NIC Teaming

welcome to the i.t free training video on NIC teaming NIC teaming allows multiple network cards to be combined together to form one network card this video will look at what can be achieved using NIC teaming and some of the limitations the video at times will refer to options that are available in Windows Server 2012 r2 however the theory applies to other operating systems and network hardware NIC teaming is the process of combining multiple network cards together to form one network card this is done for performance and redundancy reasons although Windows Server refers to this as NIC teaming other vendors may refer to it as bonding balancing or aggregation regardless of which modern operating system you use the result is the same that is multiple physical network cards are combined together using a teaming solution the team in solution will then present these multiple network cards to the operating system as the one virtual network adapter you then would configure this virtual adapter just like a standard physical network card if you consider this example in this case the server has three network cards in it if all three network cards are added to a network team solution the server can distribute the outgoing network packets between the three network cards if one of the network cards were to fail the team solution would then distribute the load between the remaining two network cards by using a system like this any application running on the server using networking does not need to know how the teaming solution is configured and which network cards are currently being used in that teaming solution the teaming solution handles additional network cards that may be added or removed for example a network card is disconnected from the network for this reason the administrator never needs to worry about configuring the physical network card other than configuring it to be part of the teaming solution when creating a nic team you will need to decide if you want to use all the available network adapters at the same time or if you want one or more network cards as standbys there are a number of different reasons that may affect your choice all things considered equal using multiple network adapters together will improve the max throughput of your server in practice this may not happen as much as you would expect and the reasons for this I will look at later in the video in the left example two network cards have been combined to improve throughput in the right example notice that the server is connected to a fast network switch and a slow network switch in this case you would want to use the slow network switch as a backup in case the connection to the fast network switch were to fail this is one example of why you would want to use a standby network adapter but later in the video I will show you some more reasons why you would want to use standby network adapters rather than making all the network adapters active when creating a nic team in Windows Server 2012 there are two settings that can be configured for the NIC team the first is load balancing mode windows can support up to 32 network adapters in the one NIC team and the load balancing mode determines how traffic will be distributed between the different network cards in the neck team when it leaves the server there are three different load balancing modes that are supported each of these I will go into more detail in just a moment the first is address hash this mode essentially uses attributes of the network traffic like the IP address and port to determine which network adapter will be used given the same input values the same network adapter will always be used if you are using virtual machines the second setting hyper-v switch port can be used and can control incoming and outgoing traffic it does this by routing each virtual machine through the same network adapter each time the third setting dynamic this is a new setting that was added in Windows Server 2012 our to this setting combines the features of the previous two balancing options and can change network streams from one network card to another as required the next setting is the teaming mode some network switches have support for load balancing if your network switch does not have that ability or you do not want to use it the first mode switch independent should be used this mode does not require any hardware support on the switch the next two modes are static teaming and LACP both these modes require your network switch to support the required load balancing protocol if you're prepared to configure your network switches these options allow better utilization of your network cards I will now have a closer look at these settings starting with the load balancing option followed by the teaming mode options the first load balancing mode that I will look at is address hash this mode creates a hash value from the values like the MAC IP address import used in the network connection this essentially means that given the same input values the same network adapter will always be used let's consider the example of a server with four network cards communicating with four desktops on the network if you consider the server is sending web traffic to four computers on the network the traffic for each desktop will use the same network adapter each time in this case you will notice that one network card was not used at all the more traffic that the server sends the more the traffic should even itself between all the network cards however this only works best if all the network streams from the server have about the same amount of data in them if you have a network stream that has a lot of data in it compared to the other network streams this network stream will always use the same network adapter this means the data leaving the server will not be well balanced between the available network if I were to have a look at incoming traffic all packets arrive using the same network card using address hash the server has no control over incoming packets and thus if you use this load balancing mode this needs to be taken into account you can start to see why some administrators may decide to use standby network cards rather than combine multiple network cards together this is because if there is no advantage to incoming data a network adapter may be set up as a standby network card for redundancy reasons rather than for performance reasons it all depends on how much traffic the server is receiving to address these issues the next algorithm hyper-v switch port can be used to some degree hyper-v switch port provides load balancing for income in traffic unfortunately it is aimed at virtual machines using hyper-v so it is limited to where it can be used essentially each virtual machine has its traffic routed out the same network card this is for incoming and outgoing traffic let's consider an example where the server has two network cards each virtual machine will use the same network card for incoming