Storage Channel Pipeline » FCoE

Transparent Interconnection of Lots of Links (TRILL): IP networking’s future

Eric Slack — Mon, 20 Sep 2010 15:17:25 +0000

Spanning Tree Protocol (STP) was invented years ago as a networking technology to prevent bridge loops (a destructive feedback type of condition) by allowing only one path between network switches or ports. A Layer 2 network protocol, STP computes a plan for routing traffic between every connected device through a “root bridge” such that only one path is used. This path, which is based upon rules configured by the user of the protocol, may not always be the most direct. The plan, or “spanning tree,” describes this set of nonredundant paths and disables all others. When a network segment goes down, an alternate path is chosen, but this process can take a few seconds, something that may be OK for communications but can be unacceptable in a storage network. Another standard, Transparent Interconnection of Lots of Links (TRILL), is designed to address this problem.

As networks have grown and become more complex due to technologies like convergence and virtualization, STP has proved to be inefficient. The indirect paths that a spanning tree contains cause data (or migrated VMs) to be routed through more network “hops,” adding latency into the storage environment. This concentration of traffic through the subset of available links that form the tree reduces effective network bandwidth and increases latency. Finally, blocked paths represent unused bandwidth, which only gets worse with each new generation of network hardware. Now, with FCoE bringing more storage traffic to the IP network, these problems will intensify.

The Internet Engineering Task Force’s Transparent Interconnection of Lots of Links network standard applies network layer routing protocols to the link layer and, unlike STP, incorporates a knowledge of the entire network and uses that information to support Layer 2 multi-pathing. This can enable multi-hop FCoE, reduce latency and improve overall network bandwidth utilization.

Network performance and efficiency are pretty fundamental concerns for IT organizations. For a VAR, when these issues with STP come up or when the term TRILL is mentioned, some background may be very useful. For more information, take a look at the webinar on TRILL by Storage Switzerland.

Follow me on Twitter: EricSSwiss

Tech Refresh: The FCoE protocol

Eric Slack — Mon, 05 Apr 2010 14:10:44 +0000

Fibre Channel over Ethernet, as the name implies, allows Fibre Channel traffic to be run over Ethernet networks. Technically, the FCoE protocol encapsulates FC frames inside jumbo Ethernet frames, replacing the FC 0 and FC 1 layers in the stack with Ethernet. At the server, it replaces the Ethernet NICs and FC HBAs with a single converged network adapter (CNA), which provides all the network connectivity for that host via a single 10 Gigabit Ethernet (10 GbE) connection. This cable connects to an FCoE switch, which consolidates FCoE cables from multiple servers, as it connects to the existing data center FC SAN and Ethernet LAN.

FCoE is a consolidation technology, designed to simplify the network connections to servers and provide some options for increasing I/O bandwidth. The proliferation of low-profile blade servers is reducing the real estate available for network cards, while at the same time, server virtualization is increasing the need for bandwidth. The FCoE protocol can address both these problems by providing 10 gigabits of bandwidth to the server and replace its multiple 1 GbE NICs and FC HBAs with a single CNA card. This can also reduce infrastructure costs as well as power and cooling costs and greatly simplify cabling.

The FCoE protocol is appealing for data centers with existing FC SANs and an interest in reducing network connectivity issues. Another alternative, especially If FC is not present, can be I/O virtualization (IOV). There are different IOV technology approaches available, using PCIe and InfiniBand, but they also provide a single network connection from a consolidation switch to servers. This solution can increase bandwidth while providing flexibility and simplified management.

For VARs, the point is to understand that server network connectivity and I/O bandwidth are issues that are very real for an increasing set of customers. There are a number of factors to be considered before planning to deploy this new technology, which makes it a discussion that’s ideal for VARs. It involves larger environments, FC SANs and enough complexity to generate real integration projects. And like most IT problems, there are multiple options; VARs that can present more than a single solution are at a distinct advantage.

Follow me on Twitter: EricSSwiss.