- Iain Kenney, Director of Product
Marketing at LG Ericsson USA (www.lgericssonus.com), says:
In today’s world, where access to
information is king, the ‘palace’ is under constant renovation. Data centers
are expanding at a rate beyond what anyone would have predicted while the 21st
century data center makes the “massive server farms” of the information age seem
like so much wasted space. Data centers today are optimized to the nth degree
and careful planning is critical to ensuring that these massive data centers
are protected from potential failures and hazards.
Ensuring that accessibility to
the servers and data is as efficient as the floor plan is critical to success;
enabling easy adjustments and optimization is paramount. Software-defined
networking (SDN) will be key to unlocking the full potential of the next
generation 21st century data center.
SDN facilitates on-the-fly network reconfiguration, and even enables
creative server placement for highly-responsive expansion and just-in-time
capacity adjustment. Top of Rack (ToR) architecture, set-up and managed via
software, sets the stage for responsiveness, efficiency and, ultimately,
cost-effectiveness, makes good sense for business.
Evolving
for Efficiency
Today’s networks must be built
for the ultimate in efficiency; software-defined networking is the key to
achieving that efficiency. Whereas it
used to be that servers were organized by specific application or data types or
sets, today, due to the amazing volume of data and activity, every bit of
processing power and every bit of data on every server in every data center
needs to be optimized. Networks have evolved from being composed of small
numbers of single-purpose networked servers to large numbers of highly
virtualized servers, all of which can share pieces of the same application or
data set.
To accommodate that need for
efficiency, the network itself needs to evolve—beyond the networking hardware
itself. Bandwidth and ‘speed’, the
buzzwords for efficient networking in the past, don’t cut it any more:
networking hardware that has all of the right features to enable speed gets
bogged-down by a large, proprietary code base.
That code base overcomes the ‘specs’, slowing the network’s
responsiveness and aggravating users.
The
Search is On
Now, data that used to reside on
a single server in an ‘organized’ fashion can be distributed in an ‘optimized’
form across multiple massively high-performance servers—sometimes all in one physical
data center, sometimes across town, sometimes around the globe. Use of those servers, which are doing lots of
things—and lots of parts of things—needs to be optimized and the left hand, so
to speak, needs to know what the right hand is doing. And, of course, in
addition to use and storage, redundancy and backup need to be managed and
tracked, too. With everything so spread out, information location becomes more
complex. Virtualization holds the key.
Word
of the Day: Hadoop
Introduced as a standard, then
absorbed into the ever-growing collection of trendy tech jargon, Hadoop, quite
simply, describes the distributed processing and cross-sharing of application
data across servers. Designed as an open standard by the Apache Group, Hadoop helps a server farm full
of machines act as one: like a borg, its single consciousness means that
killing one will not affect another. Any
server can be a master to or a servant of any other. This construct makes for the ultimate in
efficiency – in storage usage, and in access.
Factoring-in Failure
Servers are cheap, so building a data center that can ensure redundant storage and fail-safe availability is easily accomplished with simple replication. Access is a different story. The network enables users to access all of that replicated data, so the ToR switches are critical to getting at the data: no matter how many copies of the data are stored, if users can’t access it, it isn’t ‘there.’
Servers are cheap, so building a data center that can ensure redundant storage and fail-safe availability is easily accomplished with simple replication. Access is a different story. The network enables users to access all of that replicated data, so the ToR switches are critical to getting at the data: no matter how many copies of the data are stored, if users can’t access it, it isn’t ‘there.’
In, Across and Between the Racks
Connecting all of those servers through a core networking system gets users to the data and applications they need, re-routes requests when servers go down, and keeps data flowing. Point-to-point networking doesn’t make sense, because the cabling nightmare to make it all happen would be overwhelming. Top of Rack networking facilitates access across racks and racks of servers, often each with 50 or more servers running 10 or more virtual machines—500 devices— each in a single 72U rack. Keeping track of a single rack, let alone a data center full of rows of racks, is a formidable task. Network switches—ToR switches—are the critical link –that connects everyone to everything.
