This page contains a Flash digital edition of a book.
FOCUS on RACKS AND CABINETS


TOP OF RACK CONSIDERATIONS


Top-of-rack considerations


by Enterprise Strategy Group principal analyst Jon Oltisk


As the enterprise data center has grown in size and importance, the minimum unit of deployment has moved from the individual server to the blade frame - and now to a whole rack. To support this move to rack-level deployment, pre-provisioned racks and scale- out networking architectures have become imperative.


Top-of-rack network designs deliver a standard, flexible and low-cost way to incrementally scale out capacity and automate provisioning and change management. BLADE Network Technologies is leading this transition to on-demand networking for meeting near-term deployment of public, and private, clouds.


In top-of-rack designs, a switch fabric is installed at the end of each row. As racks are configured, a pair of switches is installed at the top of each new rack and wired back to the end of rack fabric (using either 10Gb or 40Gb structured interconnect uplinks). As servers or blade frames are installed in the rack they are wired directly into the top of


rack switches via patch leads.


Top-of-rack designs are designed to wire once and forget. Configuration and routing of LAN and SAN traffic are done centrally - either manually via a console or more likely as part of an integrated virtualization product such as VMware, Microsoft Hyper-V or Citrix Xen. In fact, the more “virtual awareness” in the network the more server configuration tasks can be automated.


For example, BLADE Network’s VMready can detect a VM move from one server to another; when this happens, virtualization management and the network can coordinate to preserve attributes like VLAN or 802.1p tags. With this virtualization intelligence, a large organization can leverage the flexibility and mobility features of server virtualization without breaking the network architecture or opening up security holes.


DATA CENTER SCALE-OUT The top-of-rack converged design is conceived to enable rack loads of servers and storage to be scaled out in a modern enterprise data center. As each new rack is populated, a small set of uplink connections


Rittal – The System.


are made to an end-of-row switch, usually with a fan out ratio of five to one (five top of rack ports for each uplink port).


In a BLADE infrastructure, this can radically improve data center network economics. How? Rack-based networks perform the role of access and aggregation switches, alleviating the need for more expensive external core/aggregation switches within the data center.


This also reduces cabling costs as multiple network tiers are connected within a rack rather than through patch panels under the data center floor.


Switches are acquired just-in-time as the rack needs to be populated, improving capital efficiency and reducing power wasted by orphan ports.


Conventional three-tier switch designs require full network population at data center inception, dramatically impacting capital efficiency and increasing waste power and heat. Newer high performance (40GbE and 100GbE) parts cannot be integrated into the conventional design as all equipment must


IT RACKS


IT COOLING


IT POWER


8


www.datacenterdynamics.com


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20