Building up a public cloud is quite a challenge. Oracle has faced and overcome many challenges – and it has built up tremendous momentum to tackle the current challenge. It is stretching the company – threatening to make it spread its resources thinly. The goal is clear, the focus is strong. And the belief as well – the belief that Oracle can offer its customers an enterprise grade cloud experience – in functional and non-functional terms. And at all levels of the cloud stack.
Note: read my previous article on the value proposition of the public cloud to companies around the world.
Compared to a year ago, large strides were made. Many aspects of the cloud were hardly more than plans and ideas a 12 months ago and have now come a long way towards realization. Apart from the substance that has been added, there seems to be much more coherence and vision. A year ago, dozens of teams within Oracle – under the threat of ‘be in the cloud very soon or perish’- were racing to the cloud, quickly lifting and shifting their on premises product to a cloud offering or somewhat hastily building a cloud offering from scratch. There was little integration between all offerings and a lot of overlap. Several important foundational pieces were not available. And the underlying infrastructure was overburdened as were its operations teams.
Many lessons have been learned. And an architecture vision has been applied at various levels. And much required progress has been with core pieces in the cloud puzzle. And a number of designated capabilities is now in place – to be reused across the cloud. Document Cloud Service for example handles digital asset management. Any document – or other form of unstructured digital asset – is taken care of by Doc CS. Just as all user management and authentication is handled by Identity Cloud Service (IDCS) – to be leveraged by every other cloud service. And all events are to be exchanged through the Apache Kafka based PaaS Event Bus – that was alluded to.
There is a PaaS “fabric” – PaaS Service Manager – that provides facilities such as patching, notification, logging, scaling and command line management to all PaaS services.
The user experience is more consistent with all services acting in the same way, the development is more efficient and the evolution can be much faster. A win-win situation.
One of the next steps is that these cross the cloud capabilities are implemented, exposed and managed as microservices, making reuse and independent development even easier.
Cloud First or From Lift & Shift to Migrate & Extend
Last year, Oracle stressed the lift & shift [workloads from on premises to the public cloud and vice versa] approach – facilitated by the parity between cloud and on premises: same software, same architecture and same skills. Anything that runs on premises can run unchanged in the cloud. This year, instead of lift & shift, the motto is migrate & extend (& modernize).
This seems much less intended to be bi-directional. The suggestion that the same PaaS software would be available in the cloud and on premises and would evolve along similar timelines and release schedules that we took away from Oracle OpenWorld 2015 is no longer valid.
Oracle has explicitly stated a cloud first strategy: new functionality will first be released on the cloud. And perhaps later on premises as well. The first example of this approach: Oracle Database 12cR2. The next major release of the flagship product. It is available in the cloud, since September. We are still waiting to be informed at what moment this software can be downloaded to be installed on premises.
For some products it seems that the strategy is not just cloud first but probably cloud only or at least cloud mainly. A number of products – for example WebCenter Content and Sites, BPM Suite – seem to have come to a complete stop in terms of their on-premises life cycle. Other products will see some evolution, but are quite clearly no longer at the core of their team’s strategy; SOA Suite for example will gradually make way of ICS (Integration Cloud Service) and RUEI and BTM are the poor step-siblings of Oracle Management Cloud. This may force some organizations to perhaps sooner reconsider their on premises vs. cloud strategy than they intended.
The facilities in both Oracle Database – Pluggable Databases (PDBs) – and WebLogic – partitions – to move workloads from on premise to cloud and back again are quite strong. From tools like SQL Developer and Enterprise Manager – it is quite simple to move PDBs and Partitions back and forth from on premises to DBaaS and JCS.
The App2Cloud (sometimes AppToCloud) tooling is available in various forms and shapes, including in Enterprise Manager 13c Cloud Control , to help quickly migrate existing Java applications and their supporting Oracle WebLogic Server resources to Oracle Java Cloud Service. App2Cloud will reconfigure mutual URLs and IP references between multiple managed servers. There is no Cloud2App counterpart.
If you would like to subscribe to the services Oracle offers from the public cloud, but either you are bound by legal rules or regulations that mandate your data to stay behind your firewall or your network connections to the internet and the public cloud data center is of limited capabilities, there is the cloud@customer proposition. It means: the public cloud service under the public cloud conditions – running behind your firewall, locally attached to your network.
And yes: hardware is involved, installed by Oracle on your premises. However, you subscribe to the service that is fully managed by Oracle. The fact that this machine is sitting there and is making calls out to the Oracle public cloud data center to allow Oracle staff to manage the machine should be irrelevant and invisible. The main considerations for engaging with this cloud@customer are the minimum 3 year subscription period and minimum volume: you subscribe to the same services that you subscribe to in the real public cloud – and you start with at least 288 OCPUs of Compute. How you make use of these OCPUs is up to you – you can subscribe to and run DBaaS, JCS, SOA CS, ACCS and other PaaS services on that local public cloud. Or your own workloads on virtualized compute.
Becoming a Public Cloud Provider
Oracle is struggling with several fundamental aspects of becoming a public cloud provider. Just how to define the usage metrics on which to base the pay per usage models remains a tough question. It is clear what customers want from the cloud provider: pay for real usage, scale up and down in very small chunks and not by large chunks such as whole OCPU at a time, and start and stop with very short notice. Long term subscriptions are understandable with cloud@customer which involves manual work and dedicated resources. And a coarse grained large chunk approach is also evident for the BareMetal Cloud Service and the Exadata Cloud Service. But these should be the exceptions.
With many cloud services now leveraging other services for their specific capabilities – such as PCS (Process Cloud Service) using ICS for all interactions with external systems and Doc CS for the associated digital assets and shortly Application Builder CS for creating the user interface for the human task steps in the processes – the question becomes relevant whether customers pay just for the core capability they sign up for and get limited usage access to associated services or if they have to subscribe to each service that is required for using the one service they are interested in.
Some overtures were made around suites and editions. A suite is a cluster of various services that customers can subscribe together – instead of to individual services. What that means in terms of pricing was not revealed – and may not even have been decided. Another dimension discussed were the editions: for example personal, workgroup and enterprise edition. It did not become apparent what editions represent: shapes or usage levels, restricted functionality within the suite or a limited number of users for unlimited functionality or something else altogether.
The challenge is even bigger than most of us realize: whatever model we can think of – how sensible it may sound – also needs to fit in with the Oracle organization: the public cloud infrastructure needs to be able to record, measure and bill according to the subscription definitions.
And: the compensation for the Oracle sales force somehow needs to be redefined in such a way that sales people are stimulated in the right way. Traditionally, targets are set on license based revenue, on a quarterly basis. With scalable subscriptions that can be stopped at any time – there is no such concept of ‘when is a deal made for a specific chunk of revenue’. Even though we may consider this a somewhat silly challenge – it is still very important to Oracle Corporation and probably one of the tough nuts to crack.
Download the AMIS OOW16 Highlights for an overview of announcements at OOW16.