Connecting all of those servers through a core networking system gets users to the data and applications they need, re-routes requests when servers go down, and keeps data flowing. Point-to-point networking doesn’t make sense, because the cabling nightmare to make it all happen would be overwhelming. Top of Rack networking facilitates access across racks and racks of servers, often each with 50 or more servers running 10 or more virtual machines—500 devices— each in a single 72U rack. Keeping track of a single rack, let alone a data center full of rows of racks, is a formidable task. Network switches—ToR switches—are the critical link –that connects everyone to everything.
Network
switches don’t fit the server model of cheap redundancy. Cheap, so-called ‘white-box’ switches can’t
be relied upon to shoulder the weight of a mission-critical network, but buying
the most expensive isn’t the best alternative, either. Ironically, high-priced solutions come not
just with a higher initial price tag, but also incorporate proprietary
solutions that can lock the data center into unintentional allegiance that
locks-up the budget.
In
today’s high-volume, high-accessibility data center, density is so massive that
the cables alone for a traditional point-to-point network would be
overwhelming. ToR networking aggregates
access at each rack, connecting all of the servers in the rack to each and
every other rack. Managing the access
within and across racks is a challenge accommodated by solid switching
solutions: 1G or 10G connections accomplish efficient server-to-server links
within the rack; 1G and 10G manage Northbound traffic, rack to core; and
connections of up to 40G manage East-West
(rack to rack) traffic, in a 2-3 level mesh network.
Putting
the Pieces in Place: High-Capacity, High-Performance, High-Availability
Switches
Evolution
is bringing us to the next generation of network devices. Those devices will take better advantage of
the advanced, high-speed network hardware architecture—and advances to it—by
using a slimmer and more open code base that enables data center managers to
define and implement new architectures and change the architecture as business
needs change. And, it will make it easy to make those changes by utilizing SDN.
Whereas it used to be that server
capacity and network bandwidth were the critical factors. Now, being constrained by inability to make
efficient use of the massive capacity and bandwidth that are available, we need
more sophisticated solutions for traffic management. Constant monitoring and on-the-fly adjustment
are the way to efficient computing, and that can best be accomplished when data
center managers have the tools to expand, contract, and re-configure the data
center architecture easily and without waiting for a vendor to act on a tech
support or modification request job ticket.
No one knows what will work best
for a given network better than that network’s administrator—and fine-tuning,
combined with a little bit of trial and error, often makes good even
better. High-quality, high-capacity,
highly-reliable, standards-based, SDN configurable network switches, within, on
top of and between racks of servers, are the answer. SDN provides the potential for a
‘do-it-yourself’ framework with dynamic drag-and-drop configurability that
makes the most dynamic, responsive network a reality. Acknowledging the reality
of today’s dynamic data center, SDN provides a responsive platform for data
center growth and adjustment. Separating
the configuration and programming of ”the network” from the underlying hardware
enables managers to avoid vendor lock-in, lower costs and, most importantly, to
be more responsive to user needs.
Commodity Plus: LG-Ericsson USA –
Solid Hardware and Warranties with SDN
Designed for Dynamic Environments
Designed for Dynamic Environments
LG-Ericsson
USA brings the best of both worlds to bear for Top of Rack switching in today’s
dynamic data center with high-density, ultra-high bandwidth, standards-based
ToR switches, backed with great software and firmware, superior warranties and
US-based support.
LG-Ericsson
USA’s powerful, high-bandwidth ToR switches keep it all flowing—inside the
central office and out to the mobile universe.
The company’s 1G, 10G, 10G + 40G and 40G energy-efficient,
rack-mountable, hot-swappable, standards-compliant OpenFlow-ready switches
ensure efficient, reliable data centers.
With OpenFlow enabling circuit-switched networks in packet-switched
environments for easy software-driven management of data flow, our ToR switches
bring simplification, reliability, support and hot-swappable components to software-controlled
networks. And, LG-Ericsson shows our
commitment to the customer by bringing all of this functionality, efficiency in
an affordable, standards-based, software configurable (SDN) solution.


No comments:
Post a Comment