System Size

Forums Network Management ZeroShell System Size

  • This topic is empty.
Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
  • #41337

    Dear all,

    I planned to build enterprise network with load balancer to branches.
    The HQ only connect to 4 network provider with clear channel also.
    Each branch connect to 2 network from the available connection.
    I need the size of HQ system that can handle 50 branches with each branches use their own VLAN.
    Is it need Core2Duo or Xeon? How big the memory? 2 GB or 4 GB?
    The backplane of the network, is PCE-e enough for network I/O?

    I’m totally blind of it…



    This sounds unneccesarily complicated.

    Here’s a question:
    How many users and what is your total aggregate bandwidth at the main office?

    Any why do you need 4 Internet connections? It looks like you are trying to build an enterprise class system on pennies, which is most likely going to come crashing down hard unless keep someone monitoring it constantly.

    Also, are you sure you need to do VLAN, or is this more of a VPN question?



    The user around 2000 all over the enterprise.
    4 connection needed to give better assurance to the load balancing and fail-over.
    Because I have clear channel, I want all the link use bridge mode to ensure the low latency.
    With zeroshell, I can use bridge with transparent web cache and another features.



    I think you need to check the specifications of Zeroshell again.

    There is currently no web cache module in Zeroshell. Fulvio has it on the to-do list.

    Load balancing and fail-over are pretty new to this distro, as well. From what I have read, fail-over is going to be somewhat problematic, anyways, unless you manually change the gateway configuration when a problem occurs.

    Load balancing is a round-robin effect. It won’t help if people are connecting to specific IPs in the LAN sets – it is more of an outbound thing in the traditional sense. In other words, load-balancing is helpful to the users inside the network, not those connecting from outsite the network.

    Bridging does not ensure low latency. Technically, neither does clear-channel or any other wireless solution – it is the nature of the connection that you will incure a latency penalty. If you add ANY processing, be it bridge, router, virus scanner, web proxy to a connection, you can expect an increase in latency. If adding a device to your network lowers your latency it is due to the fact that the device / connection you replaced was overloaded, and thus you are relieving a problem in the network.

    If you have 2000 users, you need to consider some muscle in your hardware. This isn’t a ‘one box’ solution. If for no other reason, it should not be a one-box solution due to the fact that you are just putting everything to one point of failure again, which supposedly is the reason that you have 2 – 4 connections per site in the first place.

    Expect at the enterprise level to need to pay some money. It will be in something – maybe it is in your paycheck. If it is in your paycheck, I suggest reading a lot more on networking in general to make this solution work. While Zeroshell may be useful and even a decent solution for you, the understanding to do what you are asking is not going to be easily answered in a post on some forum. I would also be of the belief that, even if I could completely answer your question, I probably would not be willing to utilize that much time and effort for no compensation – and neither would any of the other people who probably qualify for a CSIE.

    Fulvio is a blessing to the community in that he takes time to make wonderful projects like this available to us.


    Dear abale,

    Thank you very much…
    I appreciate your input.
    I understand that single box solution is not the answer, and I will put redundant machine later.
    In this thread I just need the rough calculation concerning my network size.

    The zeroshell help me a lot in another project, so I prefer to use it in a bigger project and I need to understand the capability of zeroshell to handle kind of load. I understand that the load little bit high.

    I already planned to put Core2Quad processor, 4 GB memory and 10 x 1GE in PCI-e bus.
    With the kind of load, I just need confident feelings with my gear to handle the load…



    The hardware requirements depend on the volume of packets that the zeroshell router needs to handle.

    I’m just posting this for any other readers that wonder what equipment they need.

    I have about 120 devices on three vlans. Zeroshell routes packets between the vlans and out to the internet. I have a $600.00 2GB memory, 2.0 GHz dual core machine with 4x1GB ethernet on the motherboard, running zeroshell. It handles the load with ease:

    a sample output from ‘top’. Note that only 200mb is in use, and the load average is just about zero.

    top – 18:16:36 up 17 days, 41 min, 1 user, load average: 0.00, 0.00, 0.00

    Tasks: 68 total, 1 running, 67 sleeping, 0 stopped, 0 zombie

    Cpu(s): 0.0% user, 0.0% system, 0.0% nice, 100.0% idle, 0.0% IO-wait

    Mem: 2074028k total, 192600k used, 1881428k free, 29760k buffers

    Swap: 131064k total, 0k used, 131064k free, 48364k cached

    pinging on my local subnet, I get .04 ms ping times. Pinging across subnets with a route through zeroshell, I get .25 ms ping times.

Viewing 6 posts - 1 through 6 (of 6 total)
  • You must be logged in to reply to this topic.