Juniper QFX Series Switches

Juniper’s QFX5000 family has quietly become one of the most versatile data center switching platforms on the market. From classic 10/40GbE top-of-rack roles to modern 25/100/400GbE EVPN-VXLAN fabrics, the QFX5100 → QFX5200 generations give architects a clear, scalable upgrade path.

This guide walks through the evolution from QFX5100 to QFX5200, and explains when it makes sense to standardize on QFX5110, move to QFX5120 or QFX5130, or mix platforms inside the same fabric.

Table of Contents

• 1. QFX5100 → QFX5200: Generational Evolution at a Glance
• 2. QFX5100: The Workhorse 10/40GbE TOR
• 3. QFX5110: Transitional 10/40/100GbE Leaf & Border
• 4. QFX5120: 25GbE Leaf for Modern EVPN-VXLAN Fabrics
• 5. QFX5130: 100/400GbE Spine & High-Performance Leaf
• 6. QFX5200: High-Density 25/100GbE for Leaf/Spine at Scale
• 7. Side-by-Side Comparison Table
• 8. Recommended Fabric Design Patterns
• 9. SKU Shortlist: Best-Bet QFX Models to Standardize On
• 10. Upgrade Paths: When to Move from 5100/5110 to 5120/5130/5200
• 11. Related QFX & EX Resources

1. QFX5100 → QFX5200: Generational Evolution at a Glance

The simplest way to think about the QFX5000 evolution is:

• QFX5100 – 10/40GbE leaf/ToR with early VXLAN support; perfect for legacy server estates and smaller fabrics.
• QFX5110 – Transitional 10/40/100GbE with more scale; often used as leaf, border leaf, or small spine.
• QFX5120 – 25/100GbE EVPN-VXLAN leaf; the “sweet spot” for most new 2025 builds.
• QFX5130 – 100/400GbE flexible platform; ideal as spine or high-performance leaf for AI / HPC / storage.
• QFX5200 – High-density 25/100GbE; great for large leaf/spine fabrics where port count and consistency matter.

At a very high level, a modern Juniper EVPN fabric often looks like this:

         [ QFX5130 / QFX5200 Spine ]
                |      |      |
      ---------------------------------
      |               |               |
  [QFX5120]      [QFX5120]      [QFX5120]
  Leaf/ToR       Leaf/ToR       Leaf/ToR
      |               |               |
   25G/10G        25G/10G        25G/10G
   Servers        Storage        Appliances

 Legacy racks or remote pods:
      [QFX5100/QFX5110] providing 10G leaf roles

Older QFX5100 and QFX5110 platforms still have a place in this design, but the long-term direction is clearly toward QFX5120 and QFX5130/QFX5200 for new capex.

2. Juniper QFX5100 — Proven 10/40GbE Leaf with Long Service Life

The Juniper QFX5100 family is the original fixed-configuration data center workhorse in the QFX5000 line. It was designed for 10GbE server access with 40GbE uplinks, and it’s still a fantastic fit for:

• Legacy virtualization clusters running primarily at 10GbE
• Brownfield data centers where servers, storage, and optics are already standardized on 10/40GbE
• Price-sensitive deployments where you want EVPN/VXLAN capabilities without paying for 25GbE silicon

Typical QFX5100 Hardware Capabilities

• 48 x 10GbE SFP+ + 6 x 40GbE QSFP+ uplinks (e.g., QFX5100-48S-AFO)
• Line-rate L2/L3 switching with low microsecond-level latency
• Basic VXLAN and early EVPN support (depending on Junos version)
• Virtual Chassis or MC-LAG for simple redundancy

Best-Fit Use Cases for QFX5100

• Legacy 10G ToR leaf in brownfield DCs that won’t move to 25G soon
• Lab / test environments where you want full QFX feature richness at a lower price point
• Edge/aggregation in smaller environments where 40G uplinks are sufficient

If you are building a new 2025 fabric, the QFX5100 is usually not the primary choice—but if you already own them or can acquire high-quality pre-owned units, they can dramatically reduce capex for secondary pods.

3. Juniper QFX5110 — Transitional 10/40/100GbE with Border Roles

The QFX5110 series bridges the gap between the 10/40GbE world and modern 25/100GbE fabrics. Juniper positioned it as a flexible platform that can be a high-end leaf, a collapsed leaf/spine, or a border leaf for interconnects.

