The world of Microsoft SQL Server licensing1 and pricing is changing as per Microsoft’s recent announcement2. Reading through the SQL Server 2012 licensing datasheet3, I found multiple ways to license SQL Server Standard Edition (SE) and the treatment of virtual machines (VMs) of particular interest. The Standard Edition has basic database, reporting, and analytic capabilities.

There appear to be two ways to license VMs for the Standard Edition. (1) License individual VMs with core licenses—one core license per virtual core with a minimum of four core licenses per VM; and (2) license individual VMs with server licenses and license each user or device with a client access license (CAL). Considering these models, I decided to delineate the cost advantages of each model over a range of VMs and users. The results are pretty interesting.

I used the demonstration pricing in the datasheet³ as follows:

For my test case, I chose a range of 250-25,000 users and 4-256 VMs. I allocated 4 VMs per server with 4 cores per VM, as there is a minimum of 4 core licenses per VM. Core-based licenses don’t depend on the number of users and the calculation is straight forward. Example: 4 VMs x 4 cores per VM x $1,793 per core = $28,688.

Core-based license cost:

The core-based cost per user directly falls out as:

The cost profile spikes as the number of VMs goes up.

Server + CAL License:
The Server + CAL license model is a little more complicated. I assumed 4 VMs per server and the SE Server License demonstration pricing of $898 per VM. Each CAL was $209. The cost is (#VMs x $898/VM) + (# CALs * $209/CAL) with 1 CAL per user. Example: for 16 VMs (4 Servers) and 1000 users the cost calculation is: (16 x $898) + (1000 x $209) = $223,368

The cost profile is fairly flat for 2,500 users and above, but grows significantly when the number of VMs is high and the number of users is below 2,500.

Server + CAL cost per user:

Showdown—Which is better from a cost perspective, Core or Server + CALs licensing?

It is less expense to go with core-based licenses for a total of 4 VMs, regardless of the number of users. Server plus CALs licensing creeps in as the better choice for 16 VMs and 250 users or less.

64 VMs is a mixed bag, for 1000 users or fewer, Server plus CALs is cheaper. But for 2500 users or greater, core-based licenses are less expensive.

Jumping out to 256 VMs, Server plus CALs has the advantage for 7500 users and under.

All in all, you have to take a close look at the number of users and VMs to achieve the lowest licensing cost for your estate.

References:
1. SQL Server 2012 Licensing Overview
http://www.microsoft.com/sqlserver/en/us/future-editions/sql2012-licensing.aspx
2. SQL Server 2012 Announcement
http://www.microsoft.com/presspass/press/2011/oct11/10-12PASS1PR.mspx
3. SQL Server 2012 Licensing Datasheet
http://download.microsoft.com/download/A/5/D/A5D112E1-78FF-491F-9364-F1BC6FAE7D57/SQL_Server_2012_Licensing_Datasheet_Nov2011.pdf
4. Related Blog
http://blogs.flexerasoftware.com/elo/2011/11/revolutionary-changes-to-microsoft-sql-server-2012-license-models.html

The Business Software Alliance (BSA) recently reached a settlement with a Belfast based architecture firm in which the firm paid £15,000 (about $24,000 USD) in damages and another £18,000 (~$28,700 USD) in true-up fees to purchase the necessary software licenses for Microsoft, Adobe and Autodesk products. The BSA represents independent software vendors such those mentioned above, as well as others including Progress Software, Symantec and The Mathworks.

In the article posted on the BSA website, Julian Swan, Director, Compliance Marketing EMEA, BSA, commented, “Unlicensed software often occurs when a company’s management regards software licensing as only an IT problem, rather than treating their software as a business asset. It is important for companies to implement software asset management (SAM) to ensure that not only are they legally compliant, but are deploying their software in the most cost efficient and productive way.”

