Dave Welch, CTO and Chief Evangelist

A review of Gartner’s November 24, 2014 “Oracle’s Licensing Policy for Server Virtualization Is a Substantial Audit Risk” Paper.

The paper, document G00271412, was $195 when originally published. It was free with registration as of February 16, 2015, and is again $195 at the time of this posting.

I’ve been busy with other things and haven’t prioritized authoring and publishing a comprehensive review until now. I drafted my initial analysis of the paper in January and have revised and re-written the draft since. I also confess I’ve been reluctant to post out of respect for the Gartner brand and industry colleagues there.

At House of Brick, we take no prisoners in our analysis of peer-drafted materials. I’ve decided to frame my critique of the Gartner paper similarly, as if authors Stewart Buchanan, Jane Disbrow and I were all on the same team. I don’t know that I can be any more fair and deferent than that. As such, I’ll ask for everyone’s indulgence as I play the role of Gartner team member. For this exercise, pretend the clock just moved back to early November, 2014.

(Begin role-play)

Stewart & Jane, thanks for the effort you put into drafting this piece.

Summary Impression

Overall, I see potential in some content in the draft. My concern is that, as written, it comes across to me as contradictory throughout and therefore confusing. If I’m right about that confusion, it will leave many readers with a feeling of uncertainty which could play right into Oracle’s hands with its non-contractual VMware assertions. The world looks to Gartner to make clear, authoritative statements. As such, I believe this draft, if published, would significantly erode our brand and credibility at least with respect to Oracle licensing. I also see recommendations here in terms of attempting to operationally involve Oracle in architecture that I believe are sub-optimal for licensees. Furthermore, the draft reads as if we are recommending that Oracle amend its contract to operationally bind the reseller – something I haven’t noticed precedent for in Oracle paper, and therefore am predicting Oracle will be unwilling to accommodate.

This piece absolutely cannot go out the door as written. With that, I ask for your understanding in advance for the line-item review that follows.

Detailed Review

Whether or not the draft title is accurate is irrelevant to me; all licensees should pay every penny they have contracted to pay. I believe the financial, operational, and business continuity benefits of running Oracle on VMware are so compelling so as to overwhelm any potential inconvenience of an audit. I don’t think we should be publishing anything that may imply that we are suggesting our subscribers should avoid asserting their contractual rights due to fear. As such, I suggest a reworked title. Maybe “Oracle’s Non-Contractual Licensing Policy for Server Virtualization May Increase Audit Risk, but the Financial and Operational Benefits Far Outstrip That Manageable Inconvenience.” That’s too wordy and needs to be reworded so as to not expand the scope of the piece. Let me think on it. No doubt you and/or I will come up with a better title that avoids the liability I’m concerned about with the current title.

The draft’s summary begins, “As a consequence of new virtualization capabilities, IT procurement professionals in enterprises have to buy more Oracle technology licenses to cover virtualized servers after Oracle audits…” Stop right there. That absolute, unqualified statement doesn’t work for me in any association to the piece, let alone in the first sentence of the summary. Au contraire, we understand no licensee that has Oracle programs installed and/or running on physical servers independent of virtualization in a manner that doesn’t exceed the licensee’s entitlement has to buy more licenses. The statement is invalid as is regardless of the brand of virtualization.

Yes, we’ve seen Oracle LMS’s Final Reports (audit finding) that claim that minimally all vSphere servers in a cluster must be covered by entitlement, not just the subset of vSphere servers with Oracle installed and/or running. Such a non-contractual claim doesn’t translate to “have to buy.” Furthermore, I’m betting there will be some small subset of readers that will read no further than that or skim-read the article so as to miss the necessary clarifications.

Any assertion that virtualization in and of itself induces incremental licensing obligations is invalid as virtualization is not mentioned in the Oracle contract or the documents the contract integrates by direct reference. Therefore, the assertion is locked out by the contract’s Entire Agreement clause.

The abstract’s second sentence: “Oracle software customers using virtualization face a dramatic multiplication of their licensing costs due to Oracle’s changing interpretation of its partitioning policy for virtualization.” That sentence has to be reworded and clarified. Oracle’s interpretation of its non-contractual partitioning policy is moot as Oracle’s Partitioning Policy document always has been, and still is, non-contractual. The policy is non-contractual due to the Entire Agreement clause mentioned above.

