Odds are most of us will never need the network throughput speeds of 2GB Fibre Channel or iSCSI or 10GB Ethernet.
But, if you're working at a data center with hundreds to thousands of servers, clusters and terabyte-sized databases, then every millisecond matters. This is why Intel is promoting a new, fast way of transporting data: FCoE (Fibre Channel over Ethernet) for Linux.
FCoE's purpose is to enable data centers to consolidate LAN and SAN (storage area network) traffic over 10GB Ethernet. FC, which comes in speeds from 2 to the just arriving 8G bps, is commonly used in data center SANs. In recent years it's been challenged by iSCSI. Fibre, which, despite the name can run both on copper and fiber-optic cables, is seen as faster and more reliable, while iSCSI is commonly thought of as less expensive.
Intel, along with FCoE's founder Cisco Systems, is hoping to combine the virtues of both Fibre and iSCSI with this new high-speed, dual-purpose network fabric. In addition, by making it possible to use FCoE for both LAN and SAN traffic, the companies want to gain customers who want to simplify their data center network management.
To help this happen, Intel has released the first FCoE code for Linux under the GPLv2 on the Open-FCoE site. This code is barely out of alpha. It can only run reliably, according to the Open-FCoE site, on a generic 2.6.23 Linux kernel. The installation instructions are optimized for Fedora 7 systems.
In its current state, the code is only capable of some FC protocol processing and encapsulating FC frames in Ethernet packets. Unlike iSCSI, FCoE does not run on the TCP/IP stack. This is Fibre Channel on Ethernet without the overhead or the management and analysis tools of TCP/IP.
Early days or not, Intel and Cisco are expecting great things from FCoE. "Fibre Channel over Ethernet will be a key capability for our customers, offering seamless server and storage access in the data center," said Jayshree Ullal, Cisco's senior vice president of the data center, switching and services group, in a statement. "The emergence of 10 Gigabit Ethernet bandwidth combined with Cisco's proposed extensions to Ethernet, enables a lossless and resilient fabric for data center I/O consolidation."
Will Intel and Cisco's hopes be realized? According to Dan Kusnetzky, principal analyst of the Kusnetzky Group, in an interview, it all depends on the fundamentals: "1) Price—will this approach cost more, in the end, than acquiring some other technology? 2) support—who do organizations or suppliers turn to for support? Can they get a contract for on-site, immediate support? And 3) ecosystem—who has agreed to build upon this specification? If there is no community, no products, there will be no interest."
Greg Schulz, founder and senior analyst for StorageIO, an infrastructure technology analysis and consulting firm, was more optimistic in an interview. "Intel opening up their kimonos and throwing their full support for FCoE is nothing but good for the emerging technology and surrounding ecosystem to help vendors get their software and drivers ready and avoid being a gating factor in the technology rollout."
The future, Schulz said, "is converged fabrics and networks in the data center and while ultimately that might involve TCP/IP, for now, a major stepping stone and continued evolution from propriety storage interconnects and interfaces is to get on to a common underlying network interface infrastructure, and that is data center class Ethernet that supports both Fibre Channel and TCP/IP-based traffic concurrently without having to stack or layer protocols on top of each other."
Looking ahead he sees "a continued downward trend on 2GFC and 4GFC with the advent and rollout of 8GFC on the Fibre Channel front and continued growing adoption of iSCSI at both 1GbE and in the future with 10GbE, with FCoE, that game realistically … is probably at least 18 months out."
That may not be such a bad thing though because it will give "time for those needing to jump from 4GFC to 8GFC and then for their next jump, the option of going to FCoE and continue on the Fibre Channel FCP/FICON route, or to leverage the same adapter and run iSCSI or NAS, or to jump to standard Ethernet and run iSCSI and NAS [network-attached storage]."
Schulz foresees FCoE being used for short-haul data traffic. "FCoE is not for spanning distances—that's where FCIP or Fibre Channel mapped to IP comes into play. FCoE is for the data center [that needs] low latency, quality of service and so forth for good performance. This performance will help both FCoE and iSCSI when using this new form of Ethernet."
At the end of the day, it's 10GB Ethernet and Data Center Ethernet (PDF link) that Schulz believes is more important than just FCoE.
"Bottom line is that Data Center Ethernet, which is at the heart or [the] underlying enabler for FCoE, is more than just Fibre Channel over Ethernet. Data Center Ethernet also means enhanced performance for iSCSI, NAS and for implementing I/O virtualization in general in the data center," said Schulz.
Check out eWEEK.com's Storage Center for the latest news, reviews and analysis on enterprise and small business storage hardware and software.