Software Defined Networking

RESEARCH
PAPER
Software Defined
Networking:
The solution’s here, but nobody’s listening
April 2014
Sponsored by
Software Defined Networking: The solution’s here, but nobody’s listening
CONTENTS
Executive summary
p3
Scoping the survey
p3
Surveying the network
p4
Planning for change
p6
The network without SDN
p7
The network with SDN
p8
A problem solved?
p8
A solution shunned
p9
Conclusion
p10
About the sponsor, HP
p11
This document is property of Incisive Media. Reproduction and distribution of this publication in any form without
prior written permission is forbidden.
2 Computing | research paper | sponsored by HP
Software Defined Networking: The solution’s here, but nobody’s listening
Executive summary
Courtesy mainly of regular increases in raw bandwidth plus advances in traffic shaping and routing
technology, networks have, until recently, managed to keep pace with most IT developments.
However, that may not be true for much longer. Faced with the triple whammy of cloud computing,
spiralling mobile demand and, more recently, the Internet of Things, infrastructures designed for
an era of more modest and predictable traffic growth are starting to show their age and run out of
steam.
Not only are these trends becoming increasingly important and onerous – and bear in mind we’ve
only seen the start of the Internet of Things so far – they are all completely dependent on the
underlying network technology. No network equals no cloud, no mobile apps, no BYOD or Internet
of Things and, you might add, no big data, distributed computing, social media or any of those
other good things.
Forward thinking developers and vendors spotted this weak link some time ago and began to
apply the same virtualisation technologies that revolutionised server processing and storage to
the networks that connect servers, storage and clients together. As a result the major networking
vendors have all lined up behind what has been dubbed Software Defined Networking (SDN)
confident that it will be the “next big thing” rolling down the tracks and into the data centre.
Customers, however, seem reluctant to give up their traditional network comfort blanket. Indeed
it is fair to say that, apart from a few large data intensive sectors such as telecoms, SDN has yet to
make its mark in the wider corporate world. Possibly because of an impression of it as a high-end
solution suited only to telcos and other large organisations or, perhaps, because of worries over
being locked into proprietary protocols, potential teething problems or the cost of having to rip
and replace existing hardware.
All of which is perplexing as, the issues SDN is designed to address are real and pressing with
companies across the board actively looking for solutions and wanting them now.
The aim of this white paper, therefore, is to firstly understand the networking problems companies
face and the plans they have for addressing them. But secondly, to discuss how SDN can help,
why vendors believe it should be a credible part of those plans and why customers appear not be
listening.
Scoping the survey
This investigation into why Software Defined Networking has yet to capture the corporate
imagination is based on an online survey of 100 Computing subscribers. Over half were from
companies with 500 to over 5,000 employees, spread across multiple sectors from education
and healthcare through banking to manufacturing and public sector organisations. Respondents
were also selected as involved in the specification, deployment or management of networking
technology and their organisations all had in-house server rooms or data centres of some
description.
Computing | research paper | sponsored by HP 3
Software Defined Networking: The solution’s here, but nobody’s listening
Surveying the network
Unsurprisingly the survey reflected widespread adoption of the kind of technologies now regarded
as essential within the modern data centre. To this end the companies involved were found to
employ multiple servers with 45 percent of respondents reporting populations of 100 or more
across their localities. As might be expected, too, the use of server virtualisation – now a very
mature technology – was equally widespread (Fig. 1), with virtualisation now very much a way of
life for most.
Fig. 1 : Physical and virtual severs
HOW MANY PHYSICAL SERVERS DOES YOUR ORGANISATION OPERATE?
2% 2%
9%
13%
0-5
5-10
11-20
15%
21-50
51-100
100+
45%
Don’t know
13%
HOW MANY VIRTUAL SERVERS DOES YOUR ORGANISATION OPERATE?
