President's Perspective

V for Virtualization, V for Value

The theme chosen for this edition of TrendSetter is virtualization. The term "virtualization" can mean different things to various people. This is why this quarter's edition of TrendSetter explores the various virtualization offerings in the server, software, and storage realms. In choosing to dedicate this issue of our newsletter to virtualization we are reflecting the renewed emphasis that the industry is placing on the many technologies that lie behind this concept.

Virtualization: The great optimizer

Organizations are moving at different paces in accepting or even acknowledging the emerging and maturing virtualization offerings. Some struggle with how to transition from traditional "sys admin" practices to the place they might intend to migrate, which would allow them to take command of new capability and derive benefits. Datatrend can assist clients in better understanding these maturing virtualization offerings and how to incorporate best practices and processes in concert with specific, in-place enterprise infrastructure, current initiatives and the present landscape of systems and applications.

Why do we think it is so important? Our answer lies in the fact that we only exist by adding value for our clients; and we only flourish by keeping ahead of the curve in how to maximize that value for you.

Following a decade of rapid and exciting expansion, the key to maximizing value in IT has switched away from individual projects. For most of our clients, the central challenge is in consolidating and integrating what they already have.

While the various server, software and storage platform virtualization technologies vary considerably, the common denominators and goals to achieve through the leveraging of such offerings have to do with increasing resource utilization, improving manageability, and decreasing outages in the course of performing routine upgrades and maintenance. The virtualization products go hand in hand with policy setting and monitoring of resources as tied to specific solutions/applications. High volume, high performance requirements can be routed, on the fly, to higher performing infrastructure elements, while reallocating lower performance requirements elsewhere.

Datatrend can perform assessments related to skill gap analysis, current infrastructure, system/application layers and other factors that would tend to tune the recommendations on how to ramp up to virtualizing servers, storage and other paradigms.

Value: The real objective

As you read this edition of TrendSetter, please keep in mind Datatrend's offer to you to assist in collaborating on the review and development of a strategy tailored to your organization. This can pave the way to take advantage of virtualization in the easiest and most logical manner, with your skills and experience in mind. What Datatrend really wants to deliver to you is value. We chose to devote this issue to virtualization because we believe it underpins many ways we can create value across the spectrum of your IT investments.

Virtualization - the logical thing to do

What is virtualization and why is it so important?

When IT vendors start to use six-syllable words, hang on to your wallet! Virtualization is a big word that's in fashion right now - but that doesn't mean it's new. This is an old idea that's been delivering IT value in many ways for over 40 years.

OK - but what is it?

The essential "trick" that underpins all virtualization is the separation of logical function from physical implementation. Once that is achieved all sorts of almost magical results become possible. So, the most general model of virtualization is shown in the diagram.

A few examples will quickly establish the power and pedigree of this approach.

Virtual memory: 1958

This first virtualization breakthrough is over 45 years old. Despite that, Microsoft was still struggling to overcome the 640K memory barrier in the 1990s! (For those under 30 years old - yes, we did mean 640K.) Virtualization enables memory sharing and increased utilization.

Database: 1960s

The virtualization layer introduced by Database Management Systems was often referred to as data independence

It simplified programming and operations and supported sharing of data between applications.

Virtual Machines: 1970s

With a virtualization layer called a hypervisor we were able to run several guest operating systems on one machine. This logical partitioning allows a test "machine" to run alongside a production machine or three old machines to be consolidated onto one and thousands of other powerful uses.

Storage virtualization: 1990s

Allowing every server to access any storage device obviously increases flexibility and reduces cost. This sort of storage pooling can be achieved with storage networking, but only up to a point. In a simple storage network, each server still has to be able to "understand" each of the storage devices it wants to access. But virtualization goes much further. The virtualization layer fools the server into thinking that it is accessing a familiar device when, in fact, it might be fulfilling the request from a device that wasn't even invented when this server was purchased. This is the key feature that differentiates true virtualization from other techniques. The virtualization layer takes logical requests and fulfills them exactly as the requester is expecting. However, the actual physical way that the request is handled is translated by the virtualization layer to match the physical devices or resources that are being used. In the case of storage virtualization, this means that:

• A server can successfully use data from a storage device it doesn't support directly

• You can optimize your storage utilization and performance by moving files or databases to the devices that best support them without having to even think about which servers need to access the data

• Utility functions like data backup, normal recovery and disaster recovery can be planned and optimized across the entirety of your storage assets instead of sub-optimizing by server

• Storage administration is simplified even though the tools provide more power  to optimize the storage subsystem

Benefits

The benefits listed above are for storage; however, virtualization, by its very nature, tends to provide the same type of benefits wherever it is applied. Here is a generalized list of virtualization benefits:

• Simplification: When sprawl has created the complexity inherent in variety, virtualizing the environment can enable consolidation and optimization.