Mr. Swan has hit the nail dead center in this statement. Maintaining software license compliance (reduced license liability risk) is only half of the equation when it comes to effective license management. The other half is cost reduction related to licenses, maintenance and improved operational efficiency. In another recent blog we discussed the Gartner survey results that showed that IT Asset Management (ITAM) programs lead to significant cost savings. In fact, the report says—

“Gartner research indicates that organizations that have effective ITAM disciplines will reap the benefit of cost savings, even in the current economic climate. More than 25% of the 66 respondents that measure ITAM savings indicate that their ITAM programs saved 10% to 20% of their IT spending in both the first year and over years two to five of their programs.”

10% to 20% savings on IT spending (not just on software spending) every year for five years sounds pretty darn good. But, in many cases, organizations can do even better, and more organizations can achieve at least this level of savings. The reason is that many organizations are not yet at the level of optimized license management maturity where maximum cost savings can be achieved. One third of the organizations in the Gartner survey rated themselves as Level 3 in Gartner’s 5-level maturity model, which they define as— ITAM roles are defined and processes are applied consistently across the organization. ITAM data is reliable and is starting to be used to guide tactical IT decisions. This means that there are more savings to be had through optimizations such as automatically applying software product use rights (license entitlements) to reduce license consumption, and reharvesting existing licences. (Download this whitepaper to learn more).

Many large enterprises have a multitude of IT asset management tools already in place— configuration management, service desk, change management, CMDBs, and more. So, naturally, they must have Software Asset Management (SAM) covered. After all, if organizations have tools from one or more of the big tool vendors that provide suites of products that “do everything”—they must have everything they need, right? Well, not really. The fact of the matter is that while these big framework tool suites may claim to do software asset and license management, the reality is that they barely scratch the surface.

The big tool suites are good at what they have been designed to do, including configuration and change management. They haven’t been designed to go deeply into optimized software asset management. For example, inventory and configuration management tools can provide raw inventory data covering both hardware and software installations. But typically, this inventory data is not in a form that provides actionable information—especially in the case of software. From a software license management perspective, you often need to know specific details, including not only the hardware characteristics—machine make and model, processor/core type and count, etc., but also the title, version and edition of software installations. This requires a tool that can analyze the raw inventory data and perform what is called ‘application recognition’ to generate the list of applications installed on each system, and get rid of a lot of the garbage at the same time.

And many inventory tools don’t cover all the bases even for basic software inventory. SQL Server is a case in point—most inventory tools don’t provide version and edition information for this application. Oracle database discovery and inventory is another area where many inventory tools fall short. Oracle license management requires an understanding of database instances and options in use by the organization. In addition, while the Windows platform may be well covered from an inventory standpoint, UNIX, Linux and MAC OS platforms are often another story in many organizations.

There are also special requirements for discovery and inventory in virtual environments. Collecting inventory on virtual machines (VM) and hardware partitions simply requires having an inventory agent on the VM or partition. But to accurately determine your license position, you also need to know on what physical server that VM is running (and of course, this can change as VMs are moved around using tools such as VMware’s vMotion) and what hardware resources have been allocated to the VM. This is necessary particularly when the license model is based on hardware characteristics—e.g. processor/core based licenses such as IBM Processor Value Unit (PVU) and Oracle processor based licenses. One way to get this information is to query the hypervisor, in the case of VMware ESXi/vSphere Server, for example. Many common inventory tools can’t do this.

Collecting inventory is just the very first step in the multi-step software license management process; this is the ‘know what you have’ step. You also have to ‘know what you own’—how many copies of what software did you purchase? Purchase orders, even for software, can have stock keeping unit (SKU) part numbers for each line item. Software SKUs tell you a lot about what you bought and how you bought it. For instance, they can tell you title, version and edition of the software, as well as the type of purchase agreement (e.g. Microsoft Enterprise Agreement, Select Plus, Adobe CLP, etc.) and whether you also purchased the maintenance plan along with the license. Next generation software asset management tools leverage these SKUs to help automate the process of reconciling ‘what you have installed’ versus ‘what you own’. The big framework tools don’t have this sort of SKU library.