and outgoing traffic this is achievable as each virtual machine has its own MAC address hyper-v switch port uses this MAC address to determine which network card to send the traffic out on clients on the network use this MAC address to communicate back to the server and thus incoming traffic is also routed back through the same network adapter notice however that in this example the bottom network card is being utilized a lot more than the top network card if you have a number of bandwidth heavy virtual machines that are allocated to the same network adapter this will reduce the effective throughput of the NIC team the more virtual machines you have the more traffic should balance out assuming that they have similar network needs before I look at the last load-balancing option I first want to answer a question that some of you may have been wondering and that is why can't outgoing traffic be load balanced out the available network adapters on a packet to packet basis to understand why consider this network administrator's will attempt to have a single flow of traffic through the network this is done to reduce the amount of reassembly when the traffic reaches its destination let's consider an example where there are four network adapters connected to the network each network adapter transmits one packet to the destination notice that when the packets arrive they arrive in the reverse order having packets arrive out of order means the destination device needs to hold these packets in memory until the first packet arrives once this packet arrives it can process the other packets this increases the amount of RAM and processing power required on the device although at times it is possible to prevent packets arriving out of order network administrator's will do their best to prevent this from happening you can see why load balancing algorithms cannot simply divide up the outgoing packets amongst the available network cards in the NIC team the last load balancing algorithm dynamic attempts to address some of these issues by combining the best features of address hash and hyper-v switch port this algorithm is only available on Windows Server 2012 r2 the difference between this algorithm and the others is that it can dynamically change which network card a network stream is using let's consider an example the server has three network streams in use on the server currently the network traffic is about even even though one network card has two network streams and the second one has one they are sending about the same amount of packets so there is good load balancing occurring if the top network stream was suddenly to become lotmore utilized the traffic is no longer divided equally between the network adapters although dynamic is not a true load balancing solution it can make changes as required to help even out the load it does this by looking for what are referred to as flow let's if you consider me speaking out loud at times I will pause like I just did and just it again network traffic has similar breaks in it called flow let's either a brief break in transmission or some sort of logical change to the transmission when windows detects one of these it will change the network stream to use another network card this helps prevent packets arriving out of order at the destination notice that the network traffic is now more evenly distributed traffic can only be redistributed based on network streams if you have one string sending 90% of the traffic from a server you will still have 90% of traffic traveling over the one network card inbound traffic works the same as hyper-v switch port if you are using virtual machines with different MAC addresses on the server traffic can be routed using the same network adapter just like it could with hyper-v port if you want better load balancing control over incoming and outgoing packets you will need to consider a hardware based solution this brings me to the next topic teaming modes the first teaming mode is switch independent this teaming mode does not require any configuration of your network switches or any special hardware support it can use one or multiple switches that is you can plug your server network cards into the one switch or into individual switches or anything in between this is the simplest teaming mode to configure since it requires no special hardware configuration if you are new to nic teaming you should try this mode first the problem with this mode is that true load balancing cannot be achieved as I have highlighted already in this video in order to do this you need hardware support the first hardware team mode is static teaming this protocol uses the network protocol 802 3 ad to combine multiple ports on the switch together when this protocol is used the switch ports need to be configured to use this protocol and the server plugged into the correct ports on the switch depending on which switch you are using this may or may not need to be configured sometimes the port may be pre-configured in other times it may need to be enabled on the switch before it can be used since the ports on the switch need to be configured all network adapters on that server in that neck team need to be plugged into the same switch you cannot use multiple switches like the previous teaming mode of switch independent in some cases some vendors may allow you to combine multiple switches together to form a single virtual switch this feature will generally only be found on high-end network equipment you would need to check the specifications of the switch to see if it can do this the advantage of static teaming is that it allows incoming packets to be load balanced as shown half of the incoming packet data is going to one network card and half to the other the switch is able to control the incoming packet flow using both network cards in the real world how balanced the incoming traffic is depends on the algorithm used in the switch and the switches effectiveness this is something that Windows has no control over the last teaming mode is LACP which stands for link aggregation control protocol LACP uses the 802 1ax protocol which is effectively a newer version of the protocol used for static teaming like static teaming this creates a single incoming pipe for multiple ports just like static teaming the network switch needs to be configured LACP is harder to configure and get working than static teaming so is generally only used in high availability enterprise environments you will also tend to find less switches supported when compared with static teaming well that covers it for all the theory for NIC teaming in the next video I will perform a demonstration on Windows Server 2012 r2 to show how you configure NIC teaming I hope you found this video useful and I hope to see you in the next one thanks and see you next time

Channel: itfreetraining

Views: 167,563

Keywords: NIC Teaming, ITFreeTraining

Id: 8qRMwpgLJek

Length: 16min 3sec (963 seconds)

Published: Mon Apr 07 2014