Skip to main content
S
someonemobile
Participant
April 16, 2010
Question

What no of concurrent connections can each server serv?

  • April 16, 2010
  • 1 reply
  • 2458 views

Ok we're a startup and currently thinking about whether to use flash media streaming server or red5, or progressive streaming with lighttpd.

In order to make a decision I would like to know how many concurrent streams can flash media server server stream until one has to purchase another license. Do any of you guys have some numbers.

We will be streaming mp3 at low bit rates, service will be hosted within AWS EC2 (with whatever instances are available there). So if you can post some performance specs that would be great....

Oddly enough Adobe Sales does not even know the answer... I have heard rumors that high number of concurrent connections is not a really a feature that this product is developed towards.

Thanks for you answers.

    This topic has been closed for replies.

    1 reply

    A
    Anonymous
    April 18, 2010

    There is not limit to concurrent connections or streams as far a the license goes.... the FMS license is per machine, so you just need a license for each machine you install on.

    The number of clients/streams you can serve really depends on your hardware and network capacity. I've had upwards of 1500 streams running on a single box with a 1Gbps interface and a lot of hardware power (12 cores / 16GB ram). All of my experience has been with windows enterprise, YMMV on other operating systems and environments.

    Also, read the docs on configuring core processes... spreading out the load across multiple FMS core processes will give you greater capacity in some cases.

    All that said, EC2 isn't a great envorinment for FMS.... for a few reasons:

    1. FMS runs a lot of clock based operations (particularly when it comes to audio and video data), and you tend to get a lot of jitter in a virtualized environment. Stuttery video, loss of a/v sync, and loss of audio in streams are not uncommon in these situations.

    2. On EC2, you're sharing hardware. Although processor cores and memory are reserved for each instance, you're in competition for disk time and network resources, so you can't count on having bandwidth available when you need it.

    3. I don't know how FMS licensing plays against a virtualized environment... but if you have two VM's running on separate machines on the same FMS license, I suspect you'd be at conflict with the EULA.

    S
    someonemobileAuthor
    Participant
    April 19, 2010

    Dear JayCharles

    Wow I TRULY APPRECIATE your answer, and I am glad you pointed out the issues with EC2. I did not know that. I assumed because aws is using Adobe for their cloudfront product it would work, good to know that there are issues. I think in that case wowza for ec2 is probably better suited for our application with streaming straight out of s3 (which is where our files are stored). You just saved me a ton of money and headaches, THANKS SO MUCH!!!!!

    A
    Anonymous
    April 19, 2010

    I'd do some more research before commiting to EC2/S3 at all. I'd also confirm that comment about Wowza streaming from S3... I've never worked with it, but if you're streaming, chances are S3 is just storage, and you're actually streaming from EC2 (I don't think you can run an application on S3... but maybe I'm wrong there)

    The trouble is, regardless of the software, you're still in competition for bandwidth. There's only so much capacity at the network interface of each shared server, and if you happen to land on a busy server, you may find there isn't enough to go around. For stateless applications like HTTP request/response, that's fine, but for apps that require statefull client/server connections, bandwidth and latency issues can prove catastrophic.

    The biggest problem I have with running a streaming server on EC2 (virtualization issues aside), is that they won't disclose hardware or network specs. So, while you can be sure you'll have the memory and processor allocation, it's a crapshoot when it comes to disk time and bandwidth, as you have no real idea how many other instances you're sharing a server with, or what those other instances are doing in terms of bandwidth and disk usage.

    While I don't know any of the specs on the cloudfront service, I suspect those servers are load balanced in such a way that bandwidth is accounted for. In the general population EC2 servers, that's not the case.

    My advice... really research this before you make your deployment decisions. I've come to find that while scaling out is a bit more challenging, dedicated hardware and bandwidth isn't all that much more expensive than EC2, and when you go dedicated, you can actually count on your resources being available to you.

    Terms & ConditionsAccessibility statement
    Using the community Terms of use Privacy Cookie preferences Do not sell or share my personal information ad choicesAdChoices