11%
5%
4%
4%
0-5
5-10
18%
11-20
21-50
51-100
100+
48%
12%
4 Computing | research paper | sponsored by HP
Don’t know
Software Defined Networking: The solution’s here, but nobody’s listening
Looking further afield MPLS and VPLS were found to predominate when it came to wide area
networking. Plus there was a high take-up of cloud services, despite the relative immaturity of
the technologies and services available in this field. Indeed, around two-thirds reported routine
deployment of some kind of cloud technology, with private clouds in the majority (52%) while
35 percent said they used public Software as a Service (SaaS) solutions and 12 percent
Infrastructure as a Service (IaaS) implementations (Fig. 2).
Fig. 2 : Which types of cloud computing does your organisation use?
Private cloud
52%
Public cloud - SaaS
35%
Public cloud - IaaS
12%
Don’t know
7%
Public cloud - Paas
7%
We don’t use cloud
28%
*Respondents could select multiple answers.
Respondents also reported good awareness of an even more recent trend – that of Bring Your Own
Device (BYOD) – and the need to be clear as to where the organsiation stood on the connection of
personal mobile devices to the corporate network. In fact over half (52%) said they had a BYOD
or mobile access policy already with a further 23 percent planning to develop one over the next
12 months (Fig. 3, see next page).
Computing | research paper | sponsored by HP 5
Software Defined Networking: The solution’s here, but nobody’s listening
Fig. 3 : Do you have a BYOD or mobile access policy?
25%
Yes
No – but we are planning to
within 12 months
52%
No
23%
Again this was far from surprising and, as confirmed in other Computing surveys mobile access
continues to be a major networking trend with machine-to-machine communication and the
Internet of Things further adding to the burden being placed on network in infrastructures.
Planning for change
When asked if they were planning to upgrade their network infrastructures to cope with increasing
complexity and demand, respondent answers were very illuminating. Not least because those
definitely not doing so were in the minority.
A mere 17 percent said they had no plans to upgrade their networks with a third (33%) planning an
upgrade within 12 months, many sooner rather than later. Another 11 percent expected upgrades
to get the green light within an 18 month window and a further 21 percent likely to follow suit
thereafter (Fig. 4).
Fig. 4 : Are you planning to upgrade your network infrastructure?
18%
7%
7%
Yes – within 3 months
Yes – within 6 months
19%
17%
Yes – within 12 months
Yes – within 18 months
Yes – at some point beyond 18 months
No
Don’t know
20%
11%
6 Computing | research paper | sponsored by HP
Software Defined Networking: The solution’s here, but nobody’s listening
When it came to reasons why they were upgrading the top two answers were the need to replace
infrastructure at the end of its life, i.e. replacing old and outdated kit, and the perennial problem
of needing more bandwidth. Better application performance was also cited with, as a result, the
majority of upgrades reasonably likely to include the adoption of the latest 10GbE switching fabrics
with many going further to 40GbE and beyond.
Support for converged communications proved yet another popular reason for upgrading. And it is
fairly safe to assume that many will also be enhancing and extending their wireless infrastructure
to cope with burgeoning mobile demands, possibly switching to the new but now ratified 802.11ac
technology, also referred to as Gigabit WiFi.
Mergers and moving to larger (or in some cases smaller) data centres were among the other
reasons why networks needed to be upgraded. Whatever the reason, however, the general picture
is one of widespread and ongoing infrastructure upgrades in progress across the whole corporate
spectrum. A picture which begs the question why, when it would seem like the ideal time to, at the
very least, evaluate SDN, so few companies seem to be lining up to take advantage of what it has
to offer?
A mere six percent of respondents to the survey were actively considering SDN with a worrying
39 percent claiming that they had never heard of it, clearly highlighting the need for more
education and marketing efforts from the vendors involved. Unfortunately we can’t fill that
void in one short white paper, but we can briefly summarise what SDN is all about and prompt a
discussion as to how it can be used to address the issues causing such frenetic upgrading activity.