• Increased utilization: Rapid growth accompanied by technological change often results in underutilization. Virtualization allows any-to-any sharing that goes beyond mere pooling and can allow you to utilize your existing capacity as an alternative to buying more.

• Integration: Removing the physical complexities and reducing each element of the system to its logical function enables many parts of a system that were previously isolated to work together in creating value.

• Ease of administration: Instead of using five different consoles to manage five different types of storage devices, an administrator can view a virtualized storage subsystem as a single entity. Similar benefits accrue in the administration of other virtualized environments.

Summary

One strong thread through the history of IT is virtualization. It started nearly 50 years ago with hardware features like memory. Looking back, we can see that, once a virtualization technique was perfected, it has taken over and become common - in the case of virtual memory, even universal. This evolution towards virtualization provides strong market evidence to supplement the logical appeal of the concept.

The evolution has accelerated exponentially in recent years. To some extent this is due to maturation of the technologies; to some extent it is due to price-performance making it affordable. However, it is also driven by need;

• The explosion of intranets and the Internet has resulted in server, storage and network sprawl that is more the rule than the exception. Today's environment is crying out for the consolidation, simplification and optimization that virtualization can enable.

• The very same explosion has created an opportunity (if not an imperative) to change the nature of the relationships between enterprises. Integration of processes between enterprises demands that physical details be stripped away to expose the true logical nature of each process and each transaction. Thus, virtualization will play a key role in enabling the development of the extended enterprise.

In short, it's logical to be logical; today it is practical to be logical; pretty soon, it will be virtually indefensible to be physical and virtually mandatory to be virtual.

Time for detail

We can only go so far in talking about virtualization in general terms. So, the following three articles take us to the next level of detail by providing examples of specific products in three areas (click the subject to jump to the article):

• pSeries - illustrating how IBM's new Virtualization Engines are scaling new heights of power and flexibility

• VMWare on BladeCenter - highly sophisticated server virtualization in a highly dense Intel-Architecture environment

• Storage virtualization - showing how hardware and software need to come together to fulfill the promised benefits.

... on how virtualization can revolutionize your IT environment.

Virtual server heaven on the pSeries

Our lead article shows that virtualization can have tremendous benefits to most organizations. This article sets out to show that IBM's p5 represents a virtual dream solution. So, if you are a pSeries user or have future interests in p5 we provide more detail to justify the virtual dream solution claim.

Why does today's server environment need virtualization?

When people talk about the benefits of "IT consolidation" or "IT optimization" they often refer to benefits like "greater utilization", "sharing resources", "any-to-any access", and "cross-application integration". Clearly these are "good" things that anyone would want. The statement that is implied by these arguments is that, today, many organizations have collections of systems that don't share resources well and don't integrate. This is true because these systems evolved at a time when the technology to "do it right" was not affordable, even if it was available.

Throwing out the old systems and recreating all the business applications in a "better" environment is clearly not practical. So, IT departments have to migrate the existing systems to an environment where sharing and integration can be retroactively imposed on legacy systems - not an easy task! As a broad simplification, there are two levels of solution:

Level 1: Switching. By putting switches between parts of systems that were previously isolated you can achieve some level of sharing and integration. For example, putting shared network switches between previously isolated servers and a common network can allow several servers (therefore, several applications) to share a single common network (sharing) and communicate with each other through the common network (integration). Equally, you can put shared switches between several servers and previously isolated storage devices to create Storage Area Networks. SANs allow some pooling of storage devices (sharing) and increase the ability of previously isolated applications to use the same data (integration). But switching alone is often disappointing. For example, SANs do not break down all the barriers between applications and the storage devices they use, so that there are limits to the resource sharing that is achieved; plus, the support and management costs of heterogeneous storage devices is not greatly simplified by merely moving them into a SAN.

Level 2: Virtualization. Switching just allows any-to-any connection between the elements. Virtualization goes one step further – one giant step. It consists in reducing the "dialog" between each element of the system to a logical rather than physical interaction. To do this, a virtualization layer is placed between the application code and the data or between the application code and the network. Once you do this some magic happens! Now any application data request can be fulfilled by any device and you no longer have limits to your resource sharing. You still have to engineer the business rules for integration, but all the previous barriers to integration caused by the physical and historical implementations disappear.

