By Barb Goldworm
published: Friday, May 02 2008
Barb Goldworm, President and
Chief Analyst, Focus Consulting
Things to watch out for...
In my last column, I wrote about things to
watch - and discussed some of the many changes relative to virtualization that
kicked off the beginning of 2008. With all the changes happening on the vendor
and product front, it's important not to take your eye off the ball on day to
day stuff. So what are the lessons from
virtualization implementations currently underway? Here are a few thoughts in the "things to
watch out for" category:
Consolidation ratios
Although multi-core processors and extended
memory are allowing more virtual machines per physical machine, performance is
not the only consideration when consolidating.
First, it is
critical to remember the "all eggs in one basket" issue. If you put lots of
eggs in one basket, make sure you shore up the basket. Planning decisions about high availability
(HA), hardware redundancy, clustering, and possibly even full system fault
tolerance should be part of the plan from day-one, so that if/when you have a
server failure, your infrastructure will handle it the way you need it to, for
the applications running on that server.
For every application that you consider moving to a virtual
infrastructure, you need to consider your "tolerance for downtime" - and where
it falls on the availability continuum - and spend your availability dollars
accordingly.
Second,
remember that if you are using High Availability restart capability, such as
VMware HA, a server failure will mean restarting all the VMs that are running
on that server. Therefore, if/when there is a failover/restart process, the
number of virtual machines running on that physical machine will determine how
long the full restart takes. The VMs at
the end of the list may not get restarted for quite a while. At some point, consolidating more may be too
many.
Security
Security is always a concern, but
unfortunately usually too far down the planning path. There are always security issues in IT
implementations, but we're not usually very good about planning it up front -
it seems to always be an afterthought. And as we add more function, we add
complexity and often more security challenges. So for example, VMware
VirtualCenter offers lots of management capabilities, but many users forget
about the security ramifications involved. When planning your implementation
and your Virtual Center management functions, consider who can do what to
whom. More than once, we've heard that
these types of security issues down the road have bogged down the process and
slowed the move to production.
On another
security note, security issues involving the virtual network should also be
considered. As you consolidate, parts of the physical network will become a
virtual network which exists inside the physical server(s). Then what happens
to the previously physical security devices?
What's guarding the virtual switches?
A number of new products are emerging here, such as Reflex and Catbird,
to perform that type of security in the new virtual infrastructure. VMware has
also just launched its VMsafe initiative to allow approved partners to have
access to VMware code and work on making the virtual infrastructure at least as
secure as (and maybe more secure than)
the current physical infrastructure. This is definitely an area to watch
and watch out for.
Servers vs. Desktops
As IT shops
start the move from virtualizing servers to virtualizing desktops, there often
comes an interesting blurring of lines. Many enterprises have been using what
has been known as Server Based Computing
(SBC) using Citrix Presentation Server (now XenApp) or Microsoft Terminal
Services for years. This model is actually using a form of virtualization
sometimes called presentation
virtualization, where the application runs in the server, and the
presentation layer is virtualized and remoted to the user on a PC or thin
client. The Citrix installed base is
over 200,000 and is by and large a very satisfied customer group. On the server
virtualization side, the bulk of the installed base to-date is running VMware,
and is also a very satisfied group.
As shops
take a new hard look at what applications should run as Server Based Computing
published applications under Citrix XenApp, and what users require a full virtual desktop (VM), the lines blur, and the
decisions are often not clear cut. And
since virtualization, in general, abstracts logical functions away from the
hardware, both of these options help address the growing need for user mobility
-the ability for users to go anywhere within the organization and still be able
to access their applications and their data.
For organizations running both VMware and Citrix, generally the server
folks love VMware and the desktop folks love Citrix. And now with Citrix offering XenDesktop
running on a back-end infrastructure based on either Citrix XenServer, VMware
or Microsoft Hyper-v (when it goes GA in Q3), the lines blur even further.
