Testing: Edge Recording (Genetec Trickling)By Benros Emata, Published on Mar 12, 2011
Edge recording is one of the most promising emerging video surveillance technologies. In many scenarios, it is highly desirable to record video (at least temporarily) on the camera. For instance:
- You have a remote site that only needs a few cameras. Rather than purchasing and deploying a recorder, it would be easier and cheaper to record video 'inside' the camera itself. Only when you need to watch the video would you recall the video from the camera.
- You are concerned about losing video during network outages. Usually, when a network (or connection from a camera to a VMS system) goes off line, video during that time period is lost. With edge recording, the camera could keep recording video and then send it back to the VMS.
- You have a situation where the camera is mobile and is not connected to any network. If you can record video on the camera and then synch it up later when you get back to your facility, operations are greatly simplified.
Historically, implementing edge recording has been challenging.
Traditionally, all video recording was 'centralized'. Cameras generated video. The video was streamed to recorders (DVRs, hDVRs, VMS systems, etc. ) and then the video was stored in the dedicated recording system. While users could deploy many recorders, video was always stored in one of those recorders (or connected storage systems). No video was stored in cameras / at the 'farthest' edge.
On the other hand, IP cameras are computers, specialized appliances of course, but computers nonetheless. Like any computer, IP cameras can manage storage. While all IP cameras always have some level of storage on board, it was generally rare for cameras to have significant storage (e.g., SD cards or hard drives).
In the last few years, support for SD cards on IP cameras has become increasingly commonplace. For example, nearly half of the 500+ cameras in our IP Camera Database support on-board storage.
However, VMS support for on-board storage has been limited. Most VMS systems offer no support for a camera's on-board storage. This meant that users deploying cameras with on-board storage would either have to use the cameras's web interface to manually recall stored video or, worse, pull the SD card from the camera. A few VMS systems offer support for a camera's on-board storage but only for the manufacturer's own cameras (e.g., March, Mobotix, VideoIQ). The most well known of these approaches is from Mobotix who emphasizes the benefits of their 'decentralized' architecture (see our Mobotix edge recording test results).
Genetec's 'trickling' feature is the first VMS offering to provide 3rd party IP camera storage support. Trickling is designed to support camera only recording or periodic transferring of recorded video from the camera to their VMS archivers.
In this test, we wanted to better understand the complexity of using trickling and the usability of accessing on-board storage. We tested Genetec's Omnicast 4.7 VMS with an Axis IP camera to better understand the interoperability between 3rd party VMS and IP camera vendors.
We believe Genetec's edge recording functionality will be very attractive for a number of reasons:
- The feature is available in any Genetec VMS version at no additional cost. Even those using the entry level Genetec VMS version can make use of this 'out of the box'.
- Video playback and video transfer from camera to recorder/archiver worked well.
- Trickling supports both 'pure' edge only recording and using edge recording for temporary / off-line storage
- It is designed and worked with 3rd party cameras (such as Axis)
However, a number of important issues / concerns should be kept in mind:
- Setting up edge recording can be time consuming as configuration is required on both camera and VMS side
- Two problems exist with integrating Axis cameras: currently, a registry edit needs to be done on the server side and continuous recording requires a 'hack' (both explained in the detailed sections below)
- Traffic / bandwidth management should be carefully consider as simultaneously trickling of many cameras can significantly increase load
- Genetec currently supports Axis and Bosch directly and other devices that integrate on board storage through their Genetec Protocol; users should check to see and note the limitations in overall 3rd party camera support
If you are already a Genetec user and comfortable with Genetec's overall complexity, this should not be a problem. However, if you are new to Genetec and find complexity to be a problem, you should carefully consider these issues (and review our Genetec test report).
Comparisons to Milestone and Mobotix
Two edge recording offerings that are cited often are Milestone and Mobotix:
Milestone offers edge storage functionality [link no longer available] for its XProtect Corporate VMS version 4. While Milestone has been promoting this feature for more than a year [link no longer available] and includes the feature in its Corporate 4.0 release, Milestone does not currently support any IP camera's on-board storage in production (confirmed directly March 2011). As such, the feature is essentially not available. Secondly, the feature is limited to only redundancy recording and only for their most expensive VMS version. As such, the Genetec version is significantly superior as it includes 'pure' edge only recording and for any VMS version.
Mobotix offers edge recording as a standard feature across its IP camera line. Two important positives for the offering are (1) no VMS licenses / no software costs at all and (2) edge recording is essentially plug n play as Mobotix is connecting to its own cameras. The key problem we found in our Mobotix test was issues with the quality of playback (stuttering, delays, etc.) and the complexity of using their VMS client.
Administration and Configuration
In this video we examine how to set-up and configure Omnicast's Trickling. Configurations are required on both the VMS side and the camera side. We start on the VMS by restricting video recording to the edge device only. Next, we explain the three methods for trickling - manual, on-schedule, and on-connection. We'll then walk through the available filters to selectively transfer or trickle video from camera to VMS/Archiver. Finally, we go directly into the camera unit to configure motion based recording of video to the camera's SD card.
