I work for a VAR and I have a customer who is interested in HCI but one of their requirements is segment routing.
Segment routing is a forwarding paradigm that provides source routing, which means that the source can define the path that the packet will take. Segment routing still uses MPLS to forward the packets, but the labels are carried by an IGP. In segment routing, every node has a unique identifier called the node SID.
I could not find anything around segment routing and HCI solutions so I was wondering if you were aware of any HCI solutions that supported segment routing?
Thanks! I appreciate the help.
An HCI solution with VMware vSAN does not have all the network virtualization features like VMware NSX, it would be necessary to analyze the scope of the service you want to have and combine it with the features offered by the VMware NSX solution.
I think that the question of if HCI supports SR is not what you should be asking because it really does not, but SR supports HCI. Let me explain. SR is a network function and HCI is a computer function. Therefore, HCI can use the function of SR to provide faster data transfers for Cloud computing. The HCI virtual systems can use the efficiency of a SR network to deliver data where it needs to go by preselecting the desired path. This could be used to provide remote virtual computing if you can imaging a HCI system functioning on a SR network.
It depends. In segment routing node SID can be either a NODE or broadcast segment gateway. If HCI cluster is behind gateway in the L2 switching domain - its easy config and don't need to explain If a few parts are behind a different SID GW - you just have to be sure that all traffic from HCI cluster nodes allowed and they can communicate between in acceptable latency budget (up to 5 ms, faster is better). But you need to know that for storage function latency is critical parameters and if it is more than 2000 ns storage performance can be unacceptable . Multi-cluster configuration can be a case for that dispersed configuration (HQ-ROBO config or MultiCluster in mixed Mine-DR roles).
There is no explicit support in HCI for segment routing- because it is not something that a virtualization platform needs to handle. Segment routing is a set of IP packet extensions that do tag flows and provide a way to handle traffic segmentation;
So, if you need to use HCI for this, you simply need to run an operating system that supports it.
You could use Nutanix with flow for microsegmentation!
Yes, our BPaaS technology supports hyperconvergence on Dell VxRail.
Hi Mike, I have consulted with an individual that is very knowledgeable and works in this industry. His answer to me was that he was not aware of any HCI solution that has this capability. Sorry.
I think the question is not relevant to the HCI in general, MPLS or IGP is not applicable by HCI but large networks, HCI is going to converge your computer and storage and simplify the management aspect. In every HCI offerings, every node in a single cluster should be in the single broadcast domain, so routing is not applicable here. I would suggest if you ask for a specific use case or bussiness case scenario people will be able to understand and guide properly.
Does VMware support segment routing? If so, HCI supports it. This isn't a storage issue, it's a VM/network issue so support will be purely upon the hypervisor.
If you need to know if AHV or Hyper-V support it, I sadly don't have any resources.
I would like to point to a combination of cisco's solutions, HypeFlex, ACI, etc...
www.cisco.com
You might want to check the advanced networking and security of Nutanix called Nutanix Flow, and Sangfor's NGAF.
You might want to take a look at Sunlight.io's HCI stack. Sunlight is a very-high performance HCI stack, designed to deliver near-bare-metal performance for fast networking and NVMe storage.
Segment routing (SR) is fully supported on Sunlight. SR is built as an IPv6 header extension to enable the source host to send a forwarding list of Layer 3 endpoints to guide a packet through the network. The first hop is handled as a normal ethernet + IPv6 packet and will be forwarded to the first hop SR-enabled destination. On the Sunlight cluster you would configure a normal direct connected ethernet segment or a VLAN network segment and attach the virtual machine VIF to that network. It would be advised to configure an external IP address allocation so that the IPv6 address can be managed via an external DHCP service. Within the VM guest (assumed to be some flavour of Linux) it would be necessary to configure a custom kernel with SR support.
[Note - I work for Sunlight.io]