Cloud computing is high on the agenda for all UK businesses, particularly for companies with substantial IT infrastructure – hosting providers, and large enterprises. Ditlev Bredahl, CEO at OnApp explains why hosting companies have a substantial market opportunity selling more flexible, profitable cloud services to traditional dedicated and shared hosting customers – who in turn want the increased reliability, convenience and pay-per-use pricing of the cloud. In the enterprise, IT infrastructure is in transition from silos of compute and storage to private clouds capable of flexible resource allocation, greater scalability, increased utilisation and better cost efficiency.
Whether building a private cloud for internal users, or a public cloud on which to host paying customers, the first challenge is finding the best way to “cloud-enable” that IT infrastructure.
While some companies build their own clouds from scratch, the quickest path to the cloud, for most, is to invest in a third party cloud deployment and management tool. There are many available, and this is a guide to the ten things a would-be cloud businesses should consider when choosing the right tool, and the right provider, to help build and manage their cloud.
Top of the list of concerns for most cloud deployment projects: how much it will cost, how long it will take, and how can we ensure return on investment?
A good cloud software provider will know what you need now to address these fundamentals, with clear pricing and implementation timelines, and a clear view of how your cloud will provide maximum value to your business.
Their product roadmap is just as important, however. They should already be working on what your customers will need in future and what development you will require as your cloud and the market matures. Find out what features are planned, including specific integrations, hypervisor support, available virtual machine templates, or support for a range of hardware or billing models. Identify which will add commercial value to your business, and challenge your provider to do the same.
Key takeaway: as with any critical business application, make sure you choose a provider with real understanding and experience of what businesses need to successfully deploy a public or private cloud.
2. True cloud, not just virtualisation
Whilst most data centres are already virtualised, at least in part, this is not the same as cloud computing.
There are lots of different products that claim to be able to “cloud enable” your infrastructure. Not all clouds are created equal however. One example of this is high availability and failover: if you don’t have failover, you don’t really have a cloud – it’s a fundamental characteristic.
The type of cloud capability also varies greatly between software platforms. Check if it supports public, private, hybrid or VPS cloud, or all of them. Other key areas of differentiation are set-up time, load balancing, auto scaling, automated billing, secure multi-tenancy, firewalling, hypervisor support and speed of provisioning for virtual machines.
Look for as much automated capability as possible for the kind of cloud hosting you plan to deliver, to avoid costly human input for provisioning and management.
Key takeaway: not all cloud products are created equal. Insist on functionality that gives you maximum automation and efficiency. Above all, find a product that gives you the flexibility to adapt to the market as it evolves.
3. Time to “cloud-readiness”
Now is the time to bring public cloud services to market for hosts. Today, cloud hosting margins are as high as 50% with low customer acquisition costs. A land grab is underway fuelled by customer demand; rapid deployment of public cloud services is not just a “nice to have”. Even six months from now the market landscape will be dramatically more competitive. Similarly, demand for private cloud infrastructure within the enterprise is accelerating.
Guaranteed speed of deployment to cloud-readiness from your chosen cloud software provider is essential. They need to show evidence of a bug-free platform, proven deployment and examples of the software in action. This is where expertise in real-world cloud deployment, and your industry, really starts to count. Cloud software providers with a tailored product have built in understanding of what it takes to deliver IT services on-demand: user and billing management, skills in server, network and SAN set-up, security, and integration with major technology vendors. Aim for a complete implementation within a couple of weeks; in six months you could be too late.
Key takeaway: a lengthy cloud deployment increases the risk of late entry into the public cloud hosting market where early adopters have become established leaders. For enterprises, slow cloud transition risks both the need for fresh capital expenditure on new hardware and operational inefficiency of the existing infrastructure. Measure your deployment in days, not months.
Every planned transition to cloud computing will be carefully scrutinised in terms of cost.
Launching a public cloud business for hosts requires complete transparency from your cloud software provider. Similarly, going live with an internal cloud for the enterprise demands clear cost structures to accurately budget for the lifetime of the project. Examine published prices carefully and ensure financial modelling on an operational basis fits with business requirements.
The ideal scenario is to run a proof of concept with the cloud software provider before committing. There may be a small investment required to build a working cloud and test the software, but you should not need to invest huge sums up front: best-in-class functionality can be had without much investment, or on a free trial basis.
Get the basics right by understanding what investment is required from you up front, the ongoing pricing model and contractual terms. Then, keep a keen eye for any hidden costs in set-up, deployment and integration. SAN set-up can often be an additional hidden cost, and support also appears as an extra with some providers.
Key takeaway: the cloud management software market is increasingly competitive, and best-in-class functionality can be yours for little or no up-front investment. You need to have a very good reason to insist on software with monolithic licensing and prices for integration, implementation and support.
Billing flexibility is critical in any cloud computing environment. In public clouds, billing flexibility drives profitability for the host, and at the same time gives cloud customers more choice. The more things a host can measure and charge for – CPU, RAM, storage space and so on – the better they will be able to design cloud services with the features and pricing different kinds of customer require, at a price point that makes sense. In private clouds, billing flexibility is more about efficient use of resources, by enabling accurate measurement of resource usage, and financial transparency for reporting, cross-charging or cost allocation purposes.
In each scenario, your cloud management software’s billing functionality is critical. The minimum standard should include options to bill for CPU cores and share; disk space, both as primary and backup; RAM; IPs and bandwidth; and IOPS. Billing ability for multiple tiers of resource and the ability to charge for idle or reserved resources is also advantageous. Finally, check the terms of utility billing. Not every customer wants to pay as they go, so look for a cloud management product that offers daily, weekly and monthly models, or a mixture.