Live and Playback Video
In this video we examine the workflows of live viewing and playback in a trickling environment. First, we show cameras configured for 'trickling' or edge only recording are still available for live viewing. We point out that live viewing and edge recorded video quality settings are configured independently (e.g., live view at 720p and record on edge at VGA). Next, we'll demonstrate the Archive Player directly query the camera's storage for video incidents/clips, and play them back directly from the camera through the Archive Player as well. Lastly, we'll demonstrate an actual 'trickling' operation to select/filter and transfer video from camera to VMS/Archiver.
We recommend you keep a number of important implementation considerations in mind when considering or deploying Genetec's edge recording:
Date and Time
Date and Time need to be synchronized between the camera and VMS. Best practice is to use an NTP server. Alternatively you can synchronize the camera units manually with the VMS/Archiver system.
Scalability and Management Issues
The distributed architecture benefits of trickling come with some trade-offs in scalability and management. During initial deployment for example administrators will have to interface each camera directly to configure a host of items, such as on-board storage, motion detection, event recording, NTP etc. Also, the camera systems must be monitored and managed for not only video but storage, motion filters, time sync, CPU usage, and other issues now that the camera is providing many more services. These associated tasks and considerations present a considerable challenge and will generally require stronger IT-centric organizations to manage/support.
*It is much simpler to monitor camera reliability/uptime when used for strickly video purposes, because an operator can easily confirm if he/she sees the video on a monitor/PC. However, on-board storage issues inside a camera are not so easily detected or monitored. The challenge is compounded as number of cameras increases as each camera represents yet another storage system to manage.
Network Traffic Issues
In larger 'trickling' deployments simultaneous downloads from camera storage to the VMS or Archiver may result in network traffic issues. This consideration is more relevant in scheduled trickling applications, where all cameras associated to the Archiver will initiate downloads according to a single schedule. The Omnicast strategy is to configure a maximum number of simultaneous downloads (e.g. a max of 5 cameras at one time). This will effectively stagger the downloads and help to avoid network traffic issues. Other approaches can include limiting the resolution, fps etc. being recorded on the camera, and/or making use of filters to limit what video is transferred.
Techniques to minimize network impact:
- apply (e.g. motion, bookmarks, shorter time frame etc.) filters to trim video 'fat'
- limit the number of cameras that can simultaneously trickle to the VMS
- determine and configure the lowest acceptable resolution and fps on the edge recording
Edge determines image quality for archived video (resolution, fps, etc.)
When video is trickled from edge to archiver it is of the same quality as configured on the edge device. This makes the record settings on the camera itself the limiting factor for image quality. As such, the 'raw' video to work with during investigations on historic video (on camera or on VMS post-trickling) will be limited by the edge recording parameters.
VMS determines image quality for live viewing
For live viewing purposes the image quality is configured independently from edge recording. This allows real-time operators the potential to view video feeds at higher, lesser, or equal image quality settings as configured on the edge device.
Continuous Record vs. Event Based
With 64GB capacity SD cards now available and efficient H.264 codecs, continuous recording is much more feasible than before. A 64GB SD card, for example, can provide nearly 72 hours (3 days) of continuously recorded HD video (2 Mb/s). Furthermore, 'tags' or metadata can still be placed on both live and playback video to help filter out non-interesting video during trickling (transfer to permanent centralized storage on Archiver/VMS).
On the other hand, an event based recording strategy at the edge will generally entail greater up front investments of time for set-up and maintenance, but promotes more efficient use of storage and network resources. A downside is that reliable collection of important video will be dependent on motion detection, and/or other sensors and triggers having low false negative rates. Also, the distributed network of motion and other sensors represents additional complexity to optimize and manage across many devices.
Note that the continuous record method reduces the complexity of the distributed environment. Filtering via metadata can be accomplished more centrally (e.g. via bookmarks and playback requests).
Continuous Record Advantages/Disadvantages:
- all video is captured
- no motion detection event or other sensor configuration hassles
- can still apply metadata for filtering video that is trickled
Event Record Advantages/Disadvantages:
- must configure scheduling, motion filters, sensor/triggers on each camera
- false negatives result in loss of 'interesting' video
Known Issues with Axis Cameras
There a two known issues with edge recording in Axis cameras:
- camera storage may fill up and stop recording
- continuous record method requires work-around
The first issue is addressed with a Windows registry edit on the VMS. Note that we did not test the registry edit. A stoppage of recording on the unit/camera can easily go undetected for a period of time, and may result in the loss of critical video.
During our testing we were able to continuous record on the camera itself, but could not achieve functional integration with the VMS (i.e. unable to neither view video directly from the camera nor download the video to the VMS using trickling). We were informed that currently there is an issue using continuous recording in a trickling environment with Axis cameras that will be fixed in the near future. A work around using scheduled recording of (8) 3 hour events was recommended that provides a degree of similar functionality but is time consuming to implement.