From time to time, I am sent proactive, preemptive commentary from a supplier. That supplier hopes to focus my attention on something, change my opinion to something closer to their opinion, or in some other away have an impact on my research. Since I'm open to learn from executives, I almost always read their commentary. On occasion, I republish something here. This is just such an occasion. Ian Knox, Director of Product Management at Skytap, had the following commentary sent along to me.
What Ian Knox of Skytap has to say about VMware's launch of vSphere
VMware is set to release its virtual infrastructure platform under the ‘vSphere’ brand. VMware has encouraged customers to virtualize their IT infrastructure within ‘private clouds’ and automate provisioning and management to lower costs and increase business agility.
VMware, of course, also has its vCloud initiative which is designed to enable service providers (such as Skytap) to deliver external clouds to provide on-demand capacity. Given VMware’s announcement and recent McKinsey analysis showing external clouds are too expensive, should customers now rush to build private clouds? What about public clouds, should they still be used? The answer, of course, is ‘it depends’.
It’s useful to go back to basics and remind ourselves of the key characteristics that define a ‘cloud’. I’ve listed these below with a subjective ‘grade’ for private clouds and public clouds:
The characteristic common to both private and public clouds is that they are virtualized, which enables provisioning of flexible computing resources. Aside from that, we see some key differences, which can be categorized into 3 areas: (1) Time-to-Value, (2) Economic and (3) Service Levels.
Time-to-ValuePublic clouds are on-demand, which means new resources can typically be deployed very quickly (typically hours, if not minutes) through a self-service portal or an API. They are also scalable, which enables incremental resources can be added as required by the business. In contrast, private clouds take time to implement and have a fixed capacity. Unless a private cloud is very large (for instance, the type of private cloud found at a Wall St investment bank), a public cloud will almost always be faster to provision and scale than a private cloud. In addition, private clouds are typically not self-service and don’t offer an API, although this will change over time.
EconomicPrivate clouds require an up-front capital investment and internal resources to build and manage the infrastructure. In contrast, public clouds are offered as a utility, enabling capital expenses to be exchanged for ‘pay-per-use’ operational expenses. In some cases, private clouds offer a similar ‘utility billing’ option, but this is usually only for very large IT deployments that provide IT services to internal customers.
Aside from cap-ex, the killer economic benefit for public clouds is that they are multi-tenant and scalable, which allows ‘time sharing’ and ‘bursting’ of resources. There are several usage scenarios that are highly dynamic (see below) that make public clouds much more compelling option than private clouds. However there are also some steady-state usage scenarios for which private clouds are very cost competitive (as McKinsey pointed out in its analysis).
Service LevelsA public cloud typically publishes an SLA agreement, with a goal of uptime (e.g. 99.9%) and rebates to a customer if this service level is not met. If a customer is not happy with a public cloud supplier, they can take their business elsewhere. In contrast, an IT organization sets SLAs for internal customers. Although it’s possible for an IT organization to meet or exceed the service levels of a public cloud, if they don’t there’s very little flexibility or competitive pressure to improve the situation. In essence, internal customers can be ‘locked in’ to a private cloud with no option to move.
William Fellows at the 451 Group, recently published some research that included a survey on what public clouds are currently being used for. The top five usage scenarios are:
(1) Internet Application Hosting (2) Databases (3) Disaster Recovery (4) Remote Storage (5) Application Testing & Development
Not unsurprisingly, these scenarios take full advantage of the on-demand and scalable nature of the cloud, especially when the scenario is highly dynamic. Our customers, for instance, utilize our cloud for virtual IT labs, which are used for application testing and development, virtual learning and sales demos. Typically, these usage scenarios are very dynamic, in that utilization varies from one application development project to the next, or from training class to training class. In addition, resources often have to be scaled up and down to cope with variable demand. This is a perfect match for public cloud computing and Gartner commented recently that Skytap has the ‘potential to dramatically reduce total cost of ownership for test labs by 25% to 50% annually’.
So, it’s not time to say farewell to the public cloud – in fact, there are some use cases which are very compelling from a business and financial perspective and trump almost any internal cloud. However, it’s important to recognize the strong usage scenarios for private clouds, including steady-state workloads such as file/print servers, predictable production applications etc.
In our discussions with customers, we’ve found those taking a ‘hybrid approach’, i.e. using public and private clouds together are gleaning the most financial benefit. As an example, we showcased ‘hybrid’ cloud computing back at VMworld in September, whereby application development and testing can be done in the cloud with access back to onsite resources as required. So, the bottom line is to do the business and financial analysis and pick either a public or private cloud based on your criteria and circumstances.
While it is clear that Ian is trying to use VMware's launch of vSphere to benefit his own firm, some of the ideas presented in his commentary appear quite sound. What do you say? Do you agree with Ian?