In this multi-part blog series I’m going to try to clear some fog around the networking parts needed in Windows Server 2016, especially when using Storage Spaces (Direct).
I’ll be touching multiple areas, things that were new to me (I’m not a network guy) such as DCB and and PFC, but also recommendations on the known RSS , Jumbo frames, VMQ and VMMQ.
Sometimes I refer to an external source because they explain it much better than I can 😉 The goal is to have you understand the whole spectrum of things needed when your done with the last part of this series.
Hyper-converged infrastructures are forcing us to change the way we configure networking, especially with technologies such as RDMA (Remote Direct Memory Access) coming in to play which allows us to enhance network speeds drastically. Windows Server 2016 supports all features needed to manage your network and traffic flows to ensure your systems handle traffic efficiently.
Remote Direct Memory Access & SMB Direct
RDMA is a networking technology that provides high-throughput, low-latency communication that minimizes CPU usage. Windows Server currently supports the following RDMA technologies:
- Infiniband (IB)
- Internet Wide Area RDMA Protocol (iWARP)
- RDMA over Converged Ethernet (RoCE)
RDMA is optimizing the network by placing data directly into the host destination memory thus bypassing the memory and CPU bus. This way the overhead for network traffic is minimal and helps a lot latency wise. With the constant evolving network hardware, which makes higher speeds possible, we want to skip the CPU and memory in the server because it could eventually cause scalability issues. In other words, the CPU and memory cannot keep up with the network capabilities.
SMB Direct is an extension of the SMB technology by Microsoft used for file operations. The Direct part implies the use of various high speed Remote Direct Memory Access (RDMA) methods to transfer large amounts of data with little CPU intervention. In Windows Server 2012 and later, the Network Direct Kernel Provider Interface (NDKPI) enables the SMB server and client to use remote direct memory access (RDMA) functionality that is provided by the hardware vendors.
I highly recommend RDMA capable NICs when doing Storage Spaces Direct clusters.
Data-Center Bridging (DCB) is an extension to the Ethernet protocol that makes dedicated traffic flows possible in a converged network scenario. DCB distinguishes traffic flows by tagging the traffic with a specific value (0-7) called a “CoS” value which stands for Class of Service. CoS values can also be referred to as “Priority” or “Tag”. Note that every node in the network (switches and servers) needs to have DCB enabled and configured consistently for DCB to work.
Enabling DCB also brings support for lossless transmission of network packets, with latency sensitive network traffic such as Storage Spaces Direct or Storage Replica traffic over the SMB protocol we would like to have a lossless network; making sure that every packet that has been send actually reaches “the other side”. Normally, lost data packets are not a big issue as they would get retransmitted by TCP/IP but with storage traffic, lost data packets are killing for your performance as they introduce I/O delays which is unacceptable. DCB uses the CoS values to specify which value needs to be lossless transmitted through the network, typically network switches support up to three lossless enabled CoS values.
Priority-based Flow Control
In standard Ethernet we have the ability to pause traffic when the receive buffers are getting full. The downside of ethernet pause is that it will pause all traffic on the link. As the name already gives it away, Priority-based Flow Control (PFC) can pause the traffic per flow based on that specific Priority, in other words; PFC creates Pause Frames based on a traffic CoS value. This way we can manage flow control selectively for the traffic that requires it, such as storage traffic, without impacting other traffic on the link.
In the following example we’re using 3 CoS values, as you can see only the storage traffic is paused when the buffers are getting full while cluster traffic just keeps going.
This makes sure the server is not unreachable for the other servers in the cluster which would cause a failover of all the cluster roles.
The receive buffers could get full when I/O cannot be done (fast enough) to disks for example.
Enhanced Transmission Selection
With DCB in-place the traffic flows are nicely separated from each other and can pause independently because of PFC but, PFC does not provide any Quality-of-Service (QoS). If your servers are able to fully utilize the full network pipe with only storage traffic, other traffic such as cluster heartbeat or tenant traffic may come in jeopardy.
The purpose of Enhanced Transmission Selection (ETS) is to allocate bandwidth based on the different priority settings of the traffic flows, this way the network components share the same physical pipe but ETS makes sure that everyone gets the share of the pipe specified and prevent the “noisy neighbor” effect.
Note that ETS only makes sure that sufficient bandwidth is guaranteed on egress traffic.
Data Center Bridging Exchange Protocol
This protocol is better known as DCBX as in also an extension on the DCB protocol, where the “X” stands for eXchange.
DCBX can be used to share information about the DCB settings between peers (switches and servers) to ensure you have a consistent configuration on your network.
In other words, you configure DCB on the switch, enable DCBX on the switches and servers and the servers will receive the DCB information to configure the network cards.
Microsoft recommends to disable DCBX, this probably has to do with vendors implementing it differently and issues with drivers etc etc.
Configuring Windows to use DCB / PFC / ETS
Now that we got the formalities on terminology out of the way, I can finally start with showing how this “network stuff” works in combination with Windows Server.
DCB is already included out-of-the-box in Windows Server, but is not installed by default. The only thing we need to do is to enable the feature using Server Manager or our beloved Powershell:
Installing the DCB feature gets us some new Powershell cmdlets to work with, which we will touch in the next part where we will configure DCB on the Windows side.
Continue to Part 2: Configuring DCB in Windows >>
Thank you for reading my blog.
If you have any questions or feedback, leave a comment or drop me an email.
Darryl van der Peijl