Tier1 Research: Is Cloud the New Colo?

Antonio Piraino, Tier1 Research

Dallas-based Horizon Data Center Solutions, more commonly known for its colocation business, has taken the leap into the cloud space, launching its FlexSafe Cloud service. Perhaps 'leap' is too large a word for a company that has been upping its managed services ever since the acquisition of Mareechi – an ASP – in April. T1R has seen colocation companies add managed services, partly as an answer to increased demand from enterprises, but also to increase ARPU with services on top of straight colocation, generating more revenue per cabinet. But very few have managed the transition, putting Horizon's FlexSafe in the company of services like those offered form StrataScale or even Terremark, Carpathia Hosting and Savvis' enterprise cloud divisions. The difference for Horizon is that it is targeting the mid-sized enterprise and federal sector that has traditionally looked to spend millions of dollars on colocation outfits and are now expected to be sucked in by Horizon's ability to create customized solutions to these complex environments.

Horizon intends to position itself somewhere between the colocation providers, the managed hosters and the large managed outsourcing providers like CSC and IBM. It won't be as expensive nor as complete as the latter, but the firm believes that it has the professional services unit and complex management expertise in-house that many of the cookie-cutter hosters lack. It also believes that its colocation business will help it succeed in the cloud space for a number of reasons. Take, for example, the fact that the company is launching the cloud from one of its Dallas datacenters that is being leased on a wholesale basis from Digital Realty Trust, which permits for rapid expansion for its colocation and cloud business with less substantial capex requirements than other builders of the cloud. The company also has recently established a strong relationship with AboveNet for a private ring connecting both of its Dallas datacenters. AboveNet's presence, and the addition of more carriers in general, has benefits all around in the way of route diversity, redundancy and scalable gigabit-level bandwidth for Horizon and its customers.

Is that enough to make a colocation provider a cloud provider? No. The company has clearly spent some money on hardware from Hewlett-Packard in the form of blade servers and ESX licenses from VMware for virtualizing the entire cloud environment. 3PAR is the company entrusted to deliver the SAN storage for the cloud, with its virtual domain capability. Cisco's firewalling (and other IDS security practices to make FlexSafe feel safe) and F5 Networks' load balancers complete the infrastructure mix, allowing for very granular load balancing, redundancy and high availability based on physical and virtual infrastructure. Horizon currently has two Dallas-Ft. Worth area data centers, connected by the aforementioned private fiber ring, providing fast, reliable data backup and replication solutions. And this is what Horizon believes will give it the foundation to go after the federal sector and mid-tier enterprises that require a great deal of utility storage, computing power and networking bandwidth – at a cost that would be unappealing to the average SMB customer.

So what makes this infrastructure-as-a-service (IaaS) offering any different from the others? The in-house expertise brought by Mareechi and Horizon. The company has supported complex hybrid ecosystems, and has deployed and managed SAGE product line applications, and other CRM, ERP and SQL deployments. The company doesn't want to become an IBM or EDS, but it also doesn't want to offer a public cloud facility for credit card-holding ISVs. It believes that its cloud product will really be a series of virtual private clouds for its mid-tier (i.e., $300 to 500m) clientele that requires bespoke systems with customized SLAs – that currently revolve around high availability (using physical and virtual technologies to do so).

The Horizon FlexSafe Cloud is processor based – with multiple-sized processors being the initial option, and not fixed menus available through a Web portal. Scalability is not truly automated and orchestration is still relatively manual. The company will monitor workloads, each of which can have built in utilization thresholds that trigger alerts for making scale up or down decisions. The company feels that the nature of its clientele is such that they will make these decisions together with Horizon, just as they would around the addition of a new cabinet or power unit in the colocation environment. Those alarms will be set around hard disks, processors, memory and network utilization. Although bandwidth is a separate purchase for VPN purposes between off-premise locations and the Dallas datacenters – DR can be controlled and internal VPNs are controlled by Horizon. In addition, depending on the contract, the company will fire up a new image if necessary on behalf of customer.

Horizon wisely stays away from trying to prove just how cloudy its IaaS service is. Yes, it has some elements that T1R considers necessary to be a cloud, including some level of automation, management, security and storage. No, it does not speak to APIs, nor fancy Web interfaces and automated levels of initial deployment or even multi-tenancy. Yet it has the ability to scale up and down, and there is a large amount of automated workflow, and things are virtualized and dynamically changeable. More importantly, Horizon has shrewdly defined itself by the National Institute of Standards and Technology (NIST) working definition, with its key characteristics of on-demand self-service, ubiquitous network access, ability to rapidly scale resources and pay on a measured-usage basis. The relevance to this is that NIST has been instrumental in giving the government a foundation by which to have a universal definition of the way in which it looks at technologies, making the federal sector feel comfortable enough to embrace it. It includes security compliance measures and other technical guidance, using language that is fast becoming commonplace across the federal sector. Incidentally, the working definition of cloud computing is now is in version 15.

When Terremark launched its Enterprise service in Europe, Joost Metten referred to it tongue in cheek as colo 2.0. The reason being as colocation became increasingly expensive on the continent, and certain enterprises continued to hold off on its premium cousin – the fully managed service, so cloud services introduced a kind of capex to opex solution similar to dedicated hosting, with less of a commitment and a greater number of features. Interestingly, Horizon has seen its new cloud services being used as part of a migration strategy by its colocation customers. The service has not been around long enough to find out whether it has stickiness with these deployments, but the hope is that as these enterprises continue making use of the networking and virtualization capabilities, eventually they will retain full production environments on that cloud infrastructure. Plus, the typical colocation company that Horizon goes after is more concerned with compliance and scalability than true variability and spikiness of traffic – meaning the cloud revenue could become quite substantial at the level of deployment of these customers.

But Horizon makes mention of the fact that rarely do its colocation customers eventually not make use of its professional services or fully managed services. That means that high-touch services are sought after and the premiums are there to be had from low-hanging fruit. In some ways, this contradicts the need for self-managed cloud services like FlexSafe. Or so it seems. And herein lies one of the differences between the enterprise-focused cloud service and the commoditized public cloud services – that enterprise cloud is really a continuation of a very complex environment that may contain any quantity of on-premise self managed infrastructure, colocation, managed hosting and IaaS services – a combination of high-touch and low-touch services. T1R expects to see greater evolution of these kinds of hybrid environments, with SLAs becoming increasingly granular and helping to shape the index of pricing associated with each service and performance guarantee.