Key QFX5110 Characteristics

• Mix of 10GbE downlinks with 40/100GbE uplinks
• Enhanced EVPN-VXLAN scale and control-plane resources vs QFX5100
• Better suited for border, inter-pod, and edge roles

Where QFX5110 Still Makes Sense

• Border leaf / gateway between an older 10G rack domain and a newer 25/100G EVPN spine
• Small fabrics where a couple of QFX5110s can act as both leaf and spine
• MPLS or L3VPN-heavy environments that want QFX EVPN plus richer routing features

In short, QFX5110 is ideal when you need more horsepower and 100G connectivity than QFX5100, but you’re not ready to standardize the whole DC on 25/100G leafs.

4. Juniper QFX5120 — The 25GbE EVPN Leaf Sweet Spot

The QFX5120 family is where most modern Juniper data center designs land today. It’s a 25GbE server-access switch with 100GbE uplinks and full EVPN-VXLAN capabilities, making it the default choice for new leaf/ToR deployments.

For a deeper dive into model-by-model details, see BrightStar’s dedicated guide: Juniper QFX5120 Series Comparison Guide.

Representative Hardware (Leaf/ToR)

• 48 x 25GbE + 8 x 100GbE (for example, QFX5120-48Y-AFO2)
• Line-rate L2/L3 with rich buffer and QoS capabilities
• Deep EVPN-VXLAN feature set suitable for large multi-tenant fabrics

Why 25GbE Leaf is the Modern Default

• Server economics – 25GbE NICs are now mainstream; many environments are skipping 10G on new hardware.
• Optics cost & power – 25G gives a much better Gbps-per-watt ratio than older 10G solutions.
• Clean oversubscription design – 48 x 25G down / 8 x 100G up is a well-understood leaf baseline.

Best-Fit Use Cases for QFX5120

• Primary EVPN leaf/ToR in new fabrics
• High-performance virtualization clusters (VMware, KVM, container platforms)
• Storage networks where 25G to the host is sufficient and 100G uplinks handle aggregation

If you’re building a new fabric today and don’t have extreme AI/HPC requirements, QFX5120 is usually the default answer.

5. Juniper QFX5130 — 100/400GbE for Spine, AI, and High-End Leaf

The QFX5130 line moves the QFX family firmly into the 100/400GbE era. A flagship model like QFX5130-32CD-AFO provides 32 x 400GbE ports that can be broken out into 100GbE or 4 x 25GbE, making it extremely flexible as either spine or leaf.

QFX5130 Highlights

• Cut-through, low-latency switching with modern switch silicon
• 400GbE support for AI/ML clusters, large-scale storage, and high-bandwidth spine layers
• Robust EVPN-VXLAN, segment routing, and automation hooks for intent-based fabrics

Typical Roles

• Spine switch in a QFX5120-based 25/100G fabric (100G leaf uplinks into a 100/400G spine)
• High-end leaf for dense GPU or NVMe storage racks that require 100G or 200G server links
• Data center interconnect (DCI) node where 100G/400G WAN or backbone links terminate

In competitive terms, QFX5130 plays in the same arena as Arista 7060X4/7388X4 or Cisco Nexus 9300-EX/FX series—modern 100/400GbE data center silicon with deeply integrated EVPN-VXLAN.

6. Juniper QFX5200 — High-Density 25/100GbE Fabric Building Block

The QFX5200 series predates QFX5120/5130 but still occupies an important role in many 25/100G designs. Think of it as a high-density 25/100GbE platform well suited for leaf or spine roles where you want many ports in a consistent form-factor.

Key Traits

• Mixed 25GbE and 100GbE port layouts (model dependent)
• Strong EVPN-VXLAN feature set and scale
• Often deployed as a spine or “aggregation leaf” in QFX5120-based fabrics

In many designs, QFX5120 becomes the standard 25G leaf, while QFX5200 or QFX5130 provide higher-density 100G aggregation and spine capability.

7. Side-by-Side Comparison: QFX5100 vs QFX5110 vs QFX5120 vs QFX5130 vs QFX5200

8. Recommended Fabric Design Patterns

Pattern A — Modern 25/100GbE EVPN Fabric

        Spine Layer:  QFX5130 or QFX5200 (100/400G)
                |      |      |
      ---------------------------------
      |               |               |
   Leaf Layer:   QFX5120-48Y-AFO2
   (25G down, 100G up)

• When to use: New greenfield build, server NICs at 25G, moderate to high scale.
• Pros: Clean, simple, future-proof; easy to expand.
• Cons: Requires 25G NICs and 100G optics investment.

