The battle for the public cloud is intensifying. Choices made by practitioners in the next year will have long-term business consequences. The entrants are well known: AWS, Microsoft, Google, Salesforce, ServiceNow, IBM, Oracle, SAP and a host of others contenders. The benefits to organizations are increased IT economic flexibility and reduced infrastructure constraints.

In a conversation with Wikibon chief research officer Peter Burris recently, he pointed out that one infrastructure constraint will not go away, and it’s the one set in stone about 13.82 billion years ago when the universe began: the speed of light. Peter explained that nobody’s quite sure why light travels at the same speed in all places and at all times, but it does. And all places, all times includes between the cloud and your data centers, or between your data centers.

To the edge and back

Peter emphasized that the speed of light will dictate the final shape of the cloud, which will — not might, will — include edge elements that place processing power close to a problem being processed. Internet of Things cloud solutions, for example, must include edge elements. And the most interesting computing challenges in digital business are more like IoT than OLTP (online transaction processing). No cloud strategy is complete without a clear plan to accommodate edge-like workloads.

Some cloud vendors can’t support edge requirements. Some can. But if you’re going to establish a long-term plan for running your business in the cloud, know that some relatively un-sexy technology is going to play a major role in your decision.

At the top of the list is technology that blasts huge volumes of data from place to place. These un-sexy technologies won’t help you with speed of light challenges, but they will help you with volume challenges. For example, the models that will drive machine learning applications likely will be constructed and enhanced at the center, and deployed at the edge. Many of these apps will have to blast sensor data from the edge back to model development so that the learning can take place. Once that’s done, updates to the model will be sent back to the edge, but the amount of data required to update the model usually will be a fraction of the raw data sent from edge to center.

Managing data movement is going to become a much bigger problem and will consume a lot of architecture design time. We’d like to think the internet makes networking “free,” but that is not the case. Network link costs for intercloud — or intra-hybrid cloud — applications will be substantial.

One of the vendors that’s betting big on what Wikibon calls “true hybrid cloud” — a cloud strategy that facilitates local and global execution and management under a common cloud regime — is Oracle. One of the unsexy technologies that makes Oracle’s promises feasible is good old ZFS.

ZFS has been around for years. It’s tried. It’s true. It’s industry-hardened. There are Linux open source varieties out there. It was originally designed for supporting the type of massive data transfers that for years were only encountered in high-performance and scientific computing work, but now are showing up routinely in big data applications. It’s a bandwidth beast.

ZFS is a file system. It’s not sexy. It’s basic. It gets the job done. And as you think about your cloud future, it’s one of the technologies that won’t excite your execs, but will help you sleep at night.

Oracle cloud journeys

At its annual Oracle Openworld customer conference this year, Executive Vice President David Donatelli (pictured above) laid out five so-called cloud journeys – i.e. paths to the Oracle cloud. We met with Donatelli just weeks after he joined Oracle, and we asked him why he chose to reunite with Mark Hurd, former chief executive of Hewlett-Packard and the individual who originally hired Donatelli away from EMC. His answer was cloud. He made the following statement about the competitive landscape: “If you don’t have a public cloud strategy, you’re in big trouble.”

Oracle EVP David Donatelli: Five Journeys to the Oracle Cloud

Donatelli’s point was that, unlike the so-called arms dealers of the industry — i.e. Dell EMC, HPE, Lenovo and the ODMs — Oracle owns its public cloud and can compete directly with the likes of Amazon, Microsoft and Google for cloud share. The arms dealers can sell to public cloud players but can’t offer the higher margin services associated with the public cloud directly to customers.

As we’ve pointed out many times, to compete for Infrastructure as a Service in the public cloud, you must have massive scale or some substantial differentiation from the big three public cloud players. Donatelli’s point was that Oracle’s position in database, PaaS and SaaS uniquely positions the company to compete in all layers of the stack, including core infrastructure. It remains to be seen how the recent AWS VMware deal will impact this dynamic.