The long
term result should be the best of both worlds - the ability to run functions
where they best fit, rather than force fitting them to an existing
solution. More choices are good for
users and will result in better products over time. What to watch out for here? The politics of
course. With multiple good options to
choose from, it will be important to not have this become a battle of the silos
within your IT organization - between the server team and the desktop team.
Creating cross-functional virtualization teams with support from higher up in
the organization can help prevent this from becoming a political power
struggle.
Storage
Often
overlooked in the move to server and desktop virtualization, storage is a key
factor, both in the success and potentially the failures of virtualization
projects. First, it is important to understand that in order to take advantage
of many of the advanced management capabilities and benefits of virtualization,
you will need to move to networked storage of some type. While it is possible to install server
virtualization products on direct attached storage, features such as live
migration (VMotion or XenMotion) and all the advanced features that use it
(e.g. VMware Dynamic Resource Scheduler) require the physical hosts to have
access to shared storage, meaning SAN or NAS storage. (Live migration allows a
running VM to be moved from one physical host to another, without impacting the
VM.) There are also specific feature differences between SAN support (which now
includes iSCSI as well as Fibre Channel) and NAS support, and differing
opinions on performance between the options, so check with your virtualization
vendor and your storage vendor to understand the nuances. In addition, features like thin provisioning
and SAN snapshots and cloning can be great complements to virtual servers, but
they also hold many potential pitfalls, if configured incorrectly or if certain
patches are not installed. Again, check with your vendor for the latest
information.
Something
else to think about with storage - moving 1,000 physical desktops to virtual
desktops solves a lot of management problems. But it also means taking the storage
for those 1,000 desktops and storing them all centrally on the SAN, for
example, resulting in 1,000 copies of Window XP sitting on the SAN. Addressing this issue up front can prevent
lots of trouble. Citrix has a great solution here with their Provisioning
Server golden image technology. Disk deduplication solutions can also help
here.
Conclusions
Server and
desktop virtualization offer huge benefits, both financially and operationally.
But like any complex technology we implement in IT, they have their share of
pitfalls, potholes, and things to watch out for. This article offers a few that we hear often
from users who have fallen in them, to hopefully help prevent new users from
taking the same falls. These are also
the tip of the iceberg, so if you have the option of getting help from your
vendors or channel partners who have already implemented projects like yours,
don't hesitate to get help. Those who have done it before are the best source
for finding out all the other things to watch out for...
|
Barb Goldworm is president and chief analyst of Focus
Consulting, a market research, analyst and consulting firm focused on
systems, software and storage. Barb has spent 30 years in various
senior management, marketing, sales, engineering, technical and
industry analyst roles with IBM, StorageTek, Novell, Enterprise
Management Associates and several successful startup ventures. Barb
began working with virtualization at IBM in the late 1970s (as a VM
technical specialist), and later was VP of Marketing for the company
that developed the first PC software distribution product in the
mid-1980s (acquired by Novell in 1992). In 2006, she was commissioned
by Wiley Publishing to write a book on blades - she agreed but changed
the focus of the book to include a major emphasis on virtualization -
the book was released in 2007, entitled "Blade Servers and
Virtualization: Transforming Enterprise Computing While Cutting Costs",
available on Amazon.com, Borders and Barnes and Noble.
Barb is a frequent keynote speaker, chair, and presenter at
industry events. In 2007, she chaired the Server Blade Summit on Blades
and Virtualization, and has been the keynote speaker for numerous
Virtualization Summits this year. She also created and chaired the
Interop Network Storage Track and has been one of the top 3 ranked
analyst/ knowledge expert speakers at SNW.
Barb has published extensively since the 1990s, and has been a regular contributor to Network World, Computerworld, Computerworld Storage Networking World Online, Tech Target, and now Virtual Strategy Magazine.
She has authored numerous research studies, landscape reports, and
business and technical white papers on systems, software, storage,
storage networking and enterprise management. Barb can be reached at
This e-mail address is being protected from spam bots, you need JavaScript enabled to view it
|
