Virtualization, 10G Ethernet streamline Cisco user's data center

The infrastructure implemented by Transplace is a great example of the target application for Cisco's data center switches
  • Jim Duffy (Network World)
  • 02 June, 2011 07:48

It is by no means the largest or most sophisticated Cisco network there is -- yet the infrastructure implemented by Transplace is nonetheless a great example of the target application for Cisco's data center switches.

Transplace is a Dallas-based transportation supply chain and logistics management company employing 550. It has a Cisco Nexus network of a single 7000 in the core, two 40-port-plus 5020s at the top of its racks, and two 48-port 2X48 fabric extenders for port extension in its co-located test and development data center in Lowell, Ariz. The company is currently building a similar configuration at its production data center in Dallas.

IN DEPTH: The incredible shrinking data center

The networks and data centers support Transplace's secret sauce -- its proprietary, on-demand Transportation Management System (TMS) that manages the logistics between shippers and carriers to optimize supply chain operations. TMS automates order management, pinpoints optimal shipment and tracks the process to make it more efficient.

TMS works with Transplace's Scenario Pro Technology system, the processing intelligence behind building sustainable transportation routes, increasing freight load ratio and best utilizing assets.

Hosting these platforms are nine -- soon to be 18, once the production data center project is complete -- VMware ESX servers on Dell blades, each supporting 150 virtual machines. These replace scores of Sun Unix servers.

The VMs are mirrored between the Dallas and Lowell data centers over a private OC-3 link for disaster recovery, with Silver Peak NX-7600s providing WAN optimization over that connection.

"This allows us to replicate a large amount of data between the data centers in near real time," says Scott Engel, director of IT infrastructure for Transplace.

Supporting that many workloads requires bandwidth, which the Nexus network provides via 60 ports of 10G Ethernet in each data center. But the real clincher was the cable management benefits the switches and virtual environment provided.

"The real factor here was to reduce the complexity in our cable plant, which in turn also reduces our footprint," Engel says. "And that's always good when you're in a co-lo facility."

Engel says the Nexus/ESX network has reduced Transplace's data center footprint by 40 per cent. He did not disclose the cost savings, but anytime you can cut your leased square footage by almost half, it is significant.

The Nexus network is less than a year old. It's replacing an older, 672-port Catalyst 6509 and 6509E core, running the Supervisor 720 routing engine and 48-port 10/100/1000Mbps line cards. Obviously, consolidating most of the 672 ports of Gigabit Ethernet into 120 ports of 10 Gigabit across the two data centers reduces a considerable amount of cabling.

Add to that the ability to virtualize workloads and have them share the same physical infrastructure eliminates even more cabling and footprint.

Transplace had been a Catalyst shop for 11 years but decided four years ago to embark on a data center "refresh" for both its test and development, and production facilities. In so doing, it moved from a collapsed Catalyst 6500 core infrastructure to a distributed top-of-rack configuration of Nexus 5000s and 2000s supporting the Dell/ESX servers and an IBM database.

Engel expects to scale out this infrastructure as requirements deem necessary, adding 5000s to the top of rack and higher-density modules to the 7000 core. The top of rack may see the 1RU 48-port 10G Nexus 5548 for higher-density requirements without adding modules to the 7000, Engel says.

Since the company just upgraded to 10G, Engel currently doesn't see a need for 40G or 100G Ethernet in the Transplace data centers. He is aware of Cisco's roadmap for that next-generation Ethernet technology -- which some believe is code-named "Jawbreaker" -- but his needs are met for now.

For connecting the two data centers, Transplace is also evaluating Cisco's Layer 2 Overlay Transport Virtualization (OTV) technology. OTV is an enhancement to the Nexus 7000's NX-OS operating system that encapsulates and tunnels Ethernet through a routed Layer 3 infrastructure to enable multiple dispersed data centers to look like one logical entity.

This allows IT administrators to move workloads and virtual machines across data centers without having to redesign or reconfigure their network. For an enterprise like Transplace, this is important to move data center workloads across long distances for resource allocation or disaster recovery.

"We're still looking at OTV... but we're not sure it's the route we'll go," Engel said.

Transplace is also preparing for storage convergence. Cisco's Nexus switches have a capability called Unified Ports that allows users to designate any port to be 1G or 10G Ethernet, 2/4/8G Fibre Channel or Fibre Channel over Ethernet (FCoE).

"We currently have no need for Fibre Channel or FCoE, but we see it coming down the pipeline and we want to be prepared for it," Engel says.

While these goodies await evaluation by Transplace, the company is currently enjoying its 40 per cent reduction in footprint and high-bandwidth virtual environment that eliminates a lot of cabling.

"The complexity of the infrastructure drove our refresh," Engel says. "We needed to reduce footprint, cable infrastructure and the complexity. And with 150 VMs on a single ESX host -- while running storage and application traffic through the same pipe -- you need the bandwidth."

Read more about lan and wan in Network World's LAN & WAN section.