The network without SDN
If you’re reading this then, chances are, you will be familiar with how traditional networks are
constructed. Essentially the hardware involved is self-contained and does everything itself. Not
just the physical switching and routing of network traffic but also the hosting of software to
decide where, when and how that traffic should be directed, the priority given to different traffic
streams, the level of security to apply and so on.
In the jargon of the trade, every switch and every router on a network contains both a forwarding
plane (to handle the physical task of forwarding network traffic) and a control plane (to manage
how, when and where traffic is forwarded).
The problem is that this integrated and self-contained approach doesn’t scale well and, on larger
deployments especially, can be very cumbersome. Not least because of the need for an additional
layer of communication between switches and routers to make sure they all handle traffic in the
same way, applying the same forwarding and security policies and so on.
Changing the way the network is structured and works can, as a result, be a complex, labourintensive task with a high risk of service disruption. Plus, because the control plane operates in
isolation on the forwarding hardware itself, the network lacks visibility and understanding of what
it’s being used for, particularly, when it comes to application requirements.
Computing | research paper | sponsored by HP 7
Software Defined Networking: The solution’s here, but nobody’s listening
The network with SDN
On a network based on SDN technology the forwarding plane is dedicated solely to that one task
– forwarding traffic around the network. The control plane managing the process, meanwhile,
is virtualised and implemented in software. This makes for a much more scalable and flexible
network, for example allowing rapid, even, dynamic topology changes to be made without the
need to over-provision ports as with traditional network architectures.
Administration also becomes much simpler with SDN as traffic management and security policies
only have to be defined once. Moreover, common management tasks can be automated and
with visibility of network traffic at the application level, business logic used to manage network
structure and behaviour.
With SDN the network morphs from a static monolith to a much more fluid beast able to
dynamically adapt to changes in demand.
At least that’s the theory.
A problem solved?
Of course there’s a lot more to SDN than summarised above. But, assuming the benefits outlined,
let’s just consider how it stacks up when compared to what the recent Computing survey identified
as the issues keeping network managers awake at night.
As already mentioned, the top reasons for upgrading a network appear to be hardware coming to
the end of its life and the need for more bandwidth. In which case it would seem logical to look at
alternatives, like SDN, rather than simply continuing with more of the same stuff unable to meet
current bandwidth needs, let alone cope with long term growth with any certainty.
A further but less obvious inference to be made is the amount of change there is with continual
and ongoing need for network upgrades just about everywhere. Networks need to change all the
time, an issue SDN addresses by making it possible to restructure and add to the network without
having to over provision up front or continually rip and replace infrastructure components simply
to provide more bandwidth.
SDN scores highly in that extra bandwidth can be dynamically assigned by the virtualised control
plane and traffic flows prioritised to optimise the available switching capabilities. Plus, where
extra hardware is needed, additions and changes can be made without the need for labourintensive cabling and complex configuration work.
Next on the list, just over a third (35%) of respondents also highlighted a desire to improve
application performance as a reason for upgrading their networks. By which, of course, most mean
throwing bandwidth at the problem, which may or may not land on target. Similarly, although it
is possible to logically segment and prioritise traffic on a conventional network, a great deal of
experimentation and skill is required to get the right balance without disrupting existing levels of
service.
Again, this is an issue SDN is expressly designed to address, through APIs that allow the
virtualised control plane to communicate directly with applications wanting to use the network.
More than that, it can dynamically assign bandwidth and adapt the way traffic is prioritised to
suit, and do so automatically using simple business-based policies to cope with new as well as
existing application requirements.
8 Computing | research paper | sponsored by HP
Software Defined Networking: The solution’s here, but nobody’s listening
Those respondents who said they were considering SDN seemed well aware of these benefits.
Indeed, when asked why they were investigating SDN, the top answer was increased scalability
and agility, followed by the ability to automate manual functions, increase network flexibility and
make their networks application aware.
Paradoxically, however, companies considering SDN at all were in the minority so of equal interest
are the responses given by those not looking to deploy SDN in the near future. As already reported,
over a third claimed not to have even heard of the term while the remainder highlighted a variety
of misconceptions as well as general lack of understanding of the technologies involved.