Once you have a handle on your basic purchased versus installed license position, you need to understand and apply your license entitlements to get to an optimized license management state. License entitlements—the number of licenses, the license metric (device, user, processor, etc.), and the product use rights, tell you where, how and by whom, the software can be installed and used. Product use rights range from the relatively simple (e.g. right of second use, downgrade right, etc.) to the very complex (disaster recovery/failover rights and virtual environment use rights). The maintenance plan usually provides important product use rights such as upgrade rights. Microsoft’s Software Assurance program also provides virtual desktop infrastructure (VDI) rights (see this recent blog).

An optimized software asset management program must keep track of these use rights so that you can take advantage of all the benefits provided by the vendor and minimize your license consumption. You may be over-spending on licenses and not even be aware of it in cases where product use rights allow the same license to cover multiple installations of the software. Again, none of the big framework tool suites provide a product use rights library to automate this process of applying your entitlements to calculate an optimized license position.

So, do next generation software asset and license management solutions replace your existing IT management tools? No, that’s generally not the case. Rather, they should ideally leverage your existing IT infrastructure. For example, they should make use of your inventory tools to avoid having to install another agent, where possible. Optimized license management tools also connect to your procurement systems, HR systems, Active Directory, and other tools to import purchase order data, contracts and organizational data, to name a few. All of this data is required to allow the tools to optimize your license position, maintain license compliance and reduce ongoing software costs.

With the enterprise adoption rate climbing for mobile devices like Apple iPads®, iPhones, and Android-based tablets, the question looms as how to best manage them. The question I get almost every day is how to inventory them. The answer is intertwined with process and policy. First, it should be decided if personal—brought from home—devices will be permitted to join the company network or will the company provide these devices to their employees. The latter is not too complicated. An inventory/configuration agent would be installed by IT, before delivering the device to the end user or the user would explicitly accept a request from a company designated management server (securely enroll) on its network—simple.

Once an agent is installed on a device, it can provide inventory including the hardware, operating system properties, and software applications installed. But, implementing an inventory agent alone has limited value. Enterprises need additional management capabilities to bring these devices into compliance with their internal policies through configuration (Exchange Active Sync, VPN, LDAP…), policies linked to security (password rules, auto-lock rules…), and restrictions (use of camera, ports, iTunes, YouTube, Safari…).

Let’s not forget personal devices. These may or may not get out of the gate and onto many company networks. Employees concerned about invasion of privacy will be reluctant to have their company inventorying their smart phones or tablets and their employers will be reluctant to accept these devices due to confidentiality and security concerns. This sounds like the airport pat-down dilemma—security versus privacy. If these issues are somehow overcome, then the process is the same to get these devices under management. An agent needs to be installed or a request accepted from a designated management server.

Where do you get the software to keep track of these devices? Apple, for example, does not provide remote management capabilities themselves, only the building blocks called “Mobile Device Management” or MDM for other companies to build solutions. There are many companies delivering solutions today as this list from Apple details. But what about the big dog in hardware/software inventory and configuration management—Microsoft?

With recent announcements from Microsoft at the 2011 Microsoft Management Summit, familiar software is around the corner. They plan to deliver an inventory agent for the iPad, iPhone, Android, Symbian, and other mobile devices with System Center Configuration Manager (SCCM) 2012 due out later this year. This is good news for Flexera Software and its customers since SCCM is a standard inventory feed into our FlexNet Manager Platform for asset, contract, and license management.
Is your iPhone/iPad allowed on your work network? Are there any restrictions?

On July 12th 2011, VMware announced the release of VMware vSphere 5.0. This version comes with new and improved features but it also changes the licensing structure of the product. The legacy vSphere 4.x license model was based on the number of physical processors (CPUs) or sockets, with limits on the number of physical cores per CPU and physical RAM (memory) per server. The new vSphere 5.0 license uses the same metric (CPU). All hardware limits on cores per CPU and physical RAM per server have been eliminated. A new vRAM (virtual RAM) license entitlement rule has been added: each vSphere 5.0 CPU license entitles the licensee to a specific amount of vRAM configured to virtual machines.

