Question 1. Vendors
are racing to lead the movement towards a software‐defined
data centre. Where are we up to in this journey, and how far are we from seeing
this trend widely adopted?
Considering
most organisations have still not fully virtualized or moved towards a true
Private Cloud model, SDDC is still in its infancy in terms of mainstream
adoption and certainly won’t be an overnight process. While typical early
adopters are advancing quickly down the software-defined route these are mostly
organizations with large scale multi site data centres who are already mature
in terms of their IT processes. Such large scale organisations are not the norm
and while the SDDC is certainly on the mindset of senior IT executives,
establishing such a model requires several key challenges and tasks.
Typical
environments are still characterised by numerous silos, complex & static configurations
and partially virtualized initiatives. Isolated component and operational silos
need to be replaced with expertise that cover the whole infrastructure so that
organisations can focus on defining their business policies. In this instance
the converged infrastructure model is ideal as it enables the infrastructure to
be managed, maintained and optimised as a single entity by a single silo. Subsequently
such environments also need to dramatically rearrange their IT processes to
accommodate features such as orchestration, automation, metering and billing as
they all have a knock on effect to service delivery, activation and assurance
as well as change management and release management procedures. The SDDC
necessitates a cultural shift and change to IT as much as a technical one and
the latter historically always takes longer. It could still be several years
before we really see the SDDC be adopted widely but it’s definitely being discussed
and planned for the future.
 |
| You can't have a successful software-defined model with a hardware-defined mentality |
Question 2. Looking
at all the components of a data centre, which one poses the most challenges to
being virtualized and software-defined?
The majority of data centre
components have experienced considerable technological advancements in past few
years. Yet in comparison to networking, compute and hypervisor, storage arrays
still haven’t seen that many drastic changes beyond new features of
auto-tiering, thin-provisioning, deduplication and the introduction of EFDs. Moreover
Software Defined’s focus is applications and dynamically meeting the changing
requirements of an application and service offering. Beyond quality of service
monitoring based on IOPS and backend/frontend processor utilisation, there are
still considerable limitations with storage arrays in terms of application
awareness.
Additionally with automation
being integral to a software-defined strategy that can dynamically shift
resources based on application requirements, automation technologies within
storage arrays are up to now still very limited. While storage features such as
dynamic tiering may be automated, they are still not based on real-time metrics
and consequently not responsive to real-time requirements.
This leads to the fact that
storage itself has moved beyond the array and is now encompassed in numerous
forms such as HDD, Flash, PCM and NVRAM etc. each with their own characteristics,
benefits and challenges. As of yet the challenge is still to have a software
layer that can abstract all of these various formats as a single resource pool.
The objective should be that regardless of where these formats reside whether
that’s within the server, the array cache or the backend of the array etc. they
can still dynamically be shifted across platforms to meet application needs as
well as provide resiliency and high availability.
Question 3. Why
has there been confusion about how software-defined should be interpreted, and
how has this effected the market?
Similar to when the Cloud concept first emerged in the
industry, the understanding of the software-defined model quickly became
somewhat blurred as marketing departments of traditional infrastructure vendors
jumped on the bandwagon. While they were quick to coin the Software-Defined
terminology to their offerings, there was little if anything different to their
products or product strategy. This led to various misconceptions such as
software- defined was just another term for Cloud, if it was virtualised it was
software-defined or even more ludicrously that software-defined meant the
non-existence or removal of hardware.
To elaborate, all hardware components
need software of some kind to function but this does not necessitate them to be
software-defined. For example Storage arrays use various software technologies
such as replication, snapshotting, auto-tiering and dynamic provisioning. Some
storage vendors even have the capability of virtualising third party vendor
arrays behind their own or via appliances and consequently abstracting the
storage completely from the hardware whereby an end user is merely looking at a
resource pool. But this in itself does not define the array as software defined
and herein lies the confusion that some end users face as they struggle to
understand the latest trend being directed at them by their C-level execs.
Question 4. The
idea of a software-defined data centre (virtualizing and automating the entire
infrastructure wildly disrupts the make-up of a traditional IT team. How can CIOs
handle the inevitable resistance some of their IT employees will make?
First and
foremost you can’t have a successful Software- defined model if your team still
have a hardware-defined mentality. Change is inevitable and whether it's
embraced or not it will happen. For experienced CIOs this is not the first time
they've experienced this technological and consequently cultural change in IT.
There was resistance to change from the mainframe team when open systems took
off, there was no such thing as a virtualisation team when VMware was first
introduced and only now are we seeing Converged infrastructure teams being
established despite the CI market being around for more than three years. For
the traditional IT teams to accept this change they need to recognise how it will
inevitably benefit them.
Market research is unanimous in its conclusion that currently IT administrators are far too busy
doing maintenance tasks that involve firefighting "keeping the
lights" on exercises. Generally figures point to a 77% mark of overall
time spent for IT admin on doing mundane maintenance and routine tasks with
very little time spent on innovation, optimisation and focus of delivering
value to the business. For these teams
the software-defined model offers the opportunity to move away from such tasks
and free up their time enabling them to be
proactive as opposed to reactive. With
the benefits of orchestration and automation,
IT admin
can focus on the things they are trained and specialised in such as delivering
performance optimisation, understanding application requirements and aligning
their services and work to business value.
Question 5. To what extent does a software-defined model
negate the need to deploy the public cloud? What effect will this have on the market?
The software defined model shouldn't and most likely won't negate the
public cloud, if anything it will make its use case even clearer. The SDDC is a natural evolution of cloud, and particularly the private cloud. The
private cloud is all about IT service consumption and delivery of IT services
whether this be layered upon converged infrastructure or self assembled
infrastructures. Those that have already deployed a private cloud and are also
utilising the public cloud have done so with the understanding and assessment
of their data; it's security and most typically it's criticality. The software
defined-model introduces a greater level of intelligence via software where
application awareness and requirements linked to business service levels are
met automatically and dynamically. Here the demand is being dictated by the
workload and the software is the enabler to provision the adequate resources
for that requirement.
Consequently organisations will have a greater level
of flexibility and agility to previous private cloud and even public cloud deployments,
thus providing more lucidity in the differentiation between the private and
public cloud. Instead of needing to
request from a cloud provider permission, the
software defined model will provide organisations
on-demand access to their data as well as independently dictate the level of
security. While this may not completely negate the requirement for a public
cloud, it will certainly diminish the immediate benefits and advantages
associated with it.
Question 6. For
CIOs looking for pure bottom-line incentives they can take to senior
management, what is the true value of a software-defined infrastructure?
The true
value of a software defined model is that it empowers IT to be a true business
enabler. Most business executives still see IT as an expensive overhead as
opposed to a business enabler. This is
typically because of IT’s inability to respond quicker to ever changing service
requirements, market trends and new project roll-outs that the business
demands. Much of this is caused by the deeply entrenched organizational silos
that exist within IT where typical infrastructure deployments can take up to
months. While converged infrastructure solutions have gone some way to solving
this challenge, the software defined model builds on this by providing further
speed and agility to the extent that organisations can encapsulate
their business requirements into business delivery processes. In this instance
infrastructure management processes become inherently linked to business rules
that incorporate compliances, performance metrics and business policies. In
turn via automation and orchestration these business rules dynamically drive
and provision the infrastructure resources of storage, networking and compute
in real time to the necessary workloads as the business demands it.
Question 7. To
what extent will a software-defined infrastructure change the way end-users
should approach security in the data centre?
A
software-defined model will change the way data centre security is approached
in several ways. Traditional physical
data center security architecture is renowned for being inflexible and complex
due to its reliance on segmented numbers of dedicated appliances to provide numerous
requirements such as load balancing, gateways, firewalls, wire sniffers etc.
Within a software-defined model, security can potentially not only be delivered
as a flexible and agile service but also as a feature that’s built into the
architecture. Whether that is based on an approach of security being embedded
within the servers, storage or network, a software-defined approach has to take
advantage of being able to dynamically distribute security policies and
resources that are logically managed and scaled via a single pane.
From a
security perspective a SDDC provides immediate benefits. Imagine how simplified
it will become when automation can be utilized to restructure infrastructure
components that have become vulnerable to security threats? Even the automation
of isolating malware infected network end points will drastically simplify
typical security procedures but will then consequently need to be planned for
differently.
Part of that
planning is acknowledging not just the benefits but the new types of risk they
inevitably introduce. For example, abstracting the security control plane from
the security processing and forwarding planes means that any potential
configuration errors or security issues can have far more complex consequences
than in the traditional data centre. Furthermore centralising the architecture
ultimately means a greater security threat should that central control be
compromised. These are some of the security challenges that organisations will
face and there are already movements in the software defined security space to
cater for this.
Question 8. Where
do you see the software-defined market going over the next couple of years?
The concept
of the SDDC is going to gain even more visibility and acceptance within the
industry and the technological advances that have already come about with Software-Defined
Networking will certainly galvanise this. Vendors that have adopted the
software-defined tagline will have to mature their product offerings and
roadmaps to fit such a model as growing industry awareness will empower
organizations to distinguish between genuine features and marketing hyperbole.
For
organisations that have already heavily virtualized and built private clouds the
SDDC is the next natural progression. For those that have adopted the converged
infrastructure model this transition will be even easier as they will have already
put the necessary IT processes and models in place to simplify their infrastructure
as a fully automated, centrally managed and optimized baseline from which the
SDDC will emanate from. It is fair to say that it won’t be a surprise to see a
lot of the organisations that embraced the converged infrastructure model to
also be the pioneers of a successful SDDC.
The above interview with Archie Hendryx is taken from the May 2014 issue of Information Age:
