‘Companies embrace the cloud to focus on core competencies’
 Cloud computing and virtualization are a boon to enterprises, but they also introduce a new set of challenges. Richard Clifton, Senior VP and GM, Technology Management & Solutions Organization, NetApp, tells Varun Aggarwal how cloud computing has evolved and how NetApp tackles cloud issues such as isolation on shared infrastructure
How has the concept of ‘cloud’ evolved over the years and how has the market kept up?
If you go back a few years, what you call cloud computing now was referred to by many people as ‘grid computing.’ If you look at some of our old products, you’ll see we’ve been using the grid for a long time. Now the market has moved interestingly as server virtualization has made it possible to have a greater level of resource sharing in server farms. We had a major push for storage for virtualized servers. We have partnerships with VMware and Microsoft. We have a lot of specific features for enabling server virtualization. Shared storage is what’s enabling cloud for enterprises together with the products that serve multiple users and multiple user groups.
What are the drivers for cloud computing globally?
Most companies in emerging economies want to focus on their core business. So they look at whether they really want to do it themselves, or let an expert do it for them—getting the same service levels at a lower price. The evolution of high-speed networks and virtualization has made it more possible to trust the provider in taking your key IT infrastructure into the cloud.
Now cloud may seