Data Center
Modernizing the enterprise data center
From legacy core to VXLAN / EVPN fabric, without losing a day of production.
The problem
Four problems we find in most data centers
Each one is a risk to uptime today and a blocker for the next migration. Here is what we see, and how we fix it.
What we transform
From a fragile core to a resilient fabric
Most refreshes start with an aging core where one failure takes everything down. We replace it with a fabric built to survive failures and upgrades.
Before
Traditional data center
- Aging, end-of-support core switching
- Firewalls on single links
- Management on the production network
- Manual, box-by-box changes
After
VXLAN / EVPN fabric
- Spine-and-leaf fabric with a VXLAN / EVPN overlay
- vPC dual-homing for servers and services
- Port-channeled firewalls at up to 100G
- Out-of-band management and NDFC automation
What we hand over
Documentation you can operate from
Diagrams
L1, L2 and L3 diagrams, rack elevations and patch sheets.
Records
Circuit, IP and VLAN/VRF documentation.
Dependencies
Application and server connectivity mapping.
Runbooks
Migration runbooks, MOPs, validation procedures and as-builts.
Resilience by design
Built to survive failures and upgrades
Three design choices keep production running when a switch, link or firewall goes down, and keep you able to fix it.
Dual-homed with vPC
Servers, storage and services connect to a pair of leaf switches, so one switch failure or upgrade never takes them offline.
What you get
- No single switch in the data path
- Upgrades without downtime
- Active / active links
Firewalls on port-channels
Each firewall connects to two switches over aggregated links, up to 100G, so a cable, port or switch failure doesn't take security down.
What you get
- No single link to the firewall
- Room to grow throughput
- Clean, tested failover
Out-of-band management
A separate management network reaches every console and management port, so you can still fix production when it is down.
What you get
- Built before production moves
- Separate from the data path
- Safer cutovers and rollbacks
Proof
The same method on three continents
Greenfield VXLAN/EVPN fabric
Dual spine, ~20 leaves, 100G spine-to-leaf, BGP EVPN with Nexus Dashboard and NDFC, delivered through production and since expanded.
Read the case →Case studyAshburn brownfield modernization
New core, Palo Alto aggregate Ethernet at 100G and a dedicated OOB layer in a live facility, with AWS and Azure connectivity preserved.
Read the case →Collapsed core replaced with redundant NX-OS core, vPC pairs, 100G and redundant Palo Alto connectivity.
Refresh of a facility that is also the regional WAN hub for MPLS, DMVPN and branch traffic.
Phased BGP overlay remediation with redundant paths validated across data centers.
How it runs
How a refresh runs
Discovery
Dependency mapping across cabling, circuits and applications.
Architecture
Design and bill-of-materials validation.
Staging
Configuration build before anything is racked live.
OOB first
Out-of-band built before production moves.
Migration
Production moves in controlled windows.
Validation
Testing and hypercare after every window.
Work with us
Start with one facility
Begin with a current-state assessment of one facility. Senior engineers and one point of contact from first call to final handoff.
EXPERIENCESenior engineers, one point of contact and full documentation on every engagementStart a project