A solution shunned
The top reason for eschewing SDN was a lack of in-house skills, but that would be the case with
any new development and, by its very nature, managing an SDN infrastructure should be easier
than having to grapple with a traditional set-up. A similar number also thought it too new a
technology for them, while others saw it as too expensive or, more likely, that there was no clear
ROI argument – something vendors clearly need to address (Fig. 5).
Fig. 5 : Why do you think you will not deploy SDN in the near future?
Lack skills in house
33%
It’s immature – standards are
not settled yet
31%
Too expensive, unclear ROI
29%
We’d have to rip and replace
our current infrastructure
22%
Other (please specify)
20%
It will add complexity in our
multi-vendor environment
16%
Disruption and downtime
while deploying
13%
It’s only suitable for really big
players, like telecoms companies
11%
Possible performance issues
9%
It’s cumbersome to deploy
7%
*Respondents could select
multiple answers.
Computing | research paper | sponsored by HP 9
Software Defined Networking: The solution’s here, but nobody’s listening
Others thought they would need to rip and replace their existing infrastructure to implement
SDN. An argument that has an element of truth, but much the same claim could be levelled at
a conventional network uplift. Moreover, the rejoinder from SDN vendors would be that, once on
the SDN path, future changes should be much less disruptive. Similarly, any network upgrade can
cause disruption and downtime – yet another concern – a risk that would, likewise, be greatly
minimised post-SDN deployment.
A mere 11 percent saw SDN as a high-end solution solely for big companies, indicating that part
of the message at least seems to be getting through. A larger number, however, thought it would
add complexity, especially in a multi-vendor environment. And that despite the majority of SDN
vendors lining up to support open technologies and standards, such as OpenFlow from the Open
Networking Foundation (ONF).
Some respondents even imagined SDN to be little more than a marketing term, illustrating the
breadth of ignorance that exists when it comes to what others see as the “next big thing” as far
as networking is concerned.
The networking vendors appear to have the answers, unfortunately their customers have yet to be
convinced of the questions and just aren’t listening.
Conclusion
The overriding inference to be drawn from the results of the Computing survey on which this white
paper is based is that vendors keen to promote the benefits of SDN need to address what appears
to be a fundamental paradox.
Potential customers quite clearly exist and, quite clearly face issues when it comes to networking
bandwidth, change management and application awareness that SDN is designed to address.
At the very least, SDN ought to be on the list of possible upgrade paths for such customers, but
that isn’t the case. Instead a lack of awareness and understanding is leading those with the
most to gain to stick with the devil they know and, some would say, perversely compound their
problems by upgrading to more of the same rather than try what many see as a new and unknown
quantity.
Blame for this paradox needs to be apportioned equally. SDN is a relatively new technology and
vendors, resellers and other “experts” could do a lot more to educate users, not just to the benefits
but to the nuts and bolts of how SDN works, how best to implement it without disrupting services
and how much it will actually cost.
Customers, though need to work a little harder as well and take their heads out of the traditional
network sand to discover not just what SDN can do for them but what they can do with SDN.
Maybe then it really will become the “next big thing” and take what may claim as its rightful place
in corporate datacentres everywhere.
10 Computing | research paper | sponsored by HP
Software Defined Networking: The solution’s here, but nobody’s listening
About the sponsor, HP
As a global leader in information technology, HP applies new thinking and ideas to help remove
complexity and simplify technology experiences. We strive to continuously improve the way our
customers live and work through sophisticated technology that is easy to use and manage.
HP’s quality engineering and reliable service and support gives all of our customers from individual
consumers to the largest enterprises greater freedom to focus on their priorities and confidence
to reach their goals. Because ultimately we know that each time we create a truly valuable
experience, we are helping people achieve results that really count.
For more information:
Visit:
www.hp.com/uk
Computing | research paper | sponsored by HP 11