When used for unlimited hosts, vSphere 5.0 comes in three editions: Standard (24GB per license), Enterprise (32GB per license) and Enterprise Plus (48GB per license), with each license covering a CPU or socket. The amount of vRAM used by a single edition of vSphere, can be pooled across all virtual machines managed by a vCenter instance or multiple linked vCenter instances. Once the vRAM capacity limit is reached, additional licenses must be purchased, regardless of the CPU consumption, to match the memory high water mark. But all virtual machines will still power on even if the high water mark is exceeded. VMware has defined the vRAM rule as a soft limit and will enforce it most likely through software license audits.

One area where memory per socket is likely to exceed the ratios mentioned above is desktop virtualization. Any customer under a Service and Support agreement using a previous dedicated version of vSphere for desktop virtualization can migrate to vSphere 5.0, while retaining access to unlimited vRAM entitlement. VMware has released their vSphere Desktop Edition that specializes in managing Virtual Desktop Infrastructure (VDI) environments. The product license does not carry any memory limits. There is no upgrade path from existing vSphere licenses to this product. Customers are encouraged to purchase this product to manage their existing desktops and repurpose their existing vSphere license to host server based workloads.

Before migrating to vSphere 5.0, customers must evaluate their vRAM position across the entire organization and define a new strategy. Virtual machines that are overloaded with memory should be resized. vRAM capacity could be claimed from virtual machines that usually do not consume high levels of RAM, including, for example, disaster recovery environments. A new methodology should be built to provision virtual machines and minimize vRAM consumption moving forward. This is the right time to review virtual machine templates and make sure they are sized properly. This overall vRAM analysis is not an easy exercise as many parameters must be considered including:

vRAM is pooled, so all vCenter instances must be considered
vRAM applies to allocated memory on all powered-on virtual machines (VM)
Limiting the vRAM may decrease performance

The overall feedback from VMware customers to the new licensing model has been less than enthusiastic. It is a challenge, as they have never been confronted with such virtual memory limitations in the past. Many of them won’t be impacted by the new memory entitlement, however it is anticipated that large enterprises with big host servers will need to upgrade their licenses. For some of them, it will come at a very high price tag. Over-provisioning memory is a current practice that will end with this vSphere release. It will now require additional software asset management capability and monitoring to stay compliant with VMware licensing.

The most interesting part with this new rule is the adoption by VMware of virtual memory usage based pricing. It severed the links between the license and some of the hardware characteristics of the host server. Limits on physical resources—the number of physical cores per CPU and the physical RAM per server have been eliminated. As a result, the specifications of the machines hosting the licensed software do not really matter, with the exception of the number of CPUs or sockets. VMware, with vSphere 5.0, has taken a step forward to cloud licensing with a “pay per consumption” license model.

Flexera Software announced today the immediate availability of a new Enterprise License Optimization product– FlexNet Manager for IBM. This product is part of our FlexNet Manager Suite for Enterprises solution for optimized software asset management and license compliance. FlexNet Manager for IBM is the latest addition to the list of vendor-specific license management products that also includes: FlexNet Manager for Microsoft, FlexNet Manager for Adobe, FlexNet Manager for Symantec and FlexNet Manager for SAP Business Suite. Our focus is on providing optimized license management solutions for the key software vendors. All of these vendors are in the top 20 list of software companies by revenue, with IBM coming in second on that list behind Microsoft.

FlexNet Manager for IBM provides automated license reconciliation for IBM DB2, WebSphere Application Server and Tivoli Workflow Scheduler. It supports IBM-related license types including, Processor Value Unit (PVU), Authorized User, Concurrent User, and Floating User and both full capacity and subcapacity licensing.

FlexNet Manager for IBM is built on the FlexNet Manager Platform, which provides core IT asset management and license reconciliation capabilities for more than 11,000 software vendors and 100,000 applications. The new release of the Platform (version 8.5), also available today, provides a number of new capabilities including discovery and inventory of hardware partitions (AIX LPAR, HP-UX nPar, vPar and Solaris Zones technologies).

What other software vendors are on your top 10 list for license management?

I was recently asked by a customer to provide some standard definitions of common software licensing terms and thought it might be helpful to others. So here it is…

Common License Types

Appliance: A license covering use of a specific piece of hardware, such as a hub, router, or PBX. Terms and conditions vary between vendors.

