Help us improve your experience.

Let us know what you think.

Do you have time for a two-minute survey?

 
 

Solution Architecture

The 3-Stage Fabric with Apstra Data Center Director is an EVPN-VXLAN-based validated design based upon the ERB network architecture. Using an ERB network architecture provides the design increased resilience by assigning specific functions to each device role and ensuring that each device role can be scaled independently of the others. Each network switch participating in the design must occupy one of the three roles:

  • Server Leaf Switches

The leaf switch focuses on learning and advertising the local MAC Addresses to other remote switches through the BGP EVPN control plane. This means leaf switches can discover all the “remote” hosts without flooding the overlay with ARP/ND requests.

  • Border Leaf Switches

Although a border leaf can function as a server leaf switch, it can also act as a gateway to external networks and hence require DCI features. DCI features include connecting to network overlays such as VMware NSX-T, MACSEC, deep buffers, and so on.

  • Spine Switches

The spine switch only performs IP forwarding and relaying of routes to all server and border leaf switches. As a result, spine switches in ERB network architectures are referred to as lean spines.

  • Access Switches

The access switches are connected to the leaf switches and form a Layer 2 aggregation tier to provide additional port density for end-hosts. This dedicated access layer offers flexible host connectivity, accommodating specific server-side port constraints where 1GbE or 2.5GbE copper interfaces are required.

The use of an ERB network architecture and the associated switch roles not only simplifies the data center design but also provides flexibility at the leaf layer so that new leaf switches can be introduced as traffic throughput increases. Another aspect of this design is the use of non-modular switches, such as the 1U QFX5130-32CD, which can perform high throughput functions at the leaf layer.

To summarize, the ERB network architecture, which underlies the 3-stage fabric with Apstra Data Center Director, can be thought of as a distributed chassis. In an ERB network, leaf switches are roughly analogous to a “line card” in a traditional modular chassis, while the lean spine means the network fabric is more flexible and resilient than a single modular chassis switch. This creates a network more capable and flexible than a traditional modular chassis-based switch, without requiring the purchase or maintenance of a modular chassis-based switch for most enterprise data center scenarios.

For those data centers looking for scale that can only be achieved with chassis-based switches, the HPE Juniper Networking Validation process does take this into account, validating modular chassis switch combinations in ERB network roles. The result is validated network fabrics that can scale from the needs of individual racks up to serving entire data centers and beyond.

Table 1: Supporting Devices for 3-stage EVPN/VXLAN Fabric (ERB) depicts the hardware in various roles such as spine, leaf, and border leaf. This validated design will walk through the high-level steps required to configure a 3-stage Data center, with QFX5220-32CD switches in the spine role, QFX5130-32CD switches in the border leaf role, and QFX5120-48Y switches in the server leaf role. These switches in these roles are considered the baseline design of this validated design, though other switches are qualified for these roles, as documented in the following.

HPE Juniper Networking Hardware and Software Components

This section describes the HPE Juniper Networking products and software versions for this solution.

The design documented in this validated design is considered the baseline representation for the validated solution. As part of a complete solutions suite, we routinely swap hardware devices with other models during iterative use case testing. Each switch platform validated in this document goes through the same rigorous role-based testing using specified versions of Junos OS and Apstra Data Center Director management software.

Table 1: Supporting Devices for 3-stage EVPN-VXLAN Fabric (ERB)

Server Leaf Switches Border Leaf Switches Spine Access Switch
QFX5120-48Y/
QFX5120-48YM/
QFX5120-32C
QFX5120-48Y/
QFX5120-48YM/
QFX5120-32C
QFX5120-48Y/
QFX5120-48YM/
QFX5120-32C
EX4400-48MP
QFX5130-48C/
QFX5130-48CM
QFX5130-48C/
QFX5130-48CM
QFX5130-48C/
QFX5130-48CM
EX4100-48MP
QFX5140-24CD QFX5140-24CD QFX5140-24CD
QFX5700/
QFX5700E
QFX5130-32CD/
QFX5130E-32CD
QFX5220-32CD/
QFX5220-128C
QFX5200-32C
QFX5230-64CD

Note: The other model variations of the tested device in the hardware series should also work. For instance, QFX5120-48YM covers all the rest of the variations such as QFX5120-48Y since the same chipset is used. However, there are some exceptions such as QFX5130-48C and QFX5130-32CD. Please contact your HPE Juniper Networking account representative for more information.

