40Gbps, The Next Frontier For Servers Says Network Specialist

We caught up with Charles Ferland, vice President of BLADE Network technologies to discuss about the company's plans, the strategic partnership it announced with IBM and the future of GbE networking technology worldwide.

1) Mr. Ferland, Can you first and foremost present BLADE Network Technologies to our readers?

BLADE Network technologies (BLADE), the data centre Ethernet switching company, was established as a privately held company in 2006. Today, BLADE is the industry’s number-two supplier of data centre Ethernet switches, with more than 9 million ports installed. BLADE’s 1 and 10 Gigabit Ethernet blade server and top-of-rack switches bring intelligence and speed to the edge of the network where it’s closer to business applications, users and innovation. BLADE products provide the superior performance, ultra-low latency and virtual machine awareness that is ideally suited to handle dynamic demands at the network edge, from private and public clouds to financial services and other HPC applications. BLADE’s lossless, low-latency, low-cost and low-power RackSwitch switches; VMready network virtualisation and BLADE Harmony switch management, connect and unify physical and virtual servers, storage and networks in the world’s largest data centres. BLADE’s Unified Fabric Architecture is a fast, virtual, proven and interoperable converged fabric that provides a roadmap for today’s 1/10 G Ethernet network to tomorrow’s 40/100 G Ethernet environments.

2) One of the company's recent announcements has to do with the adoption of BLADE’s hardware by IBM, can you tell us more about it?

BLADE has been IBMs’ trusted partner for over seven years. Most recently, IBM selected BLADE’s RackSwitch 1 Gigabit and 10 Gigabit Ethernet switches for inclusion in the IBM System x option portfolio. BLADE’s RackSwitch products provide enterprise data centre networking capabilities for IBM System x, including IBM Virtual Fabric, BLADE’s VMready Virtual Machine-aware networking and lossless DCB/CEE networking. In addition to BLADE’s embedded switch products, which are widely deployed Ethernet switches for IBM BladeCenter, IBM has offered BLADE’s RackSwitch as an option for iDataPlex and Cluster 1350 for more than two years. Since April 2008 IBM is expanding its popular RackSwitch offering to make the BLADE switches available for all IBM System x servers including the best-selling x3550 and x3650 rack servers and the new eX5 enterprise systems, as well as to provide an up-stream networking solution for IBM BladeCenter.

3) BLADE is a proponent of 10G Ethernet technology; other than speed, what are the advantages of moving from GbE to 10G?

10G Ethernet enables data centres to add additional intelligence to their networks. Converged data and storage networks use 10G Ethernet to provide the lossless capabilities required by Fibre Channel over Ethernet (FCoE) storage traffic. IBM Virtual Fabric uses BLADE’s 10G Ethernet switches to provide flexible connectivity, virtual port switching and support for converged networking. Compared to traditional multi-adapter 1Gb approaches, IBM Virtual Fabric can reduce acquisition costs by up to 44 percent and energy costs by up to 75 percent, reduce cable clutter by up to 75 percent and provide up to 2.5x more low-latency I/O bandwidth.

4) VMready technology appears to be one of BLADE's more interesting solutions. What can it do that its predecessors can't?

BLADE’s VMready equips BLADE’s switches with Virtual Vision that “sees” virtual machines. This enables business-critical applications to run in server virtualisation environments without running the risk that sensitive data will be exposed to theft or unauthorised access. The application performance will degrade when deployed on virtual servers, or degrade when virtual servers move from one physical server to another. BLADE’s VMready switches are the industry’s only shipping Ethernet switches that enable Virtual Machine mobility across a large data centre environment. They consist of hundreds of physical servers and thousands of virtual servers without requiring an overhaul of the VMware environment, a replacement vSwitch or proprietary packet tagging. BLADE’s VMready switches are ideal for enterprises that are extremely wary about deploying production/transactional applications in virtualised environments that may jeopardize security, restrict access or impair performance, and thus restrict virtualisation to non-business-critical applications or to a few servers in the data center that must be manually administered.

5) How do you see the next 24 months pan out in terms of network technologies globally?

The global market for Ethernet in the data centre remains vital and strong. With more and more servers connecting using 10G, it is only natural to see uplink speeds that are greater. 40G can be used for inter-switch communication or connecting to faster backbones, but will rarely be used to connect servers. The prices for 40G are dropping; it is much more cost efficient to use 40G instead of combining 4 x 10G. In the data centre, bottlenecks are moving from the CPU and memory access to the I/O of the servers. Today’s multi-core servers are now able to sustain a great amount of traffic, requiring fast networks, especially now that virtualisation is widely deployed. Analysts have predicted that the 10G market would double year-to-year in 2010 and 2011. More servers using 10G increases the requirement for 40G and 100G in upstream networks. Fast pipes like 40G are important, but just as important are the applications utilising these infrastructures. With technology like 10G and 40G, we can safely consolidate several different networks over a single wire, using FCoE for example. Technologies like BLADE’s Unified Fabric Architecture and vNIC allow BLADE customers to partition 10G pipes into virtual adapters to accommodate servers but still reducing cabling requirements. Many silicon vendors have made very interesting announcements this year which incorporate 40G components. BLADE, too is developing products using these components. 100G is mainly focused toward backbone interconnects at the moment which offer a few advantages, from a business perspective, to connecting servers. For now, it is safe to assume 100G is mainly focused at the carrier level.