VMS Multistreaming Comparison

Hi, Good report. As such I would also request you to advise the readers about impact/usage/nuiances of Multicasting to achieve more or less similar objectives.

Avinash, good questions on mutlicasting vs multistreaming (note: see our multicast for surveillance tutorial).

While the names are similar, the roles they play are different. As a rough rule of thumb:

• Lots of simultaneous viewers of a single stream - use multicast
• Lots of streams being viewed simultaneously on a single PC - use multistreaming

John, some VMS manufacturers are also using or developing transcoding to reduce the bandwidth for remote viewing and play back of HD cameras over limited bandwidth connections.

Have you road tested any and if so how well did they perform?

some complements in give during IP courses :

Save PC client CPU displaying 16 or 64 mini views.. use multistreaming

Use server video Analytic which is always working on low res : use multistreaming

Supervise several remote sites with low bandwidth ( 3G or Dsl) : use multistreaming

Record locally HD and only record alarms low res on Cloud : use multi streaming

If you wanna display 128 cameras in a security center with 128 x multicast based on the main FHD stream you will collapse a normal computer as it does use 64 x 6 Mbits or more for PTZ domes ....

Andrew, Gentec's WINK integration was reviewed in 2012, but requires an additional $250 per camera's transcoding license. The multistreaming feature is a more streamlined integrated approach in the varying VMSes with no additional hardware required,and no testing of external hardware was done as part of this report.

John, maybe also complete the trilogy by comparing the benefits of Multithreading? Exacq recently leveraged the power of multiple processors (starting at version 5.2 I believe) and we have noticed significant performance improvements at the client end. Providing of course one has multiple, multicore CPU's! One client of ours built their Exacqvision server on a Dell with 24 cores - slight overkill perhaps :) I'd be interested to see this feature evaluated and compared between brands.

This techniques was applied by Pelco in Endura since 2004... 9 years ago!

John, did any of the VMSes advertise upcoming support for cameras with H.264 SVC (Scalable Video Codec)? SVC allows extraction of lower resolutions from an incoming stream without having to decode the incoming stream, saving on CPU cycles. Transcoding using SVC could solve many of the issues, but features that add cost to the camera hardware, to save on the total system, are hard to market.
Some cameras support the simpler SVC-T (T for temporal), which means that you can extract lower frame rate streams, but that is less useful in your test cases.

Bengt : The camera U mentionned are from Vivotek which offers SVC -T in alternative to Jpeg/MP4/H264 in his camera and software (VAST). So very limited as it doesn't offer access to resolutions just frame drops to save CPU for display client PC . No other VMS editors according to what I know are supporting this. (at the time being) It requires also to use a processor able to generate this SVC (VVK like Axis manufactures its processors..)

SVC Frame/resolution is also expected in next generation set top boxes for TV over internet to allow scalable quality according to bandwitdtgh detection.

so if broadline applications use it you can be sure it will come also to CCTV players but don't expect it from Genetec/Milestones.Seetec/Exaq/Aimetis and so on till it's not in the majority of IP cameras...Their roadmap is already too thick for being in advance to the market

SVC vs Multistreaming

I'm not sure how long industry should wait till SVC will be here (I mean will be supported by most manufacturers). In fact in my opinion dualstraming is just recently started to be supported by most manufacturers. I fell like even 2 years ago it was a different picture. And still now most cameras share fps or pixel rate between first and second stream.

However let's imagine for a second that SVC is here now. What does it mean and should VMS use it? ( I also assume it’s supported by most popular video players and hardware and such ).

With SVC you still kind of get 2 streams( they just can be merged into 1 ).

Downside is such that you cannot control second stream fps in this case. It’s always equal to primary stream fps.

Some VMSes do analytics with second stream like motion detection and such. More fps you have – more intensive CPU usage for such kind of analytics.

So I assume VMS really like to have control under second stream fps. It’s not a case with SVC.

Yeah Marc, there are several cameras supporting SVC-T, but it is not very useful.
A while ago it was a way for camera vendors to say that SVC is supported in a tender checklist. What is needed is the 'real SVC' with resolution or quality scaling - which requires a more capable camera chip than the one Vivotek is using.

I agree on the delay in VMS development when it comes to using new video technology - it took some 3-4 years before H.264 was accepted as mainstream. With VMSes largely relying on PC technology and cheap storage to scale, I can see the reasons. How long will it take to before they support H.265, if ever? [shrug]

This should open up for end-to-end system companies that can optimize total system cost and ease-of-use performance through better networking optimization between cameras and the VMS. Onvif leaves to much on the table.

SVC is perfect for the camera to save power. Build only 1 stream and can save energy for analytics. On the VMS side it is less obvious as - as you said - VMS required specific fps /resolution for different purposes (record, View, Analytic) and Each SVC stream weight is the same interm of bandwith

On decentralized SMB architectures using Vaas or kind of camera companion it could be perfect.

Good discussion!

Comments on three key themes brought up: transcoding, SVC and hardware optimizations:

Transcoding: There are a number of VMS/NVRs that offer transcoding. Instead of requesting multiple streams, a single high resolution / high frame rate is ingested by the recorder and then multiple lower quality versions are generated / sent out. The main issue, and why lots of vendors do not support this, is processing load, can create severe performance issues.

That said, to Andrew's comments, we have not tested any out but we would like to in the next 6 months.

SVC: Like transcoding, SVC is an alternative to multistreaming. Instead of needing multiple streams, an SVC stream can easily be 'broken' up, sending only the bits or frames a specific client needs. Unlike transcoding, it is not CPU intense. The most well known SVC codec is JPEG, much ballyhooded and debated with Avigilon. There is an SVC for H264, however hardly anyone supports it in surveillance. If the cameras start supporting it and the VMSes add support for handling it, it could be useful. However, it is not clear if or when that will ever occur.

We had an excellent 47 comment discussion on the future of SVC in the old LinkedIn Group.

Hardware Optimizations: As Barry mentioned, some VMSes are optimizing use of more advanced hardware or hardware features - from multithreading to GPUs / using the client machines's graphic card. We have not tested this but it certainly would be interesting to try.

One good think that this discussion highlights is that there are many ways, beyond multistreaming, to handle megapixel streams. We will be evaluating more of these in the future to provide a more holistic treatment.

Looking forward to seeing that myself Murat. Axxonsoft has always been a good pioneer. Network Optix also does transcoding however we only do it in specific instances. An example being our Android app or web app where we use the VP8 codec (WebM open source project) as we prefer a much more open approach rather than using proprietary codecs.

The user can select which resolution they'd like to see through their browser or mobile devices and the server transcodes on the fly. Different approaches we take but both trying to push the envelope of video transport and delivery/display.

you should look at Geutebruck. they do a very good job of managing video remotely using multi-stream and their ICD/DLS technology.

We have tried to speak with Geutebruck, but they were reluctant to discuss technical details or product pricing. Since Geutebruck does not have wide availability nor a large user base, this makes it infeasible to cover them.

Hello Alex,

Milestone supports dual streaming in all versions of Milestone by taking a stream of the camera and dedicating one to live and one to recorded. Where as in XCorporate it is able to pull in ALL streams the camera supports and allowing the user to define each with greater control.

"Multi-streaming is only supported in XProtect Corporate. A range of the other products supports dual-streaming where you have one stream dedicated to live and one dedicated to recording."

Avigilon

"Digital zoom will not change to a higher resolution when digitally zooming in missing out on finer details."

This is false statment!