Layer 2 Data Center Interconnect: The Complete Guide to DC Connect

Running infrastructure across more than one data center solves real problems — resilience, geographic reach, disaster recovery — but it also creates a new one: how do you make two physically separate facilities behave, from a networking perspective, like a single site? That’s the exact problem Layer 2 interconnect exists to solve.

What is Layer 2 interconnect?

Layer 2 interconnect extends a single broadcast domain — effectively, a VLAN — across two or more physical locations, so devices in different data centers can communicate as if they were plugged into the same switch. This is fundamentally different from a Layer 3 (routed) connection, where traffic between sites has to be routed between separate IP subnets, each site keeping its own distinct broadcast domain.

In practice, this is delivered over dark fiber, a managed wavelength, or a carrier-grade Layer 2 service like VPLS or EVPN — the underlying transport varies, but the result is the same: your VLAN doesn’t know, and doesn’t need to know, that it now spans two buildings.

Common use cases

Layer 2 interconnect isn’t a general-purpose networking feature — it solves a specific class of problems that Layer 3 routing genuinely can’t:

• Live VM migration — technologies like VMware vMotion require the source and destination hosts to be on the same Layer 2 network, so a VM can move between data centers without changing its IP address or dropping active connections.
• Storage replication — synchronous or near-synchronous replication between SAN or NAS systems across sites often requires low-latency Layer 2 connectivity to keep replication windows tight.
• Active-active clusters — database and application clusters designed for multi-site high availability frequently require cluster nodes to share a Layer 2 segment for heartbeat and failover signaling.
• Multi-site Kubernetes and container platforms — stretched cluster networking for container orchestration across sites often relies on the same underlying Layer 2 extension.

If your architecture doesn’t need any of these — if each site can operate independently with routed connectivity between them — Layer 3 interconnect is simpler, more scalable, and the right default.

Layer 2 vs Layer 3 interconnect: what’s the difference

	Layer 2 interconnect	Layer 3 interconnect
What’s extended	A single VLAN/broadcast domain	Separate subnets, connected via routing
Typical use case	Live migration, storage replication, stretched clusters	General site-to-site connectivity, independent sites
Failure domain	Broadcast storms and loops can propagate across sites	Contained per site — routing isolates failure domains
Complexity	Higher — requires careful loop prevention, MTU consistency	Lower — standard routing protocols, well-understood
Scalability	Limited by VLAN/MAC table scale across the extension	Scales cleanly with standard routing

Dark fiber vs managed wave vs VPLS/EVPN

There are several ways to actually deliver Layer 2 connectivity between sites, and the right one depends on distance, budget, and how much control you need:

• Dark fiber gives you a physical, unlit fiber pair with no equipment in between — you own the entire transport layer and can run whatever protocol or bandwidth your own optics support. Maximum control, but you’re responsible for the electronics on both ends.
• Managed wavelength (wave) services deliver a dedicated slice of bandwidth over a shared fiber, lit and managed by the provider — you get guaranteed, isolated capacity without owning or managing the physical fiber or the optical equipment.
• VPLS/EVPN services deliver Layer 2 connectivity over a shared, carrier-operated network fabric, typically the most cost-effective option for connecting more than two sites, at the cost of slightly less deterministic performance than a dedicated wave or fiber pair.

What makes a good DC Connect solution

Evaluate any Layer 2 interconnect offering against these criteria:

• Guaranteed bandwidth — a committed, dedicated capacity rather than a best-effort share of a congested fabric.
• Low, deterministic latency — critical for synchronous storage replication and live migration, where latency variability causes real operational problems, not just slower performance.
• Path diversity and redundancy — physically diverse routes between sites so a single fiber cut doesn’t take down your interconnect entirely.
• Encryption options — MACsec or equivalent link-layer encryption for sensitive traffic crossing shared infrastructure.
• A clear SLA — for latency, jitter, and availability, not just uptime.

Design considerations

A few things are worth getting right at design time, not after deployment:

• Loop prevention. Extending Layer 2 across sites increases the blast radius of a misconfiguration — a broadcast storm or a switching loop can now take down both sites instead of one. Spanning tree, or better, a loop-free EVPN fabric, needs to be designed in from the start.
• MTU consistency. Encapsulation overhead from the underlying transport (QinQ, MPLS, VXLAN) eats into your usable MTU. Mismatched MTU across the extended segment is a classic, hard-to-diagnose source of intermittent packet loss.
• VLAN tagging strategy. QinQ (802.1ad) lets you tunnel a customer’s own VLAN tags transparently across the interconnect, which matters if you’re extending multiple VLANs across the same physical link.
• EVPN over traditional VPLS, where available, generally offers better convergence times and native multi-homing, worth prioritizing for new deployments over legacy VPLS.

Frequently asked questions

Do I need Layer 2 interconnect for a simple multi-site setup? Not necessarily. If your sites can operate independently and communicate over routed IP connectivity, Layer 3 interconnect is simpler and more resilient. Layer 2 is specifically for cases requiring the same broadcast domain across sites — live migration, storage replication, stretched clusters.

What’s the difference between DC Connect and dark fiber? Dark fiber is one way to deliver DC Connect — an unlit physical fiber pair you light yourself. DC Connect can also be delivered as a managed wavelength or a VPLS/EVPN service, depending on distance, budget, and how much of the physical layer you want to manage directly.

How much latency is acceptable for stretched VLANs? It depends on the workload — synchronous storage replication is typically far more latency-sensitive than live VM migration. As a rule of thumb, keep round-trip latency between sites as low and as consistent as possible, and confirm your specific application’s tolerance before committing to a design.

Can Layer 2 interconnect scale beyond two sites? Yes, particularly with EVPN-based services, which are built for multi-site, any-to-any connectivity. Traditional point-to-point wave or dark fiber links scale less gracefully beyond two or three sites without additional switching infrastructure.

Connect your data centers with confidence

Fiberway’s FiberLayer™ delivers flexible Layer 2 DC Connect between Fiberway facilities, with dark fiber, managed wave, and dedicated bandwidth options depending on your latency and redundancy requirements. Talk to a Fiberway engineer about the right interconnect design for your architecture.