User: A license that provides access to the software to a specific number of users. All installations of the software will be counted but installations across multiple devices for the same user will be counted as one license consumption.

Concurrent User: A license which provides wider access to the software but limits the number of simultaneous users using the software. It may or may not include compliance enforcement capabilities. Typically, a concurrent license is “checked out” from the license server when the software is run, assuming a license is available. If no license is available, the requester experiences a denial of service.

Named User: A license that allows access to the software by a specific number of named users. In some cases, these licenses can be transferred from one user to another. When you create the license, you should allocate the license to specific users. Only installations associated with allocated users are counted. For example, if the license is allocated to users Sam and Jan, the maximum installation count is two. Any other installations of the licensed application are treated as unassigned installations. For example, if May has also installed the licensed application but has not been allocated to the license, her installation will not be shown against installations of this license.

Enterprise: A license to install software an unlimited number of times within the enterprise. An Enterprise Agreement, such as the Microsoft EA, is defined separately to this in FlexNet Manager Suite (FNMS ). An Enterprise Agreement is structured as ‘all you can eat’ but the organization must be licensed for a specific quantity of licenses so this is not strictly an ‘Enterprise License’ model in its pure form.

Evaluation: A license that allows one or more users to install and use software for trial purposes. Evaluation licenses may be time limited, may offer limited functionality, or may restrict or mark output (for example, some PDF writing software includes the name of the software on every PDF document produced from a trial version). After evaluation, a user may purchase a full license, uninstall the software, or (for time-limited trials) the software will simply no longer work.

Node Locked: A license that allows access to the software on a specific number of named computers. These licenses are usually for server applications such as database or VMware products. In some cases, these licenses can be transferred from one computer to another, usually by requesting a new license key.

OEM: A license for software that is delivered with the hardware and is only for use on that piece of hardware. These licenses are tied to the lifecycle of the hardware and typically cannot be transferred to other hardware.

Processor (per Processor/CPU): A license based on the number of CPU/Processor sockets on which the software will run, and NOT the logical processors aka cores.

Client Server: A server license that is based on a device metric. In many cases this type of license may also have a Client Access License (or CAL) aspect. In a Server/CAL model a license must be purchased for the physical server (or virtual server – there are varying rules around virtualisation) and also additional ‘access’ licenses must be purchased for any users/devices that may access the server for that application.

Run-Time: A license that provides access rights to third party software embedded in an application. The use of the runtime license is limited to the application through which it has been acquired.

Site: A license to install software on an unlimited number of computers at one physical location.

Device (most common metric): A license for a defined number of software installations. The software may be uninstalled on one computer and installed on any other computer within the same enterprise, so long as the total number of installations does not exceed the number of purchased licenses.

Core/Processor points: A license based on points applied as a multiplier to the number of Cores/Processors in the physical server, or in some cases, the virtual machine. Some vendors count Processor sockets and others count logical processors, or cores, but the license model is similar. For example an application installed on a 4 processor server with 100 points per processor would require a purchase of 400 processor points to cover the license liability. These licenses are mainly used for Datacenter software licensing such as IBM.

Common Software Asset Management (SAM) Terms

Installations: The number of raw software installations without Product Use Rights or license metrics applied.

Entitlement: The number of purchased licenses available combined with any Contractual or Product Use Rights.

Consumed: The actual license liability (not to be confused with ‘Installs’), Consumption is the install count applied against the Entitlement

Compliant: If the number of licenses consumed is less than or equal to the number of licenses purchased.

Breach: If the number of licenses consumed is more than the number of licenses purchased.

Delta: The difference between the number of licenses consumed and the number purchased (for example 10 licenses purchased and 14 consumed would mean a delta of 4).

Was this useful?  Do you have other software licensing and Software Asset Management (SAM) terms and definitions you would like to share?

In first grade, we all learned that 1+1=2. In our family 1+1 = 4. That’s 1 wife plus 1 husband resulting in a family of four. Luckily for us, software doesn’t quite reproduce the same way, though I’m sure we might think so as we start to look at software install counts versus what we were expecting, in say a virtual environment. Take, for example, two installs of a piece of software that consume 4 licenses because the software is installed on systems with two processors, and the license for the software counts license consumption based on the number of processors. In that case 1+1=4. The process of counting installations, determining the correct number of licenses required and comparing to software purchases is often called software license reconciliation, and that’s a subject for a separate blog.