What's so great about pSeries virtualization?

Virtualization is a great concept, but like all concepts it only becomes a great reality through excellent implementation. To fully implement a practical and efficient virtualization engine requires a very powerful vision (or architecture) combined with prodigious amounts of very efficient coding. If you want proof of how difficult it is, reflect on the 640K memory barrier on PCs; it took Windows a decade to achieve virtual memory – the first baby step towards virtualization – a step that IBM took with its systems in the early 1970s. Or reflect on the problems that several manufacturers are still having in trying to move to 64 bit processors within Intel-based servers - something that was implemented overnight on pSeries some five years ago - because of virtualization layers between AIX and the hardware platform.

To illustrate the excellence of virtualization on the p5, here are just a few of the features and benefits that it supports.

Logical Partitioning

Feature: One physical system can be made to work like several, separate logical systems.

Examples of common benefits:

• Allows consolidation of several separate old severs onto one powerful new server reducing space, electric power, cooling, maintenance and other costs.

• Allows applications that require different OS environments to coexist on the same system, since each logical partition of the machine can run different levels of the operating system (and other system software components). This is of great benefit during testing and transition between different versions of OS or application software.

• Allows an isolated testing "machine" to coexist on the same physical server as production, removing the expense of a separate test system.

Dynamic Logical Partitioning

Feature: The ability to dynamically reassign resources between logical systems.

Examples of common benefits:

• Allows for dynamically matching the resource allocation to the changing needs of the business processes. For example, online customer transactions get most of the resources during the day and batch processing gets most of the resources overnight.

• Allows for non-disruptive reconfiguration of the system to meet special needs. Like bringing up a special system to diagnose an application error or repair a database without interrupting normal user service.

Micro-Partitioning

Feature: The ability to divide a single processor into up to ten logically separate resources.

Examples of common benefits:

• Allows even the smallest, uniprocessor p5 servers to implement the logical partitioning described above.

• Supports fine granularity in managing and balancing resources across multiple logical machines.

• Allows the creation of specialized logical machines that take almost no resources away from production. For example, a Linux test environment that is being used by just one programmer could be given one-tenth of a processor.

Virtual Input/Output

Feature: The ability to "fool" an application into thinking it has direct control of real physical devices, when in fact it is sharing the resource with other applications. The real physical resource is actually owned and managed by an I/O server partition that can optimize the utilization of the shared resource.

Examples of common benefits:

• A single, high-speed I/O path (such as a fiber channel) can be shared by several applications.

• A single disk subsystem can be shared between several applications providing better performance and/or lower overall cost. This can even allow applications to use new and better devices that they couldn't support if they were attached directly.

• The central I/O server can be used to manage functions like backup and high availability (mirroring or RAID5) on a more efficient centralized basis or even in ways that the original application could not support if it had direct control of the devices.

Mixing different operating systems

Feature: This is, perhaps, the blockbuster of server virtualization capabilities. Not only can one system simultaneously support logical partitions with different levels of AIX, now you can have completely different OSs running on the same system. A common example would be to have AIX and Linux partitions running on the same p5. But within IBM's POWER5 architecture, you can go much further and have AIX, Linux, Windows and i5/OS all running in logical partitions on the same machine. All sharing resources and being managed centrally.

Examples of common benefits:

• The ultimate version of server consolidation. Not only reduce the number of physical Windows servers, UNIX servers and so on, but move them all onto one physical platform.

• Allows all of the sharing and integration benefits outlined above to be realized across OS boundaries - without any reprogramming or conversion.

pSeries virtualization summary

IBM has taken decades of experience in virtualization technology and integrated it into a set of virtualization engines for its servers. This sharing of technology across the family of servers brings many benefits. For example, the virtualization engines of the pSeries, described above, are almost identical to the engines used on the iSeries; this sharing of technology reduces cost for everyone, but has particular benefits for organizations who want to deploy both pSeries and iSeries.

The power of virtualization on the pSeries has been growing over the years. But with the p5 generation it has taken a giant leap in function and affordability. That's why we say p5 (and i5) represent virtual dream solutions.

... on the tremendous benefits of pSeries consolidation with virtualization.

Ultimate virtualization on BladeCenter

The vital area of Wintel consolidation

As the lead article stated, the need for virtualization has exploded because of the server, network and storage sprawl that happened in the last decade. The driving force was the Internet and the majority of the sprawl happened with Wintel architecture servers. So, that's where the greatest need for consolidation and optimization is found and that's why virtualization is so important in this area.

