Network Evolution Challenges
Testing, Growth and Security Implications for Developers of Ethernet-based NetworksBy Cheryl Coupé, editor
As Ethernet gears up to 10, 40 and 100 Gigabits per second, developers better pull on their running shoes to keep up. Ron Cates, vice president of marketing, networking products for PLX Technology and Drew Sproul, director of marketing at Adax, Inc., address some of the key challenges developers will face moving forward.
EECatalog: What are some of the testing challenges developers are facing in the evolution to 40G and 100G networks?
Drew Sproul, Adax, Inc.: Expense, reliability and effort are the biggest challenges for testing 40G and 100G networks. These new speeds require much more precision in all aspects of equipment design – and therefore testing – to ensure high-quality, highly reliable products for the telecom market. Initially, test networks will be expensive to set up, requiring all-new test measurement and traffic generating equipment. Since first-generation equipment is never bug free, a substantial amount of time and effort must be spent ensuring that the test network itself is stable and reliable. Inevitable revisions to the testing equipment must be factored into the schedule. This will clearly affect the overall effort, as verifying the test harness will take time away from evaluating the device under test.
EECatalog: As everything but the kitchen sink moves to the cloud, what are some key implications for developers of Ethernet-based networks?
Ron Cates, PLX Technology: The macro implication, of course, is that the volume of traffic in cloud-based data centers grows rapidly. This increases the bandwidth requirements for links between servers, switches and storage elements within the data center. 1Gbps speeds, which dominate today, will migrate to 10Gbps rates and those will become the standard link speed between network elements. Such a migration could occur more smoothly if the connectivity technology chosen was one that capitalized on the existing cabling system in the data center and was backward-compatible with the dominant technology currently in use. Fortunately, 10GBase-T technology fits those requirements. It uses inexpensive Cat6A twisted-pair cabling and familiar RJ45 connectors, thereby lowering the cost of implementing 10GE links.
Sproul, Adax, Inc.: Simply put: security and virtualization. With so much crucial commercial and personal data flowing over the Internet through the mobile network from so many devices, security is the number one concern. Banking and financial transactions are done online and businesses are using video conferencing for the most sensitive business and technical conversations.
Virtualization opens new opportunities and challenges for managing Ethernet data streams to the appropriate applications hosted on fewer servers. Traffic and bandwidth management has become a major key to the successful operation of networks in conjunction with virtualized server farms.
EECatalog: What are some of the challenges for developers as service provider networks shift from TDM to IP-over-Ethernet?
Cates, PLX Technology: One such challenge is utilizing Ethernet transport for consolidated services, which transport both voice and data. In addition to the obvious quality-of-service priority needed for an isochronous service such as voice, the Carrier Ethernet transport needs to extract clock information associated with the voice information prior to Ethernet conversion. A standards-based system such as IEEE 1588 V2 should be used to facilitate interoperability between vendors’ systems for clock recovery.
Sproul, Adax, Inc.: If there were one word that best describes TDM networks it would be determinism. In the all-IP packet-based networks, the traditional ‘best-effort’ approach to delivery is no longer acceptable from a customer’s QoS or QoE perspective, nor from the commercial side of monetizing prioritized network services. Packet processing of IP flows for security, shallow- and deep-packet inspection is the key to bringing determinism into the new world.
EECatalog: What trends and implementation challenges are you seeing in the use of Energy Efficient Ethernet (EEE)?
Cates, PLX Technology: EEE is very important for 10GE migration. The power dissipation of 10GE PHY devices is higher than that of the 1GE counterparts and an algorithm such as EEE, which dramatically reduces the effective power dissipation based on traffic patterns, is extremely useful. The challenge is that EEE requires both ends of the link to support EEE in a fully interoperable way. While vendors of compliant 10GE PHY devices are in the process of utilizing interoperability labs to test and assure interoperability, to date this has not been achieved satisfactorily.
Sproul, Adax, Inc.: We are seeing a clear trend from our suppliers towards more energy-efficient parts. At 10/100/1000 the difference is minimal, however at 10G and emerging 40/100G rates the effect could be dramatic. We expect this trend to continue.
EECatalog: How are developers addressing demands for Fibre Channel over Ethernet (FCoE) in new designs?
Cates, PLX Technology: While the subject of FCoE is best addressed by a vendor in the 10G MAC space, from the perspective of 10GBase-T PHY suppliers such as PLX Technology, FCoE traffic looks the same as regular Ethernet and is transported similarly.
Sproul, Adax, Inc.: Our experience with FC and FCoE in particular is limited to the required co-existence of our I/O and packet-processing cards in the same system as the FC cards. In the past, this has led to bus conflicts and power draws from offending FC cards.
Cheryl Berglund Coupé is editor of EECatalog. com. Her articles have appeared in EE Times, Electronic Business, Microsoft Embedded Review and Windows Developer’s Journal and she has developed presentations for the Embedded Systems Conference and ICSPAT. She has held a variety of production, technical marketing and writing positions within technology companies and agencies in the Northwest.