Tim Anderson

Taking a Moderate Approach to FCoE

Users are trying to decide if and when they should adopt Fibre Channel over Ethernet (FCoE). Suppliers of storage technologies are positioning themselves depending on their abilities and their perception of the market need. Users need guidance in separating the hype from the realities. Should you go full steam ahead into FCoE or does FCoE in moderation make more sense?

In response to FCoE there are typically three positions suppliers are likely to take. The first is the FCoE now response. These vendors typically have products ready and available for deployment and they may even have a few published success stories.  The second position is from vendors that have product in the early phases of development awaiting ratification of the standard and for user demand to build before they can fully deliver their solutions. The final position is suppliers that make iSCSI only solutions and that believe there is already a viable IP infrastructure for storage, which they would rather see adopted.

While those supporting iSCSI have been banking on 10GbE to leverage the rapid adoption of iSCSI into the large enterprise data centers, typically iSCSI solutions have only seen wide adoption in the SMB end of the market, but have yet to really break into large enterprise data centers where Fibre Channel networks are dominant.

FCoE does have a number of benefits, especially when one thinks of the virtualized server space. A few of the key benefits include, the ability to have converged Fibre Channel and Ethernet server I/O connections or ports (which save on power and cooling through reduced cabling), reduction in network management overhead, the ability to hold onto the existing Fibre Channel storage and backup targets thereby protecting a company’s existing investment, and the ability to leverage advanced Ethernet networking QoS and other management practices inside the Fibre Channel space.

FCoE however has yet to become a ratified standard by T11 or the IEEE, which should be enough to give end-users pause when thinking about deploying something that isn’t even an agreed upon standard yet. At least the enterprise customer, and it’s the enterprise customer where FCoE converged networks make the most sense at least in the initial deployments, needs to be more conservative in their approach when looking to deploy a converged FCoE network infrastructure.

The true value of FCoE is going to be capital expenditure and operational expense (CAPEX and OPEX) savings by reducing the physical management challenges of two separate network infrastructures. That said re-architecting an existing network is not a light task and in this case we are re-architecting two well entrenched (IP and Fibre Channel) infrastructures.  Consequently there are a number of things to keep in mind:

1. Cost Savings Aspects - Seeing as there are only a handful of vendors that currently support FCoE capable switches and Converged Network Adapters (CNAs), the need to have an aggressive return on investment (ROI) and Total Cost of Ownership (TCO) is of critical importance in order to make any movement into this space. TCO is a large factor in this adoption as the impetus for  storage and network teams to come together to converge their networks and it will be on the high side of the learning curve. The network team will need to understand low-latency based networks like Fibre Channel and the storage team will need to comprehend Ethernet networking principles.
   

1. Network Management & Issue Resolution - Attempting to manage any new network technology always brings with it a list of concerns like will it have the ability to plug into existing management frameworks to handle configuration and change management concerns or will it be missing key features that already exist inside the current network frameworks, such as adaptive networking or proactive ability to determine issues before they have already occurred.
   

FCoE needs to encapsulate and then de-capsulate the Fibre Channel frames which makes one wonder how will network errors be tracked down and resolved when they occur inside the FCoE frame or vice-versa on the Ethernet side. Provisions will need to be made to ensure that a systems administrator can get down into the appropriate stack to determine the issue. As a result of the transmission of both protocols over the same transport medium, the management techniques will need to be more clear cut and transparent than ever before. The complexity of convergence is why in the past storage and Ethernet personnel elected to run completely separate network infrastructures.

1. Converged Network and Physical Layer Concerns - One of the key goals of FCoE is to reduce the number of storage and network cards in the server.  For example a typical virtualization host may have two storage HBAs and two or four network interface cards. With FCoE it is entirely possible to reduce that card count down to only two CNAs. The problem is that this potentially limits the amount of redundancy that may exist in a more traditional approach. For example, if there is a issue with one of the CNA's FC traffic which results in the need to remove one of the connections, instead of just taking down one half of the I/O stack you also end up taking down the Ethernet interfaces as well, potentially causing a larger disruption then may have been intended.
   

1. Security Concerns - Taking the case of a service provider or an end user with multiple separate network segments which address the need to isolate traffic based on business needs and/or security concerns, FCoE needs to ensure that these current security practices will be realized with CNAs and FCoE switches.
   

1. Political Landscape Issues - Understanding the enterprise customer’s organizational structure and layout is also an important and sometimes an overlooked factor of FCoE’s benefits. Obtaining the cooperation necessary to get the storage and Ethernet communities together at an organization is much more complicated than is normally realized. It is important to recognize that the storage team builds and maintains the Fibre Channel network while the network team manages the Ethernet network. Attempting to get these two traditionally separate departments to agree on the various consolidation issues, especially with a standard not yet ratified, would be a large undertaking.
   

Many of these concerns will surely be addressed as the standards are ratified and as more vendors deploy products and best practices around the technology. Until then customers should take note, watch how the new infrastructure unfolds and just like any new technology, take the time to understand not only the benefits FCoE will bring to their infrastructure, but also the amount of effort and time that will need to be spent training and managing a new network infrastructure.

FCoE in moderation seems to be the logical approach for now. FCoE will eventually get to where it needs to be and there is no doubt about its benefits and extensive capabilities. But as with anything else, all things should be taken in moderation. Do the necessary research to break through the hype and marketing jargon, take an in-depth look at the current state of a given storage networking infrastructure and then make a decision as to how this would most benefit your organization.