Never throw away a good idea, they say (or at least I do). The same applies to blogs that haven’t yet crystallised into a worthwhile topic. Generally, I put them on the back-burner until a number of paths align. After seeing the Oracle Cloud team present at Cloud Field Day 3 in April this year (2018), I knew there was a topic to cover, but I wasn’t sure exactly what it was. Hold that thought for a moment.
Last week, I attended a Virtustream press and analyst day in London. Both Oracle and Virtustream are pushing a variation of public cloud that doesn’t follow the hyper-scale model. Instead, the offerings are focused on moving traditional workloads from the data centre and running them in a managed environment. So is this idea of an “enterprise-class” public cloud something that has value – something that customers will buy into?
I’m going to make up some terminology. Hyperscale-class cloud (HCC) represents the type of service we expect from Amazon Web Services, Microsoft Azure and Google Cloud Platform. The underlying hardware is hyper-commodity and hyper-streamlined to support the IaaS, PaaS and SaaS offerings of these hyperscalers. You won’t see any SANs or enterprise-class servers in use here. Hardware failure is expected – although I would posit that it occurs at no greater frequency than in the enterprise. Resiliency should be built into the application and not be a function of the infrastructure.
The issue with this kind of platform is that “lift-and-shift” doesn’t generally work. Applications need to be re-written, redesigned or at best, modified to work with the public cloud. We expect application components (virtual instances, containers) to be easily replaceable – the pets vs cattle comparison. The component offerings by these providers aren’t generally customised – you take what they give you and continue to run operations yourself.
What if you don’t want to rewrite your applications? The cost of rewrite needs to be justified against some kind of saving. Re-writing in order to move to the public cloud is unlikely to save much. It makes sense to be “cloud first” and only update applications where rewrites are required for other business reasons. As an alternative, enterprise-class cloud (ECC) could be another option.
Rather than run with diminished SLAs and resilience from the infrastructure, why not pay for what is effectively a managed service and move applications into the likes of Oracle or Virtustream? Some applications like SAP simply won’t run effectively on the public cloud. Others may need hardware refresh but can’t justify the software rewrite costs and would make good candidates for outsourcing.
Oracle’s Cloud Infrastructure is pretty much enterprise hardware on demand. Check out the Cloud Field Day presentations from Oracle and you’ll see a service with no network over-subscription, high-performance storage and bare-metal servers. Oracle is strong in PaaS, including everything from data management (database), identity and security, systems management, application development and analytics. Obviously, you’d expect the company to be strong in databases, however (as you can see from the final presentation), Oracle has HPC offerings based bare metal servers and GPUs.
EMC acquired independent cloud provider Virtustream in 2015. With the EMC acquisition by Dell, Virtustream became a Dell property in 2016. Established in 2009, Virtustream had found a niche in cloud managed services that focuses mainly on SAP. Customers can lift and shift applications to Virtustream cloud, which runs on typical enterprise hardware from Dell. With a focus on a small number of applications, Virtustream offers deep subject matter knowledge and in fact, many of the SAP migrations (over 1000 since founding with 250 installations running today) are from non-x86 platforms to x86 hardware.
Finding a Niche
Both of these examples highlight the ability to offer customers something other than generic public cloud. In the case of Oracle, I expect they lead with databases and hope other applications will follow. In the case of Virtustream, the focus is SAP, with other areas such as healthcare now forming part of their offerings. What’s the real benefit here? In the same way as Exchange and messaging services were prime for outsourcing, applications like SAP that have a heavy set of infrastructure requirements and deep skill-set need, also flag up as good candidates. The service provider has the skill set to design, build and support the application. This is very different from the hyperscaler model where support stops at the infrastructure.
We hear a lot of talk about multi-cloud and for many large organisations, the idea of one single cloud would never make sense. It’s the technology equivalent of the metaphorical eggs in one basket. In any case, larger enterprises have never worked like this. Companies of a certain scale typically have “one of everything”. Walk around a typical data centre and it’s clear for anyone to see that over time, different vendors were in and out of fashion, as waves of technology were replaced.
- Building a Model of Hybrid Cloud
- Technology Choices for Data Mobility in Hybrid Cloud
- The Four Stages of Cloud Adoption
Why should cloud be any different? Private cloud, for example, may offer the best solution for production workloads, with test/dev in public cloud. For others, a mix of providers may be sensible. Some hyperscalers offer better SaaS/PaaS offerings than others, creating a mix of solutions being used. Here’s the issue though – interoperability. There are three main challenges:
- Data – how should data be moved around and processed? Does this mean multiple copies or exposing data access from one platform to another?
- Connectivity – how will networks be connected? How does the network remain protected? How is latency affected when services run in multiple locations?
- Security & Compliance – How can data and applications remain secure? How do security models interact? How can I audit access when my applications move around?
I’m simplifying here, as this post is already quite long. There are more issues to consider, but this gives some idea of the challenges of multi-cloud.
It may be that many organisations simply use multi-cloud tactically, placing applications statically with different providers. I can see that being the case for Virtustream and Oracle. Customers will see a use case for one or more applications and then leave them there until refresh time comes around. The challenges of interoperability will largely be ignored. As a result, do the enterprise-class cloud vendors have a good multi-cloud story? I don’t think they do – yet.
The Architect’s View
This post seems to have been a long-winded way of saying “horses for courses“. However, it does highlight that choosing a platform is about picking the best services to deliver the application. The most difficult challenge to overcome is how to stop enterprises from moving to a fragmented mess of multiple clouds, both private and public. I suspect we already have some of that, with experiments in using OpenStack, CloudStack and deployments using VMware and HCI solutions like Nutanix. Hyperscale public cloud is not the solution for everything and I think application-specific niches will continue to exist. It’s the classic maturity model challenge. On the positive side, it means we have the need for new skills and roles to address the complexity of delivering efficient information technology.
Further Reading & Watching
Comments are always welcome; please read our Comments Policy. If you have any related links of interest, please feel free to add them as a comment for consideration.
Copyright (c) 2007-2019 – Post #72b8 – Brookend Ltd, first published on https://www.architecting.it/blog, do not reproduce without permission.