VMware Cloud Foundation consists of a virtual management infrastructure designed to make the deployment of vSAN clusters and resources easier and more consistent. The Dell MX7000 chassis is a compute, storage and networking platform designed to make hardware deployment easier and more consistent.
A component of the MX7000 chassis infrastructure is SmartFabric. SmartFabric is integrated into the OS10 switch operating system and the Dell OME-Modular management software. SmartFabric provides the ability to manage the IOMs running OS10. SmartFabric provides a graphical utility for managing networks and uplinks and customizable templates. Templates are used to configure the server facing ports quickly and consistently on many switch ports. Templates are useful when deploying multiple cluster nodes which would all have the same network requirements. Templates provide value by programmatically identifying the switch ports to which a selected server is connected and configuring those ports based on the selected template. Server profiles are a combination of template and identity settings that are applied to a specific server or multiple servers. When the server template is deployed successfully, OME-M automatically creates and applies a server profile to that template. OME-M also allows you to manually create a server profile that you can apply to the designated compute sleds. Instead of deleting and recreating server templates, profiles can be used to deploy with modified attributes on server templates. A single profile can be applied to multiple server templates with modified attributes, or all attributes.
- Uplink creation and configuration
- VLT
- LACP link aggregation
- Spanning Tree configuration
- VLAN tagging (tagged and untagged)
- VLAN creation
- Server facing port configuration
SmartFabric does not provide capabilities such as Layer 3 routing configurations. Layer 3 routing is an important topic in the VCF deployment conversation. NSX-T, which is deployed as part of VCF, works best when it can exchange routing information with an upstream BGP-enabled router.
The MX7000 Scalable Fabric Architecture leverages the MX9116n Fabric Switching Engine (FSE) and the MX7116n Fabric Expansion Modules (FEM). In this design the Ethernet ports are configured on the FSE and the FEMs extend those ports to their compute resources. A single FSE can support up to nine FEMs and each FEM can connect to as many as eight hosts. Configuration of these server facing ports is implemented by the templates. SmartFabric tracks the ports to which a server is connected and applies the settings defined in a template. The implementation of a template can even track the location of a server if it is moved and dynamically configure the connected switch ports when the server is inserted into a different slot in the same multi chassis group connected to the same scalable fabric.
It is these advantages that have driven the decision to implement SmartFabric mode rather than a manually configured network environment. While manually configured IOMs provide more granular configuration capabilities, SmartFabric provides manageability, portability and scalability that match the manageability and rapid deployment capabilities of Cloud Foundation.
The following figure shows the connectivity between the data center and the MX7000 chassis. The leaf switches are manually configured to connect to a scalable fabric that can contain up to 160 server facing ports. The connection between the top of rack switches and the SmartFabric is a layer 2 connection configured as either an LACP aggregated connection or a Spanning Tree managed connection. VLT is used by SmartFabric to create a robust path when an LACP aggregated connection is deployed.
In this example the leaf switches are BGP-enabled, layer 3 devices. While the leaf switches are configured manually, they need only be configured once with minor modifications as the environment expands. The SmartFabric managed IOMs connect to the leaf switches. SmartFabric templates are used to assign network profiles to the assigned servers. Templates can be assigned to follow a specific server that may get moved from one slot to another or from one chassis to another (all chassis and slots must belong to the scalable fabric).
Using this architecture allows expansion of an 8-node cluster to a 16-node cluster by reusing the template used to deploy the first eight nodes. Templates can be cloned, and the cloned templates can be modified and saved as new templates. Using templates to deploy hosts provides fast deployment and consistent results.