Network resiliancy and redundancy is imperitive in your organization if cloud computing is going to work. In addition, co-ordination with others in the data centre is essential
A year ago, Jonathan Day-Reiner decided it was time to change the way his company delivered its information technology services and step up to the cloud.
The director of IT operations at Toronto’s 80/20 Solutions, an online interactive marketing company, knew the firm’s co-located servers were due to be replaced. And co-location, where he paid by the amount of power consumed not the space used, wasn’t a bargain.
“Blades use an incredible amount of power in a really small space, and that can kill you on the billing side,” he says.
But the company also has varying compute needs – sometimes a customer has a major marketing effort and needs more cycles for a short time.
So Day-Reiner turned to an infrastructure-as-a-service provider, CentriLogic Inc., for an answer. There he contracts for “instances” of computing power a month based on a total amount of memory – in short, a virtual sever. In his case a standard instance is 8 Gigabytes of RAM and two processors, and 80/20 Solutions normally has six instances of its core applications running.
“You can spin up and down more instances as required,” he says. “We can just go into a Web interface, hit a button and in five minutes there’ll be another server there with an IP address with our app and everything we need installed on it.”
This is the new future of cloud computing, and network managers better be prepared.
The thought of cloud computing may give them the shivers, if only because the term is so ill-defined.
But arguably for the burden will fall on them: Without an adequately prepared network, cloud computing – whether software or infrastructure as a service, public or private – can’t work.
“It’s critical that resiliency and redundancy are baked in to those Internet connections because now core business functions are in that cloud,” warns James McClosky, a senior analyst at London, Ont.-based Info-Tech Research.
That resiliency is crucial in part because service level agreements don’t cover everything. It does no good, McClosky argues, if your cloud provider’s data centre has five-9s reliability when the Internet connection is down.
That sound like cloud means more work for network administrators, but experts say there will be trade-offs. If done right, says McClosky, the cloud should be a net win.
For example, cloud means additional security risks. On the other hand, McClosky says, there are maturing security-as-a-service offerings so some of that burden can be offloaded. In return, administrators can focus on other security areas instead of network security.
Canadian organizations are familiar with software-as-a-service (Saas), but infrastructure-as-a-service (IaaS) will make big waves here this year.
In January, IBM Corp. launched a cloud delivery centre in a Toronto suburb, letting organizations buy computing resources on a pay as you go model. According to a just-released report by IDC Canada, similar offerings from BCE Inc’s Bell Canada and Telus Corp. and Hewlett-Packard Co. are on their way.
The report, by analyst Mark Schrutt, says at the moment cloud computing here is “hype and trendy.” But it notes that Canadian service providers are currently very active in designing and piloting private cloud environments for dozens of firms.
In 2009, roughly 3 per cent of a company’s IT budget was dedicated to cloud activities, says the report. This year will see this increase to over five per cent. By 2014 Canadian organizations will spend more than $1 billion on cloud services, with growth fastest among enterprises.
Ted Ritter, a senior analyst at Nemertes Research who specializes in cloud and data centres, says his research suggests that while North American companies are considering public clouds they are more interested now in private clouds. He defines a private cloud as a pooled resource that can be allocated on demand and can be provisioned by staff.
For network admins, private clouds, with their “huge” Layer 2 virtualized switched networks, can be a challenge. For one thing, with few exceptions virtual and physical switches have to be managed through separate consoles, he says.
The virtualization that makes private clouds work can also cause confusion, because the door is open to people tampering with the network who wouldn’t otherwise do so.
“If you’re not careful you can wind up with a developer deciding to load a new virtual instance, configure virtual network access and away they go. And that completely bypasses any procedural controls you have,” says Ritter.
Equipment makers are working towards single pane network management solutions, with Cisco System Inc.’s Nexus 1000V switch for VMware being one of the first. Juniper Networks Inc.’s recently announced QFabric architecture will include it, but won’t be complete until the end of the year.
Until this is worked out, he says, cross-training the network and server teams is the answer so each understands the implications of what the other does. “It’s more of a process, people and training issue than a technology issue,” Ritter maintains. “There are tools that will help facilitate this, but we find there’s a more fundamental issue.”
Among those looking at IaaS is Scot Adams, senior vice-president and CIO of Toronto-based real estate developer Cadillac-Fairview Corp. Ltd., who says the flexible CPU cycles could make sense for application testing.
“It doesn’t make sense to set up the infrastructure [in our data centre] to test all the different possible platforms that it will run on,” he explains.
CF’s cloud experience started some eight years ago when it outsourced its payroll system. About three years ago it began looking for non-strategic applications that could follow. The most recent is the company’s shift from on-premise Lotus Notes to Google Apps for email. By the end of the year all 1,800 of CF’s full-time staff will have made the move.
A small test of Salesforce.com for customer relationship management – to replace a Lotus Notes-based application — has just started.
On the other hand, he hesitates about moving strategic applications, such as CF’s decision support tools.”
Would he move all or most of his IT infrastructure to the cloud? “It it was better, faster and cheaper we might,” he says, “but at this point it’s definitely not cheaper.”
Ted Maulucci, CIO of Tridel Group of companies, which builds and manages condominiums and commercial properties in the greater Toronto area, said his company has had mixed results in the cloud.
In 2004 the company switched from internally overseeing email to the hosted email service of California-based Postini (now owned by Google Inc.). The on premise antispam devices Tridel used “took a lot of internal management, lot of hassle” in setting up rules, he explained. “Something like [outsourced] anitspam we just found was a no-brainer. You put it out there, they had a team of people working on it. There’s no why you could match what they were delivering for what you were paying.”
In addition, local spam filtering meant as spam went up so did bandwidth. Filtering in the cloud means only legitimate mail gets through. And if Tridel’s Internet connection goes down, the service stores incoming mail for later delivery.
In addition, a service provider manges Tridel’s corporate Web site, which runs on a cloud instance of Apache server. Largely a Microsoft shop, Maulucci felt hiring an Apache expert and dedicating a server wasn’t cost-effective, particularly for a mission-critical application that get significant Web traffic.
Still, a four-year test of outsourcing Tridel’s entire email system ended in 2009 when Maulucci brought it back in-house. Initially, outsourcing started when the company needed to upgrade its Microsoft Exchange environment. Maulucci calculated that while the move wouldn’t save money, certain administrative headaches would go away.
Instead they built up. “When you do this your bandwidth starts to go up, so you end up with increased bandwidth charges,” he warns. “So when you negotiate your contract you’ve go tot cover those things.”
In addition, email is one of those applications that still requires internal manangement. Because the provider is being paid for storage, staff have to be encouraged to keep control of their mailboxes, and the mailboxes of departing staff have to be closed quickly – lessons he says he learned the hard way.
In addition, the cloud puts you increasingly at the mercy of providers.
“Nobody assumes a liability,” he complained. Tridel was subject to the provider’s service level agreement, which was ‘I will deal with that [problem] in 24 hours.’ That’s difficult to explain to a senior executive, says Maulucci.
So email was brought back in-house. “You need to be careful. You need the right partner,” he advises, “and I think we didn’t get the right partner on that first round.”
Two years later, though, the lure of the cloud has him thinking about it again.
“I don’t think network managers need to panic too much” about cloud computing, says Day-Reiner.
“It’s a different way of doing things. But you still need core competencies to ensure you have a successful system.”