Validating IaaS Providers

In the last two weeks I was involved in a 3 or 4 events on Cloud Computing. It is a subject that still generates much discussion. It’s inevitable, because cloud is a revolution in the way we deliver and consume IT. If this new way to see IT was understood immediately, there would be a revolution.

Today we will talk on : “How do I choose good cloud providers of cloud computing infrastructure, IaaS?“.

It’s an interesting point. Although there are several providers of IaaS, in some other nations and global, despite the similarities when looking at its marketing brochure, they are not created equally and the details show is totally different.

Each IaaS provider is designed and architected to meet market scales and different characteristics. Thus, some propose to serve large enterprises, highly demanding in terms of levels of service and compliance to regulatory issues, while others seek to serve small businesses, less demanding on these aspects. Their data centers are also designed with this scale in mind. To answer a few hundred customers, a data center will have characteristics different from others that should withstand a few million customers. Business objectives are also different, which will be reflected in investment and financial capacity of each provider.

The scale of the provider has an immediate impact on the flexibility and price offered to the market. The provider must first invest in a given computing capacity, capacity that will be offered to the market. That is, it must make an upfront investment to enter the market. If your customers have an average ticket too small, it will need much breath financially subsidize the business until their offerings for IaaS are profitable. Moreover, their range also influences the degree of elasticity that it can provide. For example, if your customers do not vary much in terms of computational and consumption, are near the limits of use of your data center, any increase in demand will require more investment and speed of service will take longer. In these cases it probably cannot guarantee that 100% of the funds accrued will actually allocated. On the other hand, providers with excess capacity and customers with significant variation in load, may offer different prices according to demand.

Data centers providers also deserve special attention

Some global companies are able to create multiple data centers with highly sophisticated security controls. Others, without this capacity investment, may have a data center more vulnerable to attack or even downtime situations.

The servers of these data centers are not the same. In general, a IaaS service is based on Intel-based servers and AMD with Linux and / or Windows. Does this issue concerns us? However, depending on the processor, it may or may not have virtualization capabilities built into the hardware, which improves the performance of virtual servers. Also make sure that the servers and clusters that make up the data center has redundant components, such as dual power supplies “or” dual network interface cards (NICs). The level of availability offered by the provider is affected by these characteristics.

Another technology being observed is virtualization. In general we find hypervisor VMware, Xen, Hyper-V and KVM. If you’re worried about performance with a more detailed study will show which hypervisor is best for the level of expected performance and availability for your applications.

Depending on the virtualization technology and expertise of the provider, you may have different levels of availability. The worst thing is the crash of a physical server dropping all virtual servers that run on it. From there, it will be interesting to analyze how the provider will ensure availability: it can automatically remove the vitual server from one physical server unavailable to another, without affecting the operation of the users. And what is the cost of this level of availability?

As we talk about availability, we come to the point of disaster recovery. If the data center become unavailable, is there an alternative data center to continue operation of the provider? If so, how long the alternate data center can start operating?

Another point is the degree of automation and speed of the service provider provisioning user requests. If the provider’s provisioning process still require manual processes, the service may take several hours. Already a provider that offers automatic resource provisioning and self-service interfaces can meet the demands within minutes.

An important aspect is the SLA (Service Level Agreement). In selecting the provider considers the desirable levels of service and select only those who can, can be proven to offer such agreements. Important to place in contract penalties for breaches of these agreements. Consider that the level of a hosted application service provider in public IaaS traverses the network, be it Internet (for which has not always control) or a private network, where the deal must be closed with the cloud providers and network. The agreements ensure that there is something that is referring to DoS (Denial-of-service), as there is the ever present possibility of a public cloud provider to be attacked by hackers. After all, it  concentrates on it’s data center and hundreds of different companies.

So … what is the best provider?

Depends on the needs of each company. Some require a level of demand that the limit selecting a provider of global scale. Some do not demand such requirements and can live with a provider that does not provide very high levels of availability and flexibility.

ESDS