Step 1: Physical consolidation

BladeCenter represents a blockbuster contribution to Wintel server consolidation. When you see a whole server farm reduced to a single rack – or even to a single suitcase-sized chassis – you know you are witnessing an important breakthrough.

Beyond the space, cabling and power savings, BladeCenter provides many additional benefits in availability and administration. So, BladeCenter alone provides a significant solution for consolidating sprawling servers.

Step 2: Optimization through virtualization

Based on the general analysis in the lead article, it is no surprise that further benefits in utilization, integration and administration result from adding the power of virtualization to the BladeCenter. In the rest of this article, we illustrate these benefits by showing how VMware can be used to virtualize a BladeCenter and, in addition to all the usual virtualization benefits, add a few particularly advanced features.

Taking both steps with BladeCenter and VMware

VMware ESX Server is an industry-leading solution for bringing the benefits of virtualization to Intel Architecture servers. As such, it can be deployed on individual blade servers within a BladeCenter.

The diagram to the right shows how VMware ESX Server can partition a single blade server into multiple virtual machines. In this case four virtual servers are sharing the single physical server. These virtual servers could be running end-user applications or infrastructure functions (such as DNS) or a mixture of both.

This gives all of the normal benefits of virtualization, including:

• Higher server utilization allowing server consolidation. A mix of appropriate virtual machines can be added to a physical server to fully utilize its capacity.

• Sharing of storage subsystems and network infrastructure.

• Because the partitioning is dynamic, the ability to change the allocation of resources in real time to reflect the immediate needs of each application or infrastructure function.

• Additional servers can be deployed as virtual servers in minutes - rather than the hours or days needed for physical server deployment.

The dynamic nature of the partitioning is illustrated in the diagram by showing the resource allocation at 2 AM above the blade and the allocation at 10 AM below. Allocation of processor cycles and many other resources can be managed in real time using the administrator functions of VMware.

VMware ESX server summary

Deploying VMware on individual blades within a BladeCenter provides significant increases in utilization and efficiency. Combined with the flexibility provided in administering and managing the resources, this can lead to dramatic reductions in complexity and cost.

But wait ... there's more

So far we have talked about improving the operation of a single blade server within the BladeCenter. With the addition of VMware VirtualCenter we can extend the benefits across the whole BladeCenter and in new dimensions.

VirtualCenter is central infrastructure management software that plays the role of mission control across all your servers that are using VMware solutions. It will, of course, manage other server platforms but our illustrations here are focused on BladeCenter.

VirtualCenter's management is comprehensive and the list of features is far too long for us to do it any justice in this short article. So, perhaps the best way to get some insight into its power is to concentrate on one of the most advanced features and the associated benefits.

The diagram (which you can click to enlarge) shows Server A, which had been running on Blade X. It is now being dynamically moved to Blade Y. The facility that performs this virtual move is called VMotion.

We are not talking about closing down the virtual server on X and then opening it on Y – any virtualization scheme can do that. The magic of VirtualCenter's VMotion is the fact that the move is in real time without closing down the server and without interrupting service. The integrity of individual transactions and their data are maintained even if they are "in flight" as the move takes place.

It is easy to see many benefits from this advanced VMotion facility:

• Dynamic workload balancing: Dynamic logical partitioning allows us to allocate resources within a server. VMotion adds a higher dimension by allowing us to move whole virtual servers from blade to blade around the BladeCenter.

• Zero-downtime maintenance: Whether planned or emergency maintenance is needed, VMotion will allow all of the virtual servers on a blade to be moved. This releases the underlying physical blade for maintenance - without disrupting service.

• Transparent addition of new hardware: When an additional server blade is installed, it can be configured and tested separately. Then live applications or infrastructure functions can be moved over to it without disruption or interruption of the service.

A heaven made in marriage?

Separately, BladeCenter and VMware solutions are powerful and impressive. But if you marry them together you could create a virtual heaven for your end users, your CFO and your IT staff. The whole is certainly greater than the sum of the pair. Together BladeCenter and VMware could create the smallest, most efficient, most flexible server environment available today. More importantly, it could have the lowest total cost of ownership.

... on how BladeCenter can create a virtual heaven for your organization.