Table 2: HPE Juniper Networking Software and Version

HPE Juniper Networking Products Software or Image Version
Apstra Data Center Director 6.2
Junos OS Version (QFX) 25.2R2
Junos OS Evolved Version 25.2X100-D30
Junos OS Version (EX) 23.4R2-S8

Validated Functionality

Following lists the features for which 3-Stage Fabric is validated.

Table 3: Apstra Data Center Director Intent Based DC Reference

CATEGORY FEATURES HPE Juniper Networking Standard Document Link
EVPN-VXLAN with IPv6 Underlay Apstra 6.2 will provide IPv6 assignment and BGP configuration; hence BGP unnumbered is not included as Apstra manages BGP auto-discovery/unnumbering EVPN-VXLAN with an IPv6 Underlay
In case Apstra is not used then for BGP unnumbering see section Appendix A – BGP Unnumbered with IPv6
IRB type Asymmetric IRB with IPv6 underlay (At the time of writing this validated design, symmetric IRB parity across all Junos OS and Junos OS Evolved platforms is not available in 25.2 for IPv6 underlay) Symmetric Integrated Routing and Bridging with EVPN Type 2 Routes in EVPN-VXLAN Fabrics
EVPN Instance Type (MAC-VRF) MAC-VRF is an Ethernet VPN (EVPN) instance type that provides layer 2 multitenancy by creating dedicated MAC address forwarding tables per tenant VLAN-aware as Apstra provisions both VLAN-AWARE and VLAN-Bundle.

MAC-VRF is also configured for VLAN isolation to support overlapping VLAN IDs across tenants.
MAC-VRF Routing Instance Type Overview
External Gateway Routing Intra-subnet, Inter-subnet, Inter-VRF Configure External Router and Inter-VRF Routing
Underlay EBGP Underlay IPv6 eBGP Peering session Configure Interfaces and EBGP as the Routing Protocol in the IPv6 Fabric Underlay
Overlay EBGP Overlay IPv6 eBGP Peering session Configure EBGP for IPv6 Overlay Peering
Load Balancing ECMP with EBGP session Understanding BGP Multipath
Single Homing and Multi-homing ESI-LAG for single and Multi-homed (Active-Active). Understanding EVPN Multihoming Concepts
Bi-directional Forwarding Detection (BFD) liveness detection Apstra applies on EBGP underlay default BFD liveness detection with minimum-interval of 1000ms with multiplier of 3 before declaring BFD session down. BFD-Liveness-Detection (BFD) for BGP
Understanding BFD for BGP
Type2 Type 5 coexistence Type2 and Type5 routes for same destination and the preference algorithm applied by switch. Remote routes use Type5 and local routes are Type2. EVPN Type 2 and Type 5 Route Coexistence Implementation
Overlapping VLANs with Apstra Native Tag Override Apstra natively support vlan translation. To support multi-tenancy and overlapping VLANs, this architecture utilizes Junos Implicit VLAN Normalization orchestrated natively with the VLAN ID Override feature in Apstra Connectivity Templates. Overlapping VLAN Support Using Implicit VLAN Normalization in EVPN-VXLAN Networks

Table 4: Apstra Configlets for Customized Configuration

CATEGORY FEATURES HPE Juniper Networking Standard Document Link
Fast Convergence BFD underlay sub-second
Security Storm control and firewall filters for RE protection Control plane and data protection
DHCPv6 Apstra supports DHCP config natively, however additional statements were applied using configlet DHCP Relay
Junos DHCP user guide
Telemetry SFLOW (Apstra Flow) sFlow Technology Overview
sFlow with Apstra Flow

Table 5: Apstra Probes

CATEGORY FEATURES HPE Juniper Networking Standard Document Link
Telemetry Apstra Data Center Director provides a comprehensive set of predefined Intent-Based Analytics (IBA) probes designed to monitor the health, traffic, and configuration of data center fabrics. Probes