In-Depth

Networking for the Cloud: 3 Vendor Views (Part 3, Mirantis)

Three vendors, three opinions on what drives their solutions. In this second of a series, Doug Barney talks with Lloyd Bloom, senior product manager of CloudSleuth at Compuware Corp.

• By Doug Barney
• 02/27/2012

We contacted three major networking vendors to talk about bandwidth and the cloud. In the final part of this series, Mirantis executives explain the role of security.

Finally, three executives from Mirantis Inc. -- Boris Renski, executive vice president and co-founder; Alexey Abashkin, development manager, Cloud Computing Center of Expertise; and Oleg Gelbukh, IT engineer, Cloud Computing Center of Expertise -- handled our queries. Mirantis uses open source software to create cloud platforms for major clients.

Redmond: What are the most important networking/WAN issues IT has to think about when considering the cloud?
Alexey Abashkin: There are three very important issues: security, high availability and accessibility.

Usually, high availability is not very difficult to achieve and generally depends on WAN hardware costs. The main problem the IT engineer must solve is balancing security and accessibility. It's no secret that the most secure service is the one that's not connected to any network.

How does one evaluate the network/WAN's readiness for the cloud?
Boris Renski: In my opinion, this question is too broad to be answered with any degree of specificity. When you refer to network readiness for the cloud, it can imply an infinite number of scenarios, ranging from allowing your internal corporate users to access external resources -- browse the Internet -- all the way to utilizing hybrid cloud models and having your internal application workloads dynamically migrate to an external cloud provider.

Generally all networks -- and, moreover, the Internet itself -- are clouds, which means that it's not a question of network/WAN readiness for the cloud, but rather the ability of the cloud to utilize network in the most efficient way.

How do these issues differ depending on application?
Renski: Generally, it again all boils down to security and accessibility. Security is the paramount concern. The more connected your WAN to the cloud, the harder it becomes to keep it secure. If you simply use cloud applications, such as SaaS solutions, you probably don't have to do much beyond standard firewall configurations. If part of your infrastructure is in the cloud, such as with Google Apps and external e-mail hosting, additional scrutiny is required. Now, if you go as far as dynamically bursting internal workloads to a public provider, your WAN effectively becomes one with the cloud ... this is where security gets very challenging.

There's no generic advice for securing a network. All networks are different; all have their own specific characteristics, such as the primary technology stack, size, business direction and so on. Securing a network will depend greatly on all of these variables.

How do these issues differ depending on size of the enterprise?
Renski: The common belief is that the larger the enterprise, the more complex the network and associated issues that have to be managed. At Mirantis, our experience with our customers has been a bit different. There's some correlation, but it's not linear. The most important variable is not the size, but uniformity of the network. Large networks frequently and commonly consist of many silos that are a byproduct of their haphazard evolution. Various divisions use different technology stacks and different policies. Silos drive complexity, not the size of the enterprise.

What are the basic steps one can take to get the network ready for the cloud?
Abashkin: As we've discussed, the network should be highly available, accessible and secure. Let's call these three terms the "definition of done."

What are the key questions to ask and demand of your network provider?
Abashkin: In my opinion, after the questions about availability, accessibility and security are answered, the next question should be: "Does your network support IPv6?" This is important to any business starting to work with a network provider; no one wants to deal with the potential disruptions that could occur when the pool of unutilized IPv4 addresses is depleted and unprepared providers have to quickly switch to IPv6.

Which WANs are ready for the cloud and which aren't?
Abashkin: All WANs are ready; the question comes down to what one means by "the cloud."

How much are bandwidth issues holding back cloud adoption?
Oleg Gelbukh: Basically, the internal bandwidth in cloud datacenters isn't an issue, as a network can be faster than disk systems today. Between the datacenters is the gray area of WAN/Internet, and here everything depends on the design of the cloud application in question. Smart and flexible design is essential for cloud applications to gain all the benefits of cloud architecture and avoid getting caught in the limited bandwidth trap. What's holding back cloud adoption most isn't bandwidth issues, but application-design issues.

Does the cost of additional bandwidth and the ever-lower cost of on-premises servers and storage still make the cloud an economic advantage?
Renski: No. The main economic advantage of clouds is resource sharing between projects or customers. Cloud is not about cost, it's about agility. If you're a payroll provider and have burst workloads at the end of every payroll cycle, you want to burst them to the cloud. If you're a startup growing at 100 percent a month, you simply plan your internal infrastructure investments at this growth velocity and have to resort to cloud. In all cases, you'll end up paying more, but it's worth the investment.

What will change in the future to mitigate some of these issues?
Abashkin: I expect there will be no terms like bandwidth or cost. Our children will just use cloud services as they are, because we're now taking the steps to simplify the technology.

Can the Internet and network providers keep pace with a possible explosion in cloud demand?
Gelbukh: Again, it's hard to tell if network providers can keep up with exploding demands. We absolutely should not depend on whether they can or not as we build applications in clouds. Although network providers seem to do their best, we still should remember that even the fastest network can give up if we put the entire burden on it, instead of being smart and creative in our application design.