In another example, what about an IT asset management inventory that says there’s one copy of Adobe Acrobat on Steve’s Laptop under the Program Files Directory, and evidence of another copy of Adobe Acrobat in a different location, but there really is only one copy of the software with evidence being counted twice. You might know that Steve only has one copy of the software, but the software inventory says there are two copies. So, in this case 1 + 1 = 1 as long as you can reconcile the two pieces of inventory evidence correctly.

So just how is software counted, and how can the count sometimes be more than we expect (1+1=4), and sometimes less (1+1=1), and how can we resolve these issues to come up with an authoritative count we can use?

What are we counting and how are we counting it

The problem is not always the addition or reconciliation; problems can often occur in the counting well before any addition takes place. In addition, just what does it take to determine that an instance of Adobe Acrobat is actually installed, not to mention verifying the version and edition? It’s really a question of just what are we counting and how are we counting it.

So how do we assure we are counting correctly? For illustration, let’s take the example of the age old promotion retail stores do— placing a huge number of items, like marbles, in a jar and asking passersby to count them. Counting the marbles while in the jar is almost an impossibility, even if you use complex mathematical formulas (you can equate that to counting software on systems not connected to any network for example— you might as well just make a wild-eyed guess); there are just too many unknowns and variables like how large is the bowl, are the marbles all the same size, did the retailer put any fillers in the bowl. But even if you take all of those marbles out of the jar, then ask different people to count them you still might get different answers. Each person might use different methods to count and verify such a large number, and since the task is very monotonous, they will probably develop some scheme to make their job easier. So each starts using different methods to count, and no matter how accurate they are, chances are, they still could get a different number.

But then ask them not only how many total marbles there are but how many of each type there are. Chances are the variance will be huge when you start to compare the different counts for the different types. That’s because different people will have different ideas on what types of marbles there are, and as a result, one person might say they counted 285 cat’s eyes, while someone else might give you a number of 492 cat’s eyes. How, you may ask will two different people give such widely diverse answers? Well it comes down to what each person thinks a cat’s eye is.

Well, believe it or not, the same thing applies to counting software installations in your environment. Ask someone from IT how many copies of Microsoft Project are in your environment, and they will give you one number. Then ask someone from your help desk how many copies of Microsoft Project are out there, and you will probably get a completely different answer. Chance are these two employees are using two different sources to determine their answer, and each source will have its own idea of just what constitutes a copy of Microsoft Project being associated with a user or machine.

Then you need to consider that these two employees probably have a different understanding of just what you mean when you ask to count Microsoft Project. Is it the full install of Microsoft Project; the latest version of Project; and should we count editions of Microsoft Project separate from Microsoft Project Server components; or does the count include the free version of the Microsoft Project viewer as well. So your first problem is trying to come up with the definition of just what you mean by Project. Even if you were specific enough to say “How many copies of MS Project 2009 Professional do we have” you still might get two different answers because these two employees probably use two different resources to find the answer to the same question. One might use a raw inventory count, while the other might be using a reconciled count provided by some centralized help desk system or CMDB. Again, too many variables when trying to answer one simple question.

And don’t even get me started about asking Accounting how many copies of Microsoft Project they think you have in your company. Let’s just keep focused on counting how many copies of a particular software application you may have in your environment.
So the key here is to understand what you are asking for, and to make sure the people tasked with providing an answer also understand the same question. In this blog, we are trying to address the issues around counting the software installed in an IT environment. This includes the scrubbing of the raw inventory data using an application recognition library and also takes into consideration hardware characteristics.