Integration is also important here. Whatever type of cloud is being deployed, it’s important to understand the cloud software’s compatibility with existing billing engines or internal charging systems. Check which standard integrations have already been built, and who will pay for any integration work required.
Key takeaway: don’t assume that moving to the cloud means you have to adopt new billing platforms and utility billing models and a small set of billing options. Billing flexibility will be critical to the success of your cloud project.
6. Hardware compatibility
Whether a cloud is built on high-end racks or much less impressive hardware, the resources provisioned to customers and users are the same: a CPU resource is just a CPU resource. In spite of this, however, it is important to consider what hardware a cloud management product enables you to build your cloud on.
Different approaches to cloud deployment and virtualisation have different kinds of hardware requirements: some insist on homogenous hypervisor servers, for example, while other products can create clouds from a mix of hardware. Enterprise data centres tend to house a set of best-of-breed vendors, and are more likely to offer a standard hardware platform for cloud creation. Some hosting providers will have comparable hardware: others will be looking for ways to utilise older servers.
Both public and private cloud operators, however, will benefit from cloud enabling products that support the broadest possible mix of hardware for hypervisor servers and SANs. Not only does this provide more flexibility in the way a cloud is architected, it provides the potential for more efficient use of hardware in the future. Hardware that is due for a refresh may find a new lease of life provisioning CPU and RAM resources in the cloud.
Key takeaway: compatibility in server and storage hardware is a critical success factor for cloud projects. Focus on platforms that support the widest range of hardware types and performance levels; that enable you to re-use your existing servers and SANs; and providers that can help with any hardware investment you need to make.
7. IOPS monitoring
IOPS monitoring (disk Inputs/Outputs Per Second) is central to cloud efficiency and performance. Most cloud providers – public or private, have the ability to assign chunks of CPU and storage resource to a customer or internal user, and of course measure and if required bill for those resources. Being able to do the same for IOPS, however, is much rarer.
IOPS monitoring is important because of the dependence cloud computing has on storage performance: a single virtual machine may consume minimal CPU and RAM, but generate intensive disk access and reduce disk performance for other users.
In public clouds this is easy to mitigate by charging more for heavy IOPS users. In private clouds, there may be interdepartmental charges involved, and there will certainly be value in simply understanding how IOPS performance impacts overall cloud performance. If a cloud platform doesn’t handle IOPS, either find one that does or accept the potential for lower profitability and performance.
Tiered storage in the cloud is another consideration for distributing IOPS and maintaining performance levels. Multiple tiers of storage from multiple sources, and multiple disks per virtual machine, can go a long way to alleviating potential IOPS bottlenecks – and therefore, performance problems for customers. This is also a key feature to look for in cloud management software.
Key takeaway: unless you have good reason, avoid cloud management systems that cannot monitor IOPS or that don’t offer flexible tiered storage and swap disks.
8. User permissions
Another area to focus on is how well a cloud management product handles user permissions and limits. Granularity is the secret here. The more control a cloud platform provides over what users are allowed to do, the better it will be at delivering services tailored to the precise needs of those users.
As a minimum, hosts need a simple way of controlling permissions and limits to create all of the typical user types – standard users, VIP or power users, resellers, and their own L1/L2/L3 support teams. Beyond this, look for cloud management software that will enable you to offer premium and add-on services by enabling you to set up different user types with access to specific management features – higher-performing storage, remote access to virtual machines, advanced backup options, and so on.
In the enterprise, IT leaders need to balance performance to the user with resource efficiency. The ability to categorise users, and control their profiles and permissions, allows private cloud operators to optimise sharing and allocation of resources to users across the business.
Key takeaway: look for cloud software that gives you granular control of user limits and permissions, with an API that lets you exploit that control to create exactly the cloud service your customers need.
Self-service is a fundamental characteristic of the cloud. The cloud’s back-end may be complex, but the front-end UI should be simple enough for anyone to use it.
In public clouds the front-end has a direct impact on the quality of the customer experience, with a corresponding impact on sales. It also affects productivity for system administrators, support and billing teams working on the cloud behind the scenes. In private clouds the same is true: user experience and productivity depend on the quality of the interface, and just like a hosting business, the UI should meet the needs of a broad range of users with different experience levels.
As a result, you should spend time testing the UI from the point of view of different users, and investigate the extent of customisation, localisation and branding that the software provider can enable. Integration is another important consideration, particularly the ability to deploy an alternative customer portal if needed – rather than the provider’s UI, you may prefer to use an existing control panel or develop your own front end to the service.
Key takeaway: focus on the UI from an internal as well as a customers’ viewpoint, and favour cloud platforms that enable you to customise the user experience easily – either directly, or through the API.
The promise of the cloud is that it is “just there” – an always-on, instant access compute and storage resource. In the enterprise, cloud users expect their IT “to just work”, and in the public cloud, customers are also buying into the uptime benefits of the cloud.
With automatic failover, cloud users ought not to experience any service disruption. However, the reality of any IT system is that it will fail at some point, and when that happens you will need rapid, high quality support from your cloud software provider.
The key questions to ask your provider are whether support is free or an additional cost, the SLA or response time that can be expected, skill of support team, the hours they keep and whether 24×7 telephone, live and email support are in place. If things do go wrong you will need SLAs with response times measured in minutes, not hours.
Key takeaway: since this is the platform on which you’re running your cloud business, insist on free, high quality, 24×7 support from your cloud software provider.
By Ditlev Bredahl