The abstract’s third sentence: “Oracle customers are often proposed Unlimited License Agreements to settle virtualization compliance issues, but this only fixes the issue temporarily and does not overrule the partitioning policy.” Whether or not the contractual Unlimited License Amendment (rewording mine) contained anything in conflict with the partitioning policy is moot as the partitioning policy is non-contractual.

The abstract recommends pushing back on Oracle’s virtualization policy. May I suggest the best form of pushing back is ignoring Oracle’s virtualization policy? My experience and observations suggest that provides a consistent result with relatively scoped effort. The paragraph continues: “Some organizations have effectively done this when the policy caused them to fail an audit, because it is not documented or even referenced in their license agreements.” That’s the right idea, but not strongly-enough worded; recipients of such audit deficiency claims have the ability to prevail on the point 100% of the time in my experience. The abstract accurately points out “Only what is included in the actual contract will count as being valid,” and the virtualization policy is not documented or even referenced in customers’ license agreements, but then puts forth various conclusions and recommendations that are in complete contradiction to the observation.

Next, there’s a sentence that I interpret to say that virtualization makes the so-called 10-Day Rule moot. No such statement is found in the contract. Therefore any such statement is excluded by the Entire Agreement clause.

The draft’s introduction states, “Oracle no longer accepts the partitioning of servers by storage as a limit to the number of servers that require licensing.” Have a look at House of Brick’s October 3, 2014 blog post “Managing Oracle Licensing in a Shared Storage Environment” to be directed to a section of the Oracle Software Investment Guide that spells out unambiguously just how Oracle most definitely accepts the partitioning of servers by storage as a limit to the number of servers that require licensing. Whether or not we name House of Brick in the draft’s trailing Evidence header’s third party references is unimportant to me. What I am recommending is that we leverage Oracle’s own published documentation as an illustration of what Oracle must indeed accept as it is not disallowed by anything in the contract.

The document’s analysis section states, “… the rise in server virtualization and shared storage add complexity, because all physical servers that could access the shared storage need licensing.” Again, have a look at House of Brick’s October 3, 2014 blog post to understand why I have to push back on that statement’s accuracy as is.

Another sentence in the analysis section: “As the Oracle license agreement neither addresses nor defines Oracle’s partitioning policy (or even refers to its existence obliquely), many Oracle customers believe they only have to license processors where the software was actually running within a physical server.” And those customers’ belief is accurate. (See the House of Brick September 17, 2014 blog post “Oracle Licensing: Soft Partitioning is Your Contractual Right on VMware.” Should you choose to reference that, we can discuss the pros and cons of including in the footer an encouragement to read the post to see why House of Brick doesn’t currently endorse VMware Core Affinity with Oracle on VMware.)

In the analysis section: “Clients with good license management responded by building a smaller server farm dedicated to Oracle on separate storage and LANs. … Yet Oracle does not accept software configurations able to be changed with a few clicks of a mouse.” While I have seen some customers air gap due to their scale and for their own crass selfish operational interests, I don’t agree that good Oracle license management alone has ever mandated such air gapping. As for the last assertion, it comes to mind that humans live every day of their lives in situations where the capability is immediately present to injure others or themselves with ease inside a few motions. That they have the ability to do so doesn’t mean they are prospectively guilty of doing so. I could use numerous other illustrations but hopefully you get my point.

“Unfortunately … wherever vSphere 5.1 or higher is deployed, shared storage boundaries can no longer be relied on to limit licensing liability.” That’s not correct for reasons already stated above; licensees are not guilty until they violate the boundary.

“In the same way, (that comma is grammatically inappropriate) vMotion 3 overcame server boundaries, vMotion 5.1 transcended storage boundaries. VMware inadvertently gave Oracle the opportunity to charge mutual customers even more.” No VMware didn’t. Again, no one is guilty until a boundary is violated.

The draft makes these recommendations:

• “Work with Oracle user groups to push for more reasonable virtualization policies that do not penalize clients by charging for processors that will never be used in a virtualized environment.” As there is indeed is no penalty with contractual merit, I see it as a complete waste of time for Gartner to encourage licensees to badger Oracle to affirm contractual rights the licensees already enjoy.

• “Obtain examples from Oracle that illustrate how its partitioning policy has an impact on your virtual environment to ensure that you have correctly understood and applied current policy.” To what end? In our experience and observation, Oracle responds again with non-contractual assertions, which if the licensees follow, limit their operational agility and inflate their expense.