Keep things simple

So what can we do about the task of counting what software is installed in the organization? Instead of using something like MS Project, let’s start with something easier like the operating system, and let’s just use one source for our question, in this case IT. When we ask IT how many copies of Windows XP operating system we have (our standard for all end-user systems), they might look at their inventory report and answer simply 4,982. So should we believe them? Is this the number we should provide to Microsoft when they come in for their annual true-up? How do I know if this number is even close and how can I verify it. Especially if you remember that just last week, you asked IT how many end-user computers they managed, and they gave you an answer of 5,250. So in this case 1 + 1 doesn’t always equal 2. 5,250 computers with one OS each should equal 5,250 copies of Windows XP, but it doesn’t. So where does the discrepancy lie. Well the first thing to realize is there’s not one answer to the question of why your numbers often times don’t match.

If IT tells us we have 5,250 computers they manage (we are going to ignore servers to keep this easy), then logic tells us we must have 5,250 copies of the standard OS we manage as well. If Windows XP is the standard for our world, then we must have 5,250 copies of Windows XP. So we go to our inventory and the latest inventory report tells us we have 6,384 computers and only 4,982 copies of Windows XP.

So you need to verify the count before talking to Microsoft. Once you start digging a little further, you may find several different items that can change this simple logic such as:

Oh, we forgot about the 800 systems we just ordered that we included in our count of computers, but that haven’t been inventoried yet.
Also, we have about 500 or so older computers that we never upgraded to Windows XP because they were fine with Windows 95, and we didn’t want to risk crashing the systems with an upgrade.
And then there’s a bunch of systems in the closet that have never been inventoried by our new inventory tool.
Oh, and then there are the 400 systems that development uses for testing that they loaded Linux on
And the list of variances continues on and on the deeper you dig.

Solution part 1: A central Hardware and Software Asset Management Repository

So how can we solve this problem that seems to get worse and worse the more you dig into it? To start, you should put in place consistent measures of software that utilize proven methods of reconciling all of the variances and discrepancies that arise in counting software. It’s also important to note that no matter what method of counting you use, you are always dependent on the source of the original count, which in most cases is an inventory tool or multiple tools of some type, so it’s always a good practice from time to time to run comparisons on the inventory counts from different systems. For example, how many computers does the helpdesk monitor, vs. how many computers is the inventory system reporting, and how many computers does accounting tell you that are currently on the books. Also, you should put into practice the process of comparing the electronic inventory with a regular physical inventory to help clean up systems that simply are no longer found in your environment.

In addition to putting in place standards for counting and measuring your systems and software, you should also implement a central Hardware and Software Asset Management repository. It should have built-in reconciliation algorithms, including application recognition based on known libraries of software and have the ability to reconcile varying sources of inventory data into a single count. The system should also take into consideration the hardware portion of the equation— which systems are active, which systems should no longer be counted, and which systems are double counting themselves for some reason and hence should be discarded.

Well, since I’m about to get off the airplane, we’ll continue with Solution Part 2 next time. Meanwhile, happy counting, and please don’t tell my 1st grade teacher that 1+1 doesn’t always equal 2. It would break her heart.

One to watch is the workflow management component that can be a go or no go, for efficient contract management. Workflow management is the process enforcer that facilitates transparency, making deadlines, and compliance (contract, regulatory, tax…). Workflow should be tightly coupled to the lifecycle of a contract for:

Request
Sourcing
Approvals
Fulfillment
Auditing
Renewals/Cancellations

A number of key workflows have an associated people matrix. For an approval workflow it may be the approvers and their roles associated with contract amount signoff, contract type, and originating location. If changes are performed manually to approvers’ roles or the approvers themselves, this may be a flashpoint waiting to happen. Updates become reactive to process delays or break down. An example would be a workflow process that is waiting on approvers that are never going to answer the bell. Perhaps because they are on vacation, traveling, on sabbatical, have left the company, or are just off the grid. It doesn’t mean the workflow process will fail. Workflow systems can handle aging, escalation, warnings, failover, cancellation, and updates for workflows in process. In this example the workflow should trigger an escalation phase and backup approver as required. But escalation delays make the process less efficient.

