 days) depending on location and vertical.
In Europe, requirements are equally vague, because GDPR has no specific minimum/maximum retention period for the EU broadly but says the rule of thumb is to keep personal data (i.e. video recordings) for "as long as necessary, as short as possible." A 4 week retention period is typically cited by industry sources for a reasonable retention period, although there are technically no laws prohibiting a longer period.
The impact of legal mandates for varying retention periods is significant. For example, casinos typically have thousands of cameras, and longer retention periods could result in a significant increase in ongoing storage and maintenance costs.
Storage Capacity
Determining the amount of storage required is a key step in designing a surveillance system. Doing so can be difficult as bit rates can vary across cameras but most estimate conservatively.
The number of drives supported ranges from typically just 1 in a camera to generally a few in a recorder to a dozen or dozens on larger servers.
In surveillance storage, drives are typically offered in Terabyte (TB) sizes. Small systems may require as little as 1TB, while large systems can scale into hundreds of TBs to thousands of TB (1024 TB is 1 Petabyte) or multiple Petabytes (PB).
Calculating Required Capacity
Most surveillance manufacturers provide system storage (and network throughput) calculation tools to help determine the total amount of storage capacity required. However, many of these tools overestimate the amount of storage, increasing the number of hard drives and servers required, which can significantly increase system costs.
Calculating the storage required is a basic equation using the bandwidth estimated per camera, over the period of time stored. Additionally, for systems that are using motion/event-based recording, a variable is used to account for how much activity is expected on each camera.
Because network bandwidth is generally measured in bits per second, while storage is measured in Bytes, a division of 8 is included to make the conversion.
For example, 1 camera recorded continuously with 1Mbps (1,000,000bps), recorded for 30 days, requires 3.4e^11 Bytes, or 324 Gigabytes (GB) of storage.
For the same camera, but recorded on motion, with 8 hours of motion estimated (33% or 0.33 motion factor), only one-third of the storage or ~108GB is required.
These calculations become increasingly complex as different cameras often have widely different bandwidths and motion percentages, even if they are the same model, but looking at different fields of view.
Recorded video frame rate has a direct linear relationship to the amount of storage that is required. For example, reducing the number of frames being recording in half will result in half the storage being used.
An IPVM integrator survey showed that the average frame rate is ~15 fps with 70% of integrators using between 10-20 fps. With none using fewer than 5 fps and 17% using more than 20.
Because of the continued drop in the price of storage for over a decade, and decreased bandwidth requirements due to Smart Codecs, compromising storage requirements by recording lower frame rates is not efficient.
Impact Of Smart Codecs
Smart Codecs reduce camera bandwidth, and by extension recording storage requirements. Additionally, because smart codecs work within the standard framework of H.264/H.265, additional support is generally not required to use with VMSes and NVRs.
The reduction ranges from drastic to minor, depending on the rate and size of motion in the camera field of view. Because smart codecs may have little decreasing effect in high motion periods, storage requirements must be calculated for worst-case bitrates, not average, or risk estimating not sufficient storage.
Reference the IPVM Smart Codec Guide for more details and examples.
Recording Scheduling
Because there can be different priorities for recording depending on the time of day or week, schedules are primarily used for storage optimization. While a system may require continuous recording during business hours, less storage can be used by switching to motion recording after hours.
This is less commonly used as storage costs have decreased significantly.
Automatic Storage Deletion
Recorders, generally by default, follow a first-in-first-out (FIFO) method for managing video. In practice, this means recorders will typically delete the oldest video to free up space for new video to be recorded, when space is limited.
Most systems allow means to protect specific video clips from automatic deletion to preserve specific evidence of high importance (security incidents, criminal cases, etc.).
RAID / Redundancy
RAID is the most common and least expensive method to provide file redundancy for recorded video. Video file data is written in across multiple drives so if 1 drive fails (or more drives depending on the RAID), the storage can be recovered. The most common RAID architectures used in surveillance storage are RAID 5 (1 redundant drive) and RAID 6 (2 redundant drives).
Redundant drives with RAID do not add to the total storage capacity of the system, so there is an added cost for storage. Moreover, RAID storage can take hours to recover and rebuild when a drive fails, which can significantly impact system performance
Redundancy
Redundant storage means that video is recorded in 2 recorders simultaneously, while backup storage means the video is first recorded on primary storage and then copied to a secondary/backup storage location. This means recording continues even with a catastrophic hardware loss (e.g. recording server motherboard failure).
Fully redundant and/or backup servers will effectively double the cost for storage because twice the server and hard drive resources are required for both. These are uncommon in most surveillance systems.
Trickling
Trickling is a form of recording redundancy in which a video is typically recording simultaneously to the camera and a recorder. If the camera's network connection is down and the recorder cannot be reached the camera continues to record locally. Once network connectivity is restored the camera will upload the locally recorded video to fill in any gaps in the server recording.
This is a feature typically only supported in enterprise-focused VMSes and cloud-storing VSaaS systems with camera or gateway recording.
Encrypted Recording Use Uncommon in NVR and VMS
Almost all non-VSaaS systems default to password-protected but non-encrypted recording, meaning you need to log in to a client with a user and password to view recordings, but the video files are unencrypted in storage.
Encrypting recorded video adds significant computational cost, 20-30% higher than non-encrypted, which is the primary limiting factor in price-conscious NVRs and VMSes. Additionally, because many NVRs and VMSes are locally viewed and recorded, encryption is not commonly seen as necessary. However, if the network is breached, and someone gains access to recorder storage, the video files are unprotected from unauthorized viewing.
VSaaS manufacturers typically market their systems "end-to-end" encryption, meaning the video is encrypted as soon as it is captured by the camera. Because these systems are designed to stream and record over the Internet, encryption is a much more critical aspect. However, this also means that password management is highly critical, as essentially all VSaaSes are exposed to the Internet, and mass hacking can result.