• “… it is not always clear how many licenses Oracle will certify at the end of the ULA contract, so clarify any limitation on certification of licenses and ensure that it includes all licenses that Oracle requires, based on then-current policies.” In my experience and observation, the ULA Certification section needs no clarification. (To say it needs no clarification doesn’t say I like everything about it.) The contractual ULA Certification clause certainly is not subject to then-current non-contractual policies. ULA C.1.B. states the certification entitlement is based on where the Oracle programs are “installed and running” (note that’s no longer the contract’s “installed and/or running”). As such, ULA customers would have no more contractual right to inflate their entitlement upon certification through claiming for example the entire vSphere cluster beyond an Oracle sub-cluster than Oracle had to obligate licensing for the same with a standard non-ULA license.

• “Move Oracle workloads to an Oracle-approved cloud provider (Amazon or Microsoft Azure) but ensure the provider is responsible for demonstrating compliance if you are buying your licenses from these providers.” Although Oracle’s “Licensing Oracle Software in the Cloud Computing Environment” policy document is not referenced in the contract, would Gartner Legal predict that a court would find this under the category of an implied license? (Related to this, a licensor can grant privileges with an implied license but cannot take contractual privileges away through such a vehicle.) The policy clearly grants “by-the-drink” core-capacity licensing rights in the Amazon or Microsoft Azure hosting environments wherever the VM may float to within the hosting provider’s server farm. As such, licensees of those providers have no need to ensure the provider is responsible for demonstrating compliance.

• “Adopt Oracle VM with CPU binding for your Oracle environment if your requirements can be met within these limitations.” Rather, why should licensees not affirm their contractual privileges as opposed to limiting themselves to inferior features, let alone with what at least one leading virtualization integrator describes as a consistent experience with less than stable tooling?

• “If all other options are exhausted, architect Oracle out of the environment …” But we know that’s not necessary. Again, it may be that licensees rather are inclined to affirm their contractual privileges and stick with the RDBMS that they have spend untold time and money integrating and which leads the world in market share that has better tooling, speed, capacity, range, and altitude.

• “Database managers often research the many sources of potentially conflicting information available freely on the Internet, such as an incorrect historical opinion expressed on the VMware website.⁵” The footnote says, “See Oracle Virtualization: Run Databases, Middleware and Applications on VMware Virtualizing Oracle With VMware.” What incorrect historical opinion? I confess I personally have no idea what you’re referring to as the URL links to a vmware.com index page hosting several articles. We can’t publish that as is without specifically identifying the incorrect historical opinion being referred to.

• “A vendor account manager selling a virtualization product to virtualize an Oracle Database isn’t necessarily an expert in Oracle’s current licensing policy.” It should be of no concern to us that such a virtualization vendor is illiterate in non-contractual policies as long as the vendor doesn’t encourage the customer to do anything other than architect in a way that Oracle licenses are optimized to the minimum number of physical hypervisor-enabled servers where the Oracle programs are “installed and/or running.” Two sentences down: “Most importantly, it may transpire that none of these vendors are under contract to give licensing advice.” Huh? This caution appears to me to be no more useful than a comment on a Computer World article written by Robert Mitchell in early 2014. The commenter stated that only LMS is authorized to interpret the contract and went on to disparage that one of the contributors wasn’t even an Oracle partner. Even licensing “advice” tendered by Oracle Corporate employees is locked out by the contract’s Entire Agreement clause the moment they say or write anything in conflict with the contract.

• “Negotiate contract terms to hold your Oracle consultant or license reseller responsible for obtaining all the relevant information regarding your Oracle requirements. This means that any omissions on your part don’t give Oracle an excuse for getting the licensing quantities wrong.” I’m sorry, but I don’t foresee Oracle being willing to amend the contract to operationally bind a reseller in that manner. Furthermore, I’m betting Oracle would be only too happy for licensees to first specify their architecture blind to licensing and then have Oracle come along afterward and price it. Rather, Gartner subscribers should have Oracle licensing in mind in the earliest stages of architectural development. For many shops, it’s their most expensive system stack component.

As always, thanks for your patience with my review.

I’d be happy to review this with you in a call or WebEx if that would speed up the drafting/review cycle. Regardless, there’s enough here that I’m uncomfortable with that I’m asking to review your next draft revision pre-publication.

(End of role-play)

Gartner and Oracle documents referenced are copyright their respective corporations. Quotes from Gartner and Oracle documents are per U.S. copyright law fair use doctrine.