What is Ocularis X?
Ocularis X is more mysterious than its name. The new offering from VMS supplier OnSSI is marketed as 'bringing HD cameras via the cloud' and just won the best wireless video surveillance NPS award. So is it a wireless SaaS offering? Actually, it is neither. In this note, we examine the features, pricing and competitive positioning to OnSSI's competitors.
[UPDATE 2013: OnSSI has renamed 'X' to be the milquetoast 'Media Server']
Ocularis X is a transcoding service that improves remote viewing performance by reducing bandwidth needs while maintaining high frame rates. [For those not familiar, transcoding is a conversion of one encoding - like a HD surveillance camera feed - into another encoding - such as CIF.]
What is special about Ocularis X's transcoding is that it not only transcodes streams but outputs a composite stream that includes/embeds multiple video feeds. A practical example might help explain. The video below shows a 2 x 2 matrix in a web client:
While it looks like a normal playback of 4 video streams, actually only one stream is being played back and it is being sent over the public internet (in this case from NY to Vegas live).
Multiple surveillance camera feeds are sent to Ocularis X, which aggregates and transcodes them, sending out a single feed to the client. This significantly reduces the amount of bandwidth required to send video over the network, allowing the stream to maintain a higher frame rate without wasting resolution that could not be displayed on a multiplexed view.
To do this, Ocularis X must be installed as a separate application. Depending on the system size and load, it may need to run on its own dedicated machine as transcoding video can be resource intensive.
Dynamic zooming is one other special feature claimed by Ocularis X. As an operator selects a region to zoom in, Ocularis X can select that area and on the fly transcode it to a higher resolution to allow the operator to see more as they zoom down.
Currently, Ocularis X is only supported by the Ocularis web client up to a 3 x 3 matrix (no 4 x 4, etc.). OnSSI plans to add this to a mobile client in the short term.
Ocularis X is a free addition to Ocularis and is currently in controlled release, available to customers on request. It is expected to be in full production release in Summer 2012.
[Update: We have seen this live in production and the latency/delay of the transcoded video is in the tens of seconds. It is first and formost a usability issue but could also be a security issue for some users.]
VMS vendors use three main approaches to optimize remote viewing experiences:
- Stream switching: The VMS can switch from a higher resolution stream to a lower resolution stream depending on how many cameras are being displayed simultaneously. For instance, when displaying a single camera, it will be streamed at its max resolution (such as 1080p). However, in a 3 x 3 matrix, each camera might only be streamed at CIF. This should not make a discernible difference to the viewer because each pane in a 3 x 3 matrix is small. This is supported by many systems including Genetec, Milestone Corporate, Axis Camera Station, Avigilon H.264 streams (and others).
- Scalable Video Codec: When Avigilon markets the 'original' HDSM, this centers around JPEG2000, a codec that allows dynamic changes to streamed and displayed resolution. The main drawback is JPEG2000's overall storage costs - ~10x higher than H.264 and as such is not supported by many other vendors than Avigilon.
- Transcoding Individual Videos - A handful of vendors include transcoding functionalities into their recorders or VMS, that on the fly transcodes individual streams. For instance, Salient and 3xLogic actively market this.
While this looks to be a useful addition for OnSSI, other viable options exist that address the same core problem by different technical means. OnSSI has does a nice job of branding and building buzz but the marketing and (silly) award detract from the more fundamental and limited functionality that this service offers.
In terms of comparing performance of various remote viewing optimizations, we have not nor do we know anyone who has rigorously tested the different implementations. This is something we are looking into doing in the future.