I am looking after this network with around 40 clients on one Windows 2008 R2 server. These clients are connected to server via two gigabit switches (24-port dumb switches). Over the next few months the number of clients are going to increase. Let us say we want to plan for up to 100 clients.

The network traffic is already high and sometimes it gets really clogged. My goal is to double the network bandwidth in near future and quadruple the same when we procure a new server hardware.

Obviously, this would require multiple NIC or an NIC card with multiple ethernet ports. These are my thoughts about options/limitations:

1. I read that Windows Server 2012 has built-in NIC teaming feature. However, this is not at all viable as the licensing cost (server + CAL) would be too high. Besides, this cannot be a reason to upgrade the server software at all.

2. Go with multiple NIC without LACP/teaming. Each NIC will have to be on a separate subnet because Windows Server does not play nicely when two of its NICs are on same subnet. Each NIC will connect to a different switch and accordingly those clients will be on different subnets. While this will increase the total bandwidth of the server, it will also add some complexities in network configuration (routing, DHCP, DNS, firewall, etc.) Can you convince me otherwise?

3. Discard the existing dumb switches and go for something better. Netgear offers "unmanaged plus" switches which feature LACP/LAG. Each 24-port switch (highest port configuration) costs around Rs. 20k (~US$350) in India so four such switches will set us back by around Rs. 80k (~US$1400). But I think it would still be better than replacing Windows Server itself. This can be sort of last resort.

4. I read that certain server-grade NIC cards (such as those from Intel) can offer driver-based teaming/aggregation even if the server OS or the switch don't support that feature. I have done some search and most results are about Linux and FreeNAS etc. so I am not sure if this works in Windows environment. If this is really a workable option, it is my best bet in near future. Can anyone share their experience with this kind of setup and suggest any specific NIC model to go with?

Other than that, I can try for some kind of network optimization. Currently all clients are connected to the switch directly. What I can do is introduce some small switches among a group of clients and connect those small switches to the main switches. But not sure if that will work for me because the traffic is mostly real and all of it involves the server. There isn't much going on between clients. The server is mainly being used as a file server.

I am looking after this network with around 40 clients on one Windows 2008 R2 server. These clients are connected to server via two gigabit switches (24-port dumb switches). Over the next few months the number of clients are going to increase. Let us say we want to plan for up to 100 clients.

The network traffic is already high and sometimes it gets really clogged. My goal is to double the network bandwidth in near future and quadruple the same when we procure a new server hardware.

Obviously, this would require multiple NIC or an NIC card with multiple ethernet ports. These are my thoughts about options/limitations:

1. I read that Windows Server 2012 has built-in NIC teaming feature. However, this is not at all viable as the licensing cost (server + CAL) would be too high. Besides, this cannot be a reason to upgrade the server software at all.

2. Go with multiple NIC without LACP/teaming. Each NIC will have to be on a separate subnet because Windows Server does not play nicely when two of its NICs are on same subnet. Each NIC will connect to a different switch and accordingly those clients will be on different subnets. While this will increase the total bandwidth of the server, it will also add some complexities in network configuration (routing, DHCP, DNS, firewall, etc.) Can you convince me otherwise?

3. Discard the existing dumb switches and go for something better. Netgear offers "unmanaged plus" switches which feature LACP/LAG. Each 24-port switch (highest port configuration) costs around Rs. 20k (~US$350) in India so four such switches will set us back by around Rs. 80k (~US$1400). But I think it would still be better than replacing Windows Server itself. This can be sort of last resort.

4. I read that certain server-grade NIC cards (such as those from Intel) can offer driver-based teaming/aggregation even if the server OS or the switch don't support that feature. I have done some search and most results are about Linux and FreeNAS etc. so I am not sure if this works in Windows environment. If this is really a workable option, it is my best bet in near future. Can anyone share their experience with this kind of setup and suggest any specific NIC model to go with?

Other than that, I can try for some kind of network optimization. Currently all clients are connected to the switch directly. What I can do is introduce some small switches among a group of clients and connect those small switches to the main switches. But not sure if that will work for me because the traffic is mostly real and all of it involves the server. There isn't much going on between clients. The server is mainly being used as a file server.

What are your thoughts? Am I missing something here?