How do I setup my dual LAN connections for a 2GB connection?

AMD_RULES

2[H]4U
Joined
Mar 26, 2007
Messages
3,010
I recently converted my desktop into a server and I'd like to setup the two gigabit lan connections to make a 2gig connection. How do I do this? Operating System is Windows Server 2008 R2 Enterprise x64.

Thanks and sorry for my n00bieness lol
 
it's called teaming. you need a switch/router that supports it. but that is the term you should be googling, "teaming lan"
 
You need a switch and NICs that support LACD or "link aggregation". Teaming is another term for it. You're going to need a managed gigabit switch to do it.
 
it's called teaming. you need a switch/router that supports it. but that is the term you should be googling, "teaming lan"
Alright thanks!
just curious what's on your LAN that requires that speed?
Figured I'd take advantage of both NICs... don't necessarily need that much speed.
You need a switch and NICs that support LACD or "link aggregation". Teaming is another term for it. You're going to need a managed gigabit switch to do it.

Alright thanks for the info! Guess it's time to get a real switch :D
 
Teaming often means a "failover" team so you only actually use one at a team. You want to do link aggregation, or a port-channel, or an Etherchannel (old Cisco name)...or a trunk (in incorrect HP terms). But here is the thing. I've explained this a few times on here.

There is no standard way for a single client to talk to a single server and use both links. Link aggregation uses a hashing method... Hash based on source MAC. Hash based on IP. Hash based on port..all sorts of ways, but the bottom line is that a single connection between your desktop to your server will never use more than 1Gb. The reason that there is no standard way to just do packet/frame round-robin load balancing is that you can't guarantee order of delivery like that which breaks some things.

If you have two desktops connecting to a server with two NICs you MIGHT get utilization across both NICs...it just depends on your hashing method and if you the mathematical algorithm works out so one connection hashes on NIC1 and the other hashes on NIC2. If they don't you may need to adjust IP or something (again, depends on hash type) to fix that.

That's why when I talk about link aggregation in network designs I never say "load balancing", I say "load distribution".
 
Teaming often means a "failover" team so you only actually use one at a team. You want to do link aggregation, or a port-channel, or an Etherchannel (old Cisco name)...or a trunk (in incorrect HP terms). But here is the thing. I've explained this a few times on here.

There is no standard way for a single client to talk to a single server and use both links. Link aggregation uses a hashing method... Hash based on source MAC. Hash based on IP. Hash based on port..all sorts of ways, but the bottom line is that a single connection between your desktop to your server will never use more than 1Gb. The reason that there is no standard way to just do packet/frame round-robin load balancing is that you can't guarantee order of delivery like that which breaks some things.

If you have two desktops connecting to a server with two NICs you MIGHT get utilization across both NICs...it just depends on your hashing method and if you the mathematical algorithm works out so one connection hashes on NIC1 and the other hashes on NIC2. If they don't you may need to adjust IP or something (again, depends on hash type) to fix that.

That's why when I talk about link aggregation in network designs I never say "load balancing", I say "load distribution".

What if you start two separate file copies?
 
What if you start two separate file copies?

Nope. You're using the same protocol and ports for the "conversation" to the server (conversation is a term I often us when talking about this). If you hash by source/destination IP..they are the same for both connections. Hash by MAC? Same for both connections. Port? Same. See the problem?
 
Nope. You're using the same protocol and ports for the "conversation" to the server (conversation is a term I often us when talking about this). If you hash by source/destination IP..they are the same for both connections. Hash by MAC? Same for both connections. Port? Same. See the problem?

Depending on protocol the connection might originate from from a random port.
 
Depending on protocol the connection might originate from from a random port.

Possibly..but most switches will hash based on MAC or IP. Some things will do it by source/dest IP and port...just depends and both sides have to agree. VMware, for example, is IP...no port info. So sometimes you can do things between switches and get better utilization but not from end-device to switch. It's annoying.
 
AMD_RULES, Actuallt my D-Link supports it I believe, I have a DGS-1224T 24 port gigabit.

Just thought Id tell u so u know u dont have to break the bank on this.
 
He has a Cisco SG300 and an HP switch that both do link aggregation so he's good there.
 
Possibly..but most switches will hash based on MAC or IP. Some things will do it by source/dest IP and port...just depends and both sides have to agree. VMware, for example, is IP...no port info. So sometimes you can do things between switches and get better utilization but not from end-device to switch. It's annoying.

Yep, for a workstation only benefit I see to teamed/bonded connections is fail over.
 
Back
Top