Day 3/ Session

The next datacenter

Dileep Kumar, Director- Product Management (INDIA & Neighboring Countries), ADC Krone made a presentation about understanding and preparing for tomorrow's technologies

If you look at bandwidth demand forecasts, the need for bandwidth is being driven by infrastructure upgrades and fueled by video. Kumar touched upon the trends and technologies that we can expect to see in the near future including support for 10 Gigabit Ethernet and beyond – 40Gig, 100Gig standards under development by IEEE; a shift towards fiber; data dynamics and higher expectations in terms of reliability; design considerations and expectations; compliance with the datacenter tia-942 standard as well as green initiatives and energy efficiency.

He expanded on the first trend, namely that of support for 10 Gigabit Ethernet and beyond and talked about the IEEE 802.3ba 40Gbps and 100Gbps Task Force wherein the IEEE Higher Speed Study Group (HSSG) has assembled as the IEEE 802.3ba task force with the goal of assembling a standard for release in 2010. The first (of many) draft standards is set to be circulated for comment. The task force is currently behind schedule.

The second trend was about the shift towards fiber. The increasing demand for fiber in datacenters was, he felt, due to its lower power consumption at higher data rates, more efficient cooling, cheaper power bills, lower carbon footprint, higher densities of cabling per bay and lower cost of ownership.

Fiber Applications included Fiber Channel which is the primary protocol for SAN, meeting the needs of all types of storage device. The trend is for 2x fiber channel and above which needs to be supported by 10G cabling—either OM3 or OS1. FCoE and iSCSI were both gaining ground. iSCSI packages multiple SCSI data channels into IP packets and pushes them out onto the WAN, over single-mode fiber and unlike Fiber Channel it can be transmitted over long distances using existing infrastructure.

Next he touched upon data dynamics and higher reliability expectations where applications have grown in number & complexity and servers are often dedicated to specific applications. The answer lies in virtualization by transforming farms of servers, storage and networking into pools of computing resources.

This connects with cabling, he argued, because extensive background swapping of data to disk storage calls for greater bandwidth in cabling; encapsulating entire systems into single files means packet sizes will increase and virtual machines need to run safely i.e. stressed network physics should not impact data transmission.

He reminded the audience that more was at risk than ever before if physical layer reliability was not established and guaranteed.

The fourth trend was that of design considerations and expectations where the network engineer’s concerns within the datacenter had to be addressed. Here the issue was of scalability in terms of equipment density, cabinets, frames; fast and accurate moves, adds, and changes; thermal issues that are a problem in most datacenters especially when Blade Servers have been deployed. Poor Air Flow is a problem in many datacenters where reliability and uptime are a problem. You can’t afford any downtime even during expansion, he said. Cables blocking air inlets and exits will raise the temperature of switches and servers lowering their reliability.

Finally, Kumar touched on the topic of green initiatives and energy efficiency where he quoted an IDC report to state that by 2010, for every $1 spent on hardware, 70 cents will be spent on power and cooling; by 2012, for every $1 spent on hardware, $1 will be spent on power and cooling and 46% of data center managers don’t know how much they spend on power and cooling.