Optimize storage

Increase efficiency and reduce costs with storage virtualization

If you read all of the preceding articles, you will discern a pattern; optimization is achieved in stages – with virtualization being the most advanced stage – the stage that yields the ultimate benefits. Storage is no exception to this pattern; here are the stages involved in storage optimization:

Step 1: Storage networking

Not many years ago, most storage devices were directly attached to a server that "owned" them, thus making sharing difficult to impossible. Two forms of storage networking have broken down the barriers and allowed storage devices to be shared by multiple severs across the network:

• Network Addressable Storage (NAS) devices are best known for file serving –  allowing one common pool of files to be accessible to all servers and, therefore, to all users.

• Storage Area Networks (SANs) are based on switching devices that can direct individual blocks of data (as opposed to a whole file) from a storage device to any attached server. This architecture can support both file and database sharing schemes.

Step 2: Optimization through virtualization

Adding full virtualization to your storage networks yields many additional benefits. This is particularly true if you have a heterogeneous storage environment – that is, if you are trying to pool and share a range of storage devices from different generations or even different manufacturers. Full virtualization completely separates the logical data requests from the physical fulfillment of those requests allowing you to do such things as:

• Moving particular files of databases from manufacturer A's storage device to a device from manufacturer B – even though the application is on a server that doesn't directly support the product from manufacturer B.

• Splitting files or databases across multiple, different devices in order to better utilize the available capacity.

• Moving files in real time as applications continue to access them.

• Sharing backup devices and high availability schemes across a heterogeneous storage environment. Thus, reducing the cost and complexity of backup and high availability.

• Simplifying and speeding data recovery and disaster recovery. For example, giving the ability to recover the most vital data onto whatever combination of storage devices are still available after an incident. Thus, reducing the recover time and cost of your business continuity plan.

• Providing a single, simpler administration interface across a variety of disparate storage devices.

What you get from taking both steps

Through these benefits, full storage virtualization increases the life of your storage hardware by giving you the flexibility to achieve higher utilization. In other words, you can better utilize what you have rather than having to buy more. This becomes very important when you consider that many advanced users are finding that storage costs are rising rapidly to the point where storage is exceeding 50% of the total IT budget.

The administrative efficiencies are particularly significant in the optimization equation. The same advanced users are reporting that, for every dollar they spend on storage hardware, the deployment and administration costs are around $8! Thus, the flexibility and efficiency provided by storage virtualization may be the only way to reverse this trend and start to see the spectacular increases in storage hardware price-performance reflected in the bottom line of your IT budget.

Putting it all together

How you take the next steps in storage optimization depends on where you are now. Part of the challenge is to make maximum use of the storage legacy you have today. Since currently installed storage assets vary so much, it is not useful to talk about generalized migration paths. However, there are some considerations that do apply to most situations:

• From a hardware point of view you need to network your storage devices in order to build the benefits of virtualization.

• Software will usually play a significant role – especially in simplifying and unifying administration, backup and recovery. Look for Tivoli and TotalStorage software products to provide a vital layer of automation and simplification at the operational level.

With those thoughts in mind, here are some examples of the products that will be part of your solution, depending on your needs.

Storage costs - especially administrative costs – are the fastest growing part of the IT budget. You can reverse that trend by optimizing your storage subsystems. The most powerful tool for optimization is storage virtualization. And once they can optimize your utilization - you may not need to buy that extra storage after all!

 ... on how optimized storage can provide immediate benefits.

TechTip: Enabling External Connectivity on IBM BladeCenter

By Britton McGinn, Technical Engineer, Datatrend Technologies

On a new BladeCenter, remember to enable the external ports on each I/O module for external connectivity. By manufacture default these ports are disabled for security purposes.

1. Open Web browser to your Management Modules IP address (default 192.168.70.125).

2. Login (default user ID = USERID; default password = PASSW0RD)

3. In the task bar, scroll to I/O Module Tasks - Admin/Power/Restart

• In this window go to the I/O Module Setup section

• Select the desired I/O module and enable external ports

To be able to telnet for CLI, ping, and open the GUI interface to your I/O modules from the Management Module make sure of the following are configured on the same subnet:

• I/O Modules - default=192.168.70.127 to 192.168.70.130

• Management Module external network (eth0) - default=192.168.70.125

• Management Module internal network (eth1) - default=192.168.70.126

Contact us | Visit Datatrend website

All trademarks, registered trademarks and service marks are the property of their respective owners.
IBM, the IBM logo and other referenced IBM products and services are trademarks or registered trademarks of the International Business Machines Corporation in the United States, other countries, or both. All rights reserved.

UNIX is a registered trademark of The Open Group.
Windows is property of Microsoft Corporation.

Brought to you by Datatrend Technologies Inc.

6815 Meadowridge Court, Alpharetta, GA 30005