Take ZFS as an example. One of the few bright spots in the storage portfolio of Sun Microsystems when Oracle purchased the company was ZFS. Oracle has done an excellent job of evolving its ZFS architecture to create a very competitive NAS product. Recently, Oracle announced the ZS5, a fifth generation system based on the ZFS file system. It’s a mature product that is finding a place with Oracle customers, especially in high bandwidth workload environments.

ZFS storage is core infrastructure and the ZS5 is a bigger, better, faster version of the previous products in the series. Hundreds of TBs of flash, tons of capacity and super bandwidth to help with brute force  constrained apps (like backup and big data). The key for Oracle is the way the system fits into its cloud strategy. Oracle has a “Same:Same” approach meaning the same technology used on-premises is used in the cloud.

Oracle claims its public cloud uses over 400 petabytes of ZFS Storage Appliances, implying the architecture is proven at scale – (although details of those claims have not been research by Wikibon). With the ability to run the same systems on-prem and in the public cloud, Oracle provides a level of architectural equivalency that differentiates from the arms dealers.

That said, the arms dealers will attempt to mimic this capability through partnerships– emphasizing choice and openness. Oracle will pooh-pooh this as impossible without deep engineering integration. But the IT industry’s history shows that it’s achievable and sometimes more attractive because it increases customer options — think Wintel and Android — but admittedly sacrifices deep integration (think iPhone v Android).

Religious debates aside, as we see it, ZFS fits into Donatelli’s Journey slide above within three of the five scenarios:

Scenario 1. Legacy Data Center → Engineered Systems/Storage → Public Cloud, where current legacy infrastructure is “optimized” (I think that means forklift upgraded in Oracle parlance) to Oracle engineered systems or storage that has affinity to Oracle software and then can be migrated to the cloud over time.

Scenario 3. Legacy Data Center → Hybrid Cloud → Public Cloud, running between data centers and public clouds, where you install Oracle infrastructure that has database/application knowledge that allows you to run apps on prem or in the cloud.

Scenario 4: Legacy Data Center → Private Cloud → Public Cloud, where you build a so-called Private Cloud (meaning whatever you want it to be) and if you use Oracle hardware you’ll be able to move to the public cloud over time.

The key to Oracle’s strategy is homogeneity. It understands that to the extent public and on-prem infrastructure use the same technology and processes, the more efficient and higher performance applications will behave. The four questions customers must ask Oracle are:

1. How mature are Oracle’s cloud offerings?
2. Are they truly integrated and is the experience competitive to alternatives?
3. Is Oracle’s IaaS/Infrastructure pricing really as competitive as Oracle says it is?
4. Does the business value I receive from going with an all-Oracle solution offset the longer term lock-in risk?

At Oracle Openworld, Larry Ellison attempted to address some of these questions. He claimed that Oracle Cloud can run analytic queries 100X faster than Amazon Redshift and that AWS is more closed than the IBM Mainframe. He said that Oracle Database, MySQL and NoSQL DB products can all be run on-premises on any server, in the AWS cloud, in the Oracle Cloud and in Microsoft Azure, while AWS Aurora, Redshift and Dynamo DB can only run in AWS, making for a true lock-in scenario. Ellison also pointed out that mainframe software such as MVS, CICS and IMS ran on mainframe clones from Amdahl and Hitachi, whereas AWS has yet to let customers run its software in any other environment.

Do Ellison’s claims adequately address the above concerns? No, but they give a glimpse as to how Oracle plans to compete in this moving target of cloud.

There’s little debate that AWS is the “Mother of all Lock-ins” – perhaps even more so than Oracle. But customers are voting with their wallets and AWS is winning – as Oracle has for years, despite its lock-in advantage. In fact the history of the IT industry has recorded many examples of lock-in strategies winning for extended time periods. Indeed, Wikibon research shows that 85 percent of CIOs will choose to endure the risk of lock-in if the solution delivers business value.

Lock-in or not, winning approaches require investments, innovation, references and maturity. These are the factors that will adequately address the above four questions. Donatelli may prove to be correct in asserting icebergs ahead for those without a public cloud. While we don’t believe the arms dealers are all screwed, if successful, Oracle’s margin model will be significantly more attractive.

That much is not cloudy.

Wikibon is owned by the same parent company as SiliconANGLE.

Photo by SiliconANGLE