On the flip side, if the workflow system is connected to a dynamic source, a change in personnel could be accounted for and the workflow automatically corrected. The dynamic source doesn’t need to be complex, just accurate and kept up-to-date. In the foregoing example, a company directory that includes name, email address, title, department, and location is sufficient. Of course there should be an accompanying database source that associates a person’s title, department, and location to approval scope—department, business unit, subsidiary, company wide; type of contracts; and dollar amount. The approval scope database changes very infrequently, once setup. On the other hand, a source like the company directory can change almost every day in a large global organization with employees taking on new roles within a company, new hires, and people leaving. The workflow management system should “see” the changes in the company directory, and update its matrix of approver logic, based on their scope. It should then update any workflows in progress to account for the approver change. This potential flashpoint is then removed by a combination of robust workflow and a dynamic data source.

Another flashpoint is the inability to kickoff a workflow, or at least an alert, based on a contract term. Once a contract is signed off and in the firebox, the real work begins. With 1000s or 10000s of contracts, how do you know if the terms are being followed, tiered discounts are being taken, there is compliance to escrow deposits, time sensitive options are executed, or termination provisions proactively selected? An auditable closed loop process is really required. Key terms in a contract should be easily associated with a workflow, before the contract is put in the box. The associated workflow can then be the compliance watchdog to ensure the maximum benefit is achieved from each contract. This flashpoint is therefore addressed by picking a solution that doesn’t view contract management as document management alone, but rather understands the contact management lifecycle and supports the creation of closed loop auditable processes.

IBM announced on April 12, 2011, a change for selected products whereby they are moving from the Processor Value Unit (PVU) software license metric to the new Resource Value Unit (RVU) metric. This impacts several product families such as Tivoli Monitoring, Tivoli Provisioning and Service Delivery Manager. PVU and RVU metrics are radically different:

With the PVU metric, the capacity of the server where the application is installed is measured. If this application is moved to another server or the hardware properties (processors) of the server hosting it are changed, the license may be impacted. This can easily happen in virtual server environments. So, there is a risk of being out of license compliance.

With the RVU metric, the number of units of a specific resource used or managed by the application are counted. The application can be freely moved across servers and the hardware properties (processors) of the server hosting it can be modified with no impact on the license.

This makes perfect sense: the capacity of the server does not always correspond to the usage of the software installed on it. In the past, with the PVU metric, many IBM customers had to choose between price and performance when selecting hardware on which to install the software. With the RVU license metric, IBM customers will be able to configure and adjust the hardware capacity of the server hosting the application to meet their business performance requirements: there will be no impact on license compliance or license cost.

What is a Resource Value Unit? The purpose of the RVU license metric is to tie the number of licenses required, measured in RVUs, to the utilization of the software or the resources the software manages. RVUs may use– the number of Gigabytes, premium income, pages per month, number of messages, etc., as units of measure. The price of the license is based on the number of RVUs required. The RVU calculation is quite complex and usually uses a factor that is applied per quantity tier. In the case related to this IBM announcement, the number of activated processor cores for physical or virtual servers managed must be counted for RVU licensing. The factor table is as follow:

For Servers (active cores):

Quantity Factor
0 to 2,500 1.00
2,501 to 10,000 0.80
10,001 to 50,000 0.60
50,001 to 150,000 0.50
Over 150,001 0.20

Example: if a customer has 45,000 Activated Processor Cores across its managed servers, it will need 29,500 RVUs.

(2,500*1.00 + 7,500*0.80 + 35,000*0.60) = 29,500 RVUs

If the resource considered is a client device, the metric is just the number of client devices but a different factor table is applied. For managed servers, it is not that simple to get the right number of license entitlements needed, as it relies on a number of cores. It gets even more complex if the servers considered are deployed in virtual environments: the RVU license metric in this case supports sub-capacity licensing. In these environments, the measurement of Activated Processor Cores is based on the number of processor cores available for use. Rules and definitions are clearly defined by IBM but are specific to each virtual environment (virtual machines or hardware partitions).

Overall, customers will benefit from the IBM announcement and the use of the RVU software license metric: they will purchase a license from IBM that will be calculated based on what they use. However, moving managed servers across hosts in virtual environments or modifying their hardware characteristics can impact the license position in cases where the RVU is based on active cores available in those managed servers. Staying in license compliance is not going to be easier.