The UNIX Evolution - from Enterprise Viability to
Virtualization
 Looking back, organizations looked to
host workloads on UNIX with an open systems view or premise.
Seeing the marketplace's favorable view of UNIX, prompted several
prominent vendors to enter the fray with their own UNIX variant.
Furthermore, these same vendors promoted Independent Software
Vendors (ISV)developers to embrace their brand of UNIX. This led
to a proliferation of application solutions that addressed needs
throughout the enterprise and across virtually every industry
segment.
Over time, as server hardware manufacturers traded places or
positions of competitiveness in performance, scaling and other
metrics, ISVs began to offer
their solutions in more than one UNIX implementation. As most ISVs
supported multiple implementations of UNIX, and as the UNIX
vendors jockeyed to offer contrasting advanced functions,
organizations found themselves taking on the support of Solaris,
UX, AIX and other implementations.
To further compound the paradigm, larger organizations with a
robust M&A strategy, absorbed other companies and inherited
yet more UNIX implementations, workloads and servers.
Furthermore, in some companies, the relationship between line of
business and their technology services group, led to
dedicated servers per workload per internal funding entity. The
end result was a high degree of server sprawl (as servers were
frequently acquired for a single purpose) and the need to support
different implementations. While UNIX originally promised
openness, once the hardware manufacturers continually advanced
their unique UNIX implementation in the quest to attract clients
and gain share, more specialized skills were needed and few
individuals had the time to master Solaris, UX, AIX and other
implementations. Accordingly, there were very little economies of
scale that could be realized with regard to the number of system
administrators. To clarify, the Solaris group was a contrast to
those proficient in AIX or UX and in most cases, little cross
training occurred due to time constraints and the result was an
internal alignment by and between certain skills and workloads
supported by a given UNIX implementation.
Therefore, over a decade or so server sprawl, under utilization,
multiple implementations, and the need to hire UNIX variant
specific skills aligned to specific workloads....combined to
increase complexity and costs. However, as each manufacturer
continually invested more resources and dollars to further advance
their flavor of UNIX, great advancements in reliability, security,
systems management, availability, and performance were realized by
Sun, HP, IBM and others.
Today, Linux offers a low cost and developer-friendly
environment. However, as much as the Linux implementations
continue to advance, the feature content and advance function
enablement lags behind that offered by the top UNIX providers.
This is okay in many workload instances as the security with Linux
is not the concern or degree of concern it once was. The developer
aids and tools associated with Linux are compelling. End users
feel more hardware brand independence advantages aligning with
Linux. Some feel that Linux delivers more closely to the original
UNIX open oriented promises.
Yet, the top UNIX implementations of today are very advanced and
combined with their hardware platforms, offer "mainframe like"
reliability features and sophisticated virtualization offerings
and a host of advanced functions to improve workload up-time and
to minimize impact during maintenance events. While there is
disparity between organizations as to how they view Linux, UNIX
and other realms, most organizations are developing their own
internal standards or guidelines as to what workloads are to move
to Linux and what workloads should stay with UNIX. Even as Linux
will continue to bite into UNIX share, there is enough interest in
and user commitment to UNIX to promote the continuance of the top
marketplace implementations for the foreseeable future. This fuels
continued manufacturer investments and spurs collaboration by and
between AIX and POWER (IBM example). Performance,
scaling, availability and reliability all are improving by virtue
of continued advancements with the operating system and the
hardware platform.
While organizations refine their view, plans and priorities (with
respect to platforming decisions moving forward), they also are
frustrated with the progress made to date in optimization,
compared to their stated goals from a few years back. In the most
extreme examples, it may not be prudent for an organization to
move to only one UNIX implementation. Some workloads are ISV
sourced and are only supported in a given UNIX OS brand and in a
subset of these examples, either the migration cost is prohibitive
and/or the supporting ISV is not interested in supporting a
different implementation. In these cases, the compromise is to
attempt to standardize to the chosen UNIX implementation while
also making progress with the enterprise optimization goals.
