Not surprisingly, since the release of my new book, Visible Ops â€“ Private Cloud, I have been talking with a lot of people about how to deploy private cloud, where to start, what to avoid, etc. So far, the most common question has been, â€œWhat type of existing workloads are organizations putting into private cloud environments today – and what are they avoiding?â€
So I thought I would jot down some of my answers, specifically related to ‘cloud-migrant’ services, as opposed to ‘cloud-native’ services – and without getting too hung up on whether the use cases are 100% cloud or not!
One recurrent use case is to provide dynamic desktop allocation, especially for education and projects use cases. A number of schools, universities, training centers, and even some larger enterprises, have adopted private cloud to allocate servers, clients, applications and data for reusable desktop systems.
This seems especially prevalent for short-term learning facilities, repeatable one-off classroom systems, training/demo labs at conventions (or user groups), and contractor setup. It is also similar to the executive briefing centers and ‘demos on demand’ that many software sales organizations (like CA Technologies) use.
Another service-based use case I have seen in several universities is self-service access for students and faculty, using pooled resources, not only for application services but also for full VDI desktop allocation.
I have seen this in other enterprises too – most notably for home-source process workers (e.g. call center, data entry) – but mostly as a proof-of-concept, not a large-scale production deployment.
However, most cloud-migrant workloads I see deployed to private clouds today still tend to be server-based. Most of these are at ‘Phase 1’ in the Visible Ops Private Cloud – a reorientation of virtualization deployments to pilot a private cloud that works, proving results, gaining skills, and hopefully measuring opportunities. It is still focused on servers, not services, but provides a vital part of the learning curve toward private cloud.
- Dev/test/QA servers – 3-tier LAMP stacks (app/Db/WS), but also LAMP components, IDEs, source code management tools, etc. (which often results in applications that run on a private cloud in production)
- Collaboration servers – especially SharePoint, but also Web-based collaboration services like team chat servers, content repositories, blogs, wikis, and project management tools
- Engineering servers â€“ I have seen a number of engineering firms move their design project systems (especially CAD tools) into private clouds so engineers can fire up new design projects on-demand
- Web servers – popular for marketing teams who can fire up their own Web servers, especially for short-term and/or localized promotions & campaigns
- Analytics servers – short-term number crunching of ‘big data’ (including BI applications) in medical research, social marketing, pharmaceutical research, higher education, financial, logistics, etc
The workloads that are less suited to private cloud deployment are harder to identify, because it requires positive evidence of absence, so my thoughts here are much more anecdotal. I do see CIOs push back on migrating â€˜coreâ€™ applications, even to private clouds, citing lack of confidence, performance concerns, potential security and compliance issues, and lack of ROI. I would not agree these are always good reasons, but they can be, and are certainly understandable.
In my opinion, private cloud is not ideally suited to relatively large, static, predictable, and resource-saturating workloads – think ERP or Data Warehouse. After all, used internally such applications are almost never deployed â€˜on demandâ€™; they are rarely if ever â€˜multi-tenantâ€™; they have no real benefit from an â€˜infinitely scalableâ€™ infrastructure; and are mostly viewed as a cost of doing business, without any ‘resource measurement’ or chargeback.
(btw, there are certainly good arguments to deploy these applications on a public cloud, as ‘cloud-native’ services using SaaS, to outsource them to a non-cloud third-party, or to just virtualize them – even with 1:1 virtualization – without the other trappings of cloud. Such alternatives could deliver better cost savings, higher up-time, faster DR, and other benefits. However, I think the upside of putting such applications in a private cloud is less apparent.)
That said, I do think that we will see more and more strategic services – as opposed to project servers – deployed in both private and public cloud as it matures. In fact, recent IDC data suggests CIOs that are adopting private cloud will migrate many core applications in the coming years. Moreover, some of the more advanced customers I talk with are already doing this, although they are by far in the minority.
Either way, I will be very interested to see how this all pans out.
What do you think? What have I missed? What types of workloads do you see being deployed in a private cloud? What are CIOs passing over in their evaluations? Are they right, or wrong? What criteria should they use?
Please feel free to continue the discussion in the comments below, or hit me up on Twitter with your ideas.
This post was originally published on the CA Communities website.