Fibre Channel over Ethernet (FCoE) may still be a couple of years from gaining traction in the data storage market, but there are some early adopters already signing on to use the storage networking technology.
On such customer is Oregon’s Salem Hospital, which turned to an FCoE solution from Cisco Systems (NASDAQ: CSCO) to upgrade its data center and improve business continuity, reduce energy consumption, and accommodate more storage from two key systems — its electronic medical records (EMR) and picture archiving and communication system (PACS). Salem Hospital implemented Cisco Nexus family and Cisco MDS 9000 Multilayer switches to facilitate the upgrade while simultaneously moving to a virtualized environment and beginning the deployment of FCoE.
“We hit a wall on speeds and on how many virtual machines we could put in place,” said Heng Him, a technologist at Salem Hospital. “The Cisco Nexus means we have three fabrics including FCoE, higher port counts and speeds, and we have been able to consolidate our data center to reduce space.”
Salem Hospital is a not-for-profit hospital located in Oregon. This regional medical center is one of the largest of the state’s 57 acute care hospitals and operates the busiest emergency department in the state.
Previously, the organization used mainly Dell (NASDAQ: DELL) pizza box servers that provided 2 to 4 Gbps Fibre Channel and were not fully redundant. These were supported on the storage side with what Him characterizes as a hodgepodge of storage area network (SAN) gear from vendors such as EMC (NYSE: EMC), Brocade, Hitachi Data Systems (HDS) and Extreme Networks. Some switches, for instance, were six years old, and one Hitachi Lightning box had put in almost a decade of service.
The servers were upgraded to Dell PowerEdge 2950, R610 and R710 rackmount models running Windows 2008 Server in both physical and virtual environments. Some older servers remain, however. EMC Clariion CX3 units became the disk array of choice.
As Cisco dominates in the networking arena, it made sense to implement the new Cisco Nexus switches, which were designed to enable both virtualization and FCoE. Salem Hospital looked at competitive offerings from Juniper Networks (NASDAQ: JNPR) and others, but Him pointed to Nexus features such as power savings, speeds and feeds, redundancy and the value of virtualized switches as the reasons he opted for Cisco.
“The full feature set of the Nexus, including FCoE, enables us to consolidate our hardware, reduce cabling and increase overall efficiency,” said Him. “With the ability to deal with multiple fabrics, as well as higher port counts and speeds, the Nexus is future-proof.”
One pair of Cisco Nexus 5020 Switches, with FCoE support, connects over lossless 10 Gigabit Ethernet to the hospital’s newer servers, which host multiple healthcare applications and Microsoft Exchange. The Cisco Nexus 5020 separates the storage traffic from data traffic, sending the storage to the Cisco MDS-based SANs and the latter to the Cisco Nexus 7000 Switch at the core. Another pair of Cisco Nexus 5020 Switches connects to existing Gigabit Ethernet servers by way of Cisco Nexus 2000 Fabric Extenders.
QLogic CNAs connect the servers to the storage. According to Him, he only needs two cards per server (one for redundancy) compared to four in the past — two for the network and two for Fibre Channel. Now one card can deal with both functions.
Page 2: Big Power Savings
Big Power Savings
While FCoE is seen as the future, the organization hasn’t eliminated its traditional Fibre Channel fabric. Older servers continue to run FC and are being switched out gradually. Him estimates that it might be three years before they are entirely changed out. But the results of the switch to FCoE have been immediately apparent in terms of greater reliability and lower energy consumption. As a result, the organization no longer has a power crunch.
“Since we put Nexus in place six months ago, we’ve had no network downtime,” said Him. “The Nexus 2000 consumes 89 watts per box with data running across it. We were very short on power before, but now we have more margin.”
The Nexus and MDS both use the NX-OS, which helps make management and training easier. Salem Hospital has two engineers trained on this operating system. It provides them with visibility into the network and storage, so a networking technician can troubleshoot without having to call upon his storage equivalent, and vice versa. Previously, far more interaction was needed between the networking and storage disciplines to resolve problems.
Another element of the Cisco plan for global data center domination is the Cisco Unified Computing System (UCS). This compact unit combines storage, server and networking gear in one form factor and does away with a lot of cabling. The hospital currently has Cisco UCS under testing.
“We are excited about the initial numbers from Cisco UCS,” said Him. “We are looking at the possibility of using it for our entire data center.”
FCoE Learning Curve
While Him is excited about the potential of FCoE, he admitted that getting it up and running on the first couple of servers was far from plain sailing. He ran into issues with drivers and difficulties in configuration.
“As the Windows drivers were not working properly, we weren’t provisioning correctly,” he said. “It was mainly configuration glitches and a general lack of understanding.”
With the experience gained from the first couple of machines, he said setting up servers and getting VMs running over FCoE now saves significantly on overall deployment time. Him gives the example of an imaging system that ran on an older server. The hospital moved it onto a VM running over FCoE in a few hours.
His advice to others on FCoE implementation is to extensively test provisioning of LUNs to ensure the various channels are working properly.
“Once you get comfortable with the technology, it becomes so much easier,” said Him. “It’s all about testing.”
Follow Enterprise Storage Forum on Twitter