Pattern B — Brownfield Migration from 10/40G to 25/100G

 Legacy Pods:                   New Pods:
 [QFX5100/QFX5110] ----- Border ----- [QFX5120 Leaf] ---- [QFX5130 Spine]

• Keep existing QFX5100/QFX5110 racks in service.
• Introduce new pods with QFX5120 leaves and QFX5130/QFX5200 spines.
• Use QFX5110 or MX routers as EVPN border gateways between old and new domains.

Pattern C — High-Performance / AI / Storage Cluster

           [ QFX5130-32CD Spine (400G) ]
                  |      |
        -----------------------------
        |                           |
  [QFX5130 Leaf]              [QFX5120 Leaf]
  100/200G GPUs               25G general servers

• QFX5130 provides 100/200/400G to GPU or NVMe hosts.
• QFX5120 handles general compute/storage racks.
• All participate in a common EVPN fabric.

9. SKU Shortlist — QFX Models Worth Standardizing On

To keep procurement and sparing simple, many operators standardize on a small subset of SKUs:

• QFX5100-48S-AFO — Affordable 10G leaf for legacy or edge pods.
• QFX5110-48S-AFO — Transitional border leaf where 10G and 100G intersect.
• QFX5120-48Y-AFO2 — Primary 25G leaf/ToR choice for new builds.
• QFX5130-32CD-AFO — 100/400G spine or high-performance leaf for AI/HPC or core aggregation.
• QFX5200-32C-AFO — Optional, when you want a dense 25/100G platform as aggregation or spine.

Choosing two to four “golden” SKUs simplifies optics, spares, and configuration templates while still giving you room to scale.

10. Upgrade Paths — When to Move from QFX5100/5110 to 5120/5130/5200

Stay on QFX5100/QFX5110 When…

• Your racks are still 10GbE-only and you don’t anticipate upgrading servers for several years.
• You have spare optics and transceivers standardized around 10G/40G.
• The fabric size is modest and EVPN scale requirements are limited.

Plan a Migration to QFX5120 When…

• New server purchases now ship with 25GbE NICs by default.
• You’re deploying EVPN-VXLAN as the default DC architecture.
• You want modern telemetry, automation, and Junos lifecycle support for the long haul.

Introduce QFX5130 or QFX5200 When…

• Your spine layer is becoming a bottleneck at 40G or 100G.
• You’re deploying GPU/AI clusters, high-IOPS storage, or large east-west workloads.
• You want to consolidate multiple smaller fabrics into a single, larger EVPN domain.

11. Related QFX & EX Resources

• Juniper QFX vs EX Series Switches: How QFX data-center platforms differ from EX campus switches
• QFX5120 Series Comparison Guide: Deep dive into QFX5120 models and port layouts
• All Juniper QFX Series Switches at BrightStar Systems
• Juniper EX4300 Series and EX4400 Series for campus/edge roles that complement QFX fabrics

Still unsure whether you should standardize on QFX5120, mix in QFX5130, or continue leveraging existing QFX5100/5110 inventory? Contact BrightStar Systems and we’ll help you design a pragmatic, cost-optimized QFX roadmap tailored to your specific data center strategy.

Juniper QFX5100–QFX5200 Deployment Architectures & Design Patterns

Choosing between the QFX5100, QFX5110, QFX5120, QFX5130, and QFX5200 isn’t just about speeds and feeds. Each generation lines up naturally with certain data center fabric designs, upgrade paths, and lifecycle strategies. This section walks through common deployment patterns and how the QFX lineup fits into each one.

1. Classic L2/L3 Top-of-Rack with QFX5100 & QFX5110

Earlier-generation fabrics often started with 10G/40G top-of-rack (ToR) designs, where each rack has a ToR switch uplinking to an L3 aggregation or core layer. For customers in this position,the used Juniper QFX5100 switches and QFX5110 switches remain extremely practical.

• QFX5100 – Great fit for 10G server access with 40G uplinks, L2/L3 ToR, and migration from legacy EX/QFX stacks.
• QFX5110 – Adds 25G/100G options while still supporting mixed 10G environments, ideal for “refresh without forklift” projects.

In many environments, we see a pattern where legacy EX access or older EX vs QFX designs are incrementally upgraded by inserting QFX5100/QFX5110 as ToR or aggregation, then later introducing QFX5120-48Y or QFX5130-32CD in new pods for 25G/100G-heavy workloads.

2. EVPN-VXLAN Leaf–Spine Fabrics with QFX5120 & QFX5130

For modern data center designs, the most common pattern is a leaf–spine EVPN-VXLAN fabric. Here, the leaf layer terminates servers and storage, while the spine provides high-bandwidth, non-blocking interconnect. This is where the QFX5120 and QFX5130 really shine.

• QFX5120 – Excellent as a 25G access leaf with 100G uplinks. The QFX5120 comparison guide
  models (like the QFX5120-48Y-AFO2) are ideal when you’re migrating from 10G/40G to 25G/100G but still need feature-rich EVPN, advanced QoS, and flexible buffer profiles.
• QFX5130 – Higher scale and 100G-dense options (e.g., QFX5130-32CD-AFO) make this line a great
  fit for spine roles or bandwidth-heavy leafs where 50G/100G access and 100G/400G aggregation are in play.

In a typical EVPN-VXLAN fabric:

• Leaf layer: QFX5120-48Y for 10/25G servers, with 4–8×100G uplinks.
• High-performance leaf or small spine: QFX5130-32CD providing dense 100G and flexible breakout for 4×25G or 4×10G connections.
• Underlay: eBGP or IS-IS, with EVPN as the overlay control plane and VXLAN for tenant isolation.

This combination is ideal for customers building greenfield EVPN fabrics or replacing older MPLS/VPLS aggregation layers while keeping the option to interwork with Juniper MX Series routers at the core or DCI edge.

3. Collapsed Core Designs for Mid-Size Data Centers

Not every data center requires a fully distributed spine–leaf topology. In many mid-size or regional deployments, a collapsed core design can reduce hardware count and complexity while still delivering high bandwidth.

A common pattern:

• Core/Spine: Pair of QFX5130 or QFX5200-class switches acting as both spine and core.
• Leaf/Access: QFX5120-48Y leafs in each rack, especially where 25G access is the new standard.
• Edge/Services: MX204/MX480 edge for WAN, Internet, or MPLS services, connected via 100G to the collapsed core.

This design is especially attractive when you’re consolidating multiple legacy environments (for example, EX distribution plus older QFX core) into a single EVPN-VXLAN fabric without deploying a full multi-stage spine.

4. High-Density Spine & DCI with QFX5200

At the top end, the Juniper QFX5200 series comes into play as a spine switch or high-density aggregation node. In environments where 100G is the baseline and 25G is ubiquitous at the leaf, QFX5200 provides the scale needed to keep oversubscription under control.

• Role: 100G spine for multiple QFX5120/QFX5130 leafs.
• Use case: Large private clouds, service providers, and content networks with high east–west traffic.
• DCI: When paired with MX204/MX304/MX10003 at the edge, QFX5200 can serve as a high-bandwidth handoff for EVPN-based data center interconnect.

Many customers adopt a hybrid approach where a QFX5200 spine fabric is fronted by Juniper EX4400 campus switches at the edge, and QFX leafs in the data center, creating a consistent EVPN-ready architecture from access to core.

5. Practical Migration Paths from Legacy QFX to 25G/100G Fabrics

If you’re still heavily invested in earlier QFX5100 deployments, you don’t need a forklift upgrade to reach a modern EVPN fabric. Common migration paths we see:

1. Phase 1 – Stabilize & Segment: Keep existing QFX5100 ToR and aggregation, but introduce a small EVPN “pod” using QFX5120-48Y and QFX5130-32CD for new workloads or greenfield applications.
2. Phase 2 – Gradual Cutover: Migrate high-bandwidth or latency-sensitive applications to the new EVPN pod, while keeping legacy services on the existing QFX5100/QFX5110 core.
3. Phase 3 – Fabric Unification: Expand the EVPN fabric across racks, retire older aggregation layers, and consolidate services onto QFX5120/QFX5130/QFX5200 with MX routers handling DCI and WAN.

This approach keeps CAPEX manageable and allows your team to gain operational experience with EVPN and automation tools before committing to a full data center-wide migration.

6. Summary: Matching QFX Models to Your Fabric Design

As a quick mental model:

• QFX5100/QFX5110: Best for mixed 10G/40G or 10G/25G ToR, legacy integration, and incremental upgrades.
• QFX5120: Workhorse 25G leaf with 100G uplinks for EVPN fabrics and modern virtualization stacks.
• QFX5130: High-scale 100G leaf/spine for bandwidth-heavy applications and dense interconnect.
• QFX5200: Spine/aggregation role in large fabrics, especially where 100G is standard across the data center.

If you’d like help mapping specific SKUs (for example, QFX5120-48Y-AFO2 vs QFX5120-32C vs QFX5130-32CD) to your topology, contact BrightStar Systems and we can walk through port counts, buffer profiles, optics, and growth plans to recommend the right QFX mix for your environment.