Therefore, decisions are made as to what to migrate to Linux, what
to keep in UNIX and where new development efforts are to be
undertaken (Linux UNIX, or other), while also prioritizing
consolidation and virtualization tasks in the quest for improving
optimization on an ongoing basis. A reasonable projection of
anticipated progress is a reduction in the number of unique realms
to support, a clear picture as to platforming standards/policies,
while reducing costs through optimization (pertaining to the
workloads ripe for consolidation and virtualization).
Sun, HP and IBM all want your UNIX business and each has its story
to tell with respect to server line competitiveness, operating
system performance, advanced functions, high availability and
virtualization. In the historical view, all three of these vendors
have traded places in overall competitiveness / performance, and
in a specific sense viewing the various workload types
(transaction processing, web, database, etc.). Conversely, over
the last five or six years, according to Gartner and many other
independent industry authorities, IBM POWER combined with AIX are
leading overall in performance, scaling, advanced functions and
especially in highly optimized implementations where resource
utilization has reached a higher proficiency level than that of a
lower utilized/distributed reality. An organization might be
facing several other factors that could influence them to prefer
to optimize leveraging HP/UX or Sun/Solaris as their standard,
including in-house skills, relationship with the manufacturer
team, commonality to workloads, preferences, culture, cost to
retrain/retool, etc. Regardless of one's UNIX brand preference,
when considering optimization, most organizations tackle
x86/Wintel prior to moving over to working on consolidating and
virtualizing UNIX. Painting somewhat with a broad brush, one
reason for this is the typical workload complexity associated with
UNIX. The workload classification or priority tier, (read as
mission critical workloads), is often a factor. Back level OS with
UNIX workloads is yet another factor.
Today, the time is at hand for organizations to focus a bit more,
at tackling the optimization of UNIX workloads. Even as Linux
takes on more share, many organizations will continue with
platforming at least some workloads and the more critical
enterprise class workloads to UNIX. Many optimization solution
providers are more focused on or specialized in x86/Wintel. This
stems from these realms being riper during the first waves and
early years of consolidating and virtualizing.
Datatrend Technologies is an optimization solution provider
with both the x86/Wintel and UNIX workload experience. Our team of
specialists have the assessment know-how, and the assessment tools
that combine to ensure the success in efforts geared to optimizing
enterprise class workloads. While past TrendSetter articles have
addressed optimization of Linux and/or x86/Wintel workloads, for
this edition, we wanted to share some information with respect to
the viability of UNIX's role in the enterprise and our realization
that most organizations struggle a bit more when considering
optimization of UNIX workloads. While we have a few clients that
have stated a material commitment to Linux, many of our customers
view UNIX as something they will support, if not embrace for a
very long period of time. The manufacturers seem committed to
improve the argument for or the viability of UNIX based solutions
well into the future.
In addition to the previously stated issues of complexity
involving some UNIX workloads, in some cases migration from one
UNIX brand to another UNIX variant is not a viable option. Either
the application literate resources are not available to perform
the migration, outside firms' price quotations to perform
migration services are prohibitive or the people with the
application knowledge are stretched too thin to perform final
testing. In some real-world examples, this leads to collapsing
Solaris workloads within Solaris, UX to within UX and AIX to
within AIX. Therefore, assessments of UNIX workloads ahead of
optimization, might in some cases, need to yield the output of
multiple views/options. If a given organization wants to move
underutilized UNIX realms (wherever justifiable) to AIX and POWER,
certain applications might be ripe and ideal for that, while
others should remain in Solaris or UX. At the end of the day,
consolidation / virtualization has to have a quantifiable payback
and a professional assessment can consider factors other than
resource utilization, application type, application dependencies,
OS level supporting virtualization or otherwise. Many
consolidation recommendations can determine migration to AIX/POWER
while maintaining other workloads in Solaris or UX.
 Today, the trend has moved toward
consolidation and virtualization, as companies have realized the
cost they pay in space and maintenance to support these unused
resources. Companies also want to create an enterprise environment
that can respond quickly to changing demands. Virtualization
involves a shift in thinking from the physical to the logical,
looking at how to manage a logical pool of resources rather than
individual devices.
Companies first look to change or upgrade their platforms for a
number of reasons, including aging servers, questionable hardware
road maps from vendors, application specific requirements, HW
vendor stability and viability, increased performance/capacity
demands, and budgetary pressures due to energy costs, cost of
software licensing, increased maintenance on aging servers, and
server management. Whatever the driving force, the goal today is
to develop a system that will run better and accomplish more
business with fewer resources, yet still possesses the flexibility
to permit future growth.
Of course these comparisons represent a generalization and can
vary by organization when considering or applying corporate
policies, procedures and/or preferences. If you are interested in
an analysis of your UNIX environment, Datatrend can help determine
if server optimization or consolidation is right for you. In
short, we can provide a comprehensive assessment that will help
you achieve IT infrastructure optimization. To learn more, click
here.
Since during this current era, Gartner and other
independents are positioning POWER and AIX as the leading UNIX
platform, in the quest to illustrate the continued viability of
UNIX to the enterprise and to also illustrate the viability of
optimizing appropriate UNIX workloads, we will explore
advancements and capabilities of POWER and AIX combined.
With PowerVM, IBM has been a leader in developing Virtualization
technologies that provide the ability to support multiple
Operation Systems in a Dynamic LPAR environment. PowerVM
provides users with many options in developing and running
applications, simplifying operations by minimizing processor and
memory resources, and providing efficient management of these
resources, while providing flexibility and portability for future
growth.
PowerVM can help clients decrease their hardware / software costs,
increase efficiencies and optimize workloads, while also providing
the scalability and reliability needed to operate in today's
demanding environments.
Here are a few things Power VM can do for
you:
- Run multiple OS images or instances – including a mix
of: AIX levels, IBM i releases, and Linux distributions.
- Consolidate workloads from multiple physical servers
onto a single POWER6 server, thus producing a server environment
with fewer cores, less memory, and higher utilization. This
consolidation results in a significant decrease in floor space,
power & cooling, and maintenance.
- Create dynamic allocation of server processor, memory,
and IO resources. This allows you to adjust CPU / memory based on
business demand and application needs, with possible allocations
as small as 1/10th of a processor to the server. It also allows
you to share IP and storage resources, improve overall asset
utilization, and manage resources based on business service
requirements.
- Provide complete partition isolation as if running on
physically separate servers. One partition can not have an
effect on another partition.
- Provide in conjunction with the Power Hardware,
extensive RAS functionality, memory management, energy
management, predictive analysis and alert management.
- Better utilization of processors and memory while
reducing the footprint and administrative effects.
- Enhance partition and workload mobility by allowing
real-time movement of workloads without any application downtime.
This is an effective means for managing HW System Maintenance,
Firmware updates, etc. which has traditionally required full
server outages for lengthy periods of time. This can also
effectively reduce server power by combining off-hour loads onto
single servers and powering down others until the demand
increases.
As an example of how PowerVM virtualization can
streamline operations, consider the steps required to add a
server in a traditional environment:
Review requirements
Determine server size
Purchase equipment
Find floor space
Determine power/cooling/UPS
Run cables/acquire patch panel
Schedule delivery
Set-up HW
Apply outstanding firmware/driver upgrades
Burn In
Create appropriate backups
Turn over to user
Compare that to the steps required to configure the same
information under a POWER virtualized paradigm:
Review requirements
Determine resource sizing
Provision new Image Logical Partition (LPAR) – image backups are
taken as part of image creation
Turn over to user
Clearly UNIX virtualization can offer many benefits to a company
but sometimes the task of accomplishing this virtualization is
daunting. The assistance of a partner such as Datatrend can be
invaluable in helping a company achieve its goals. Click here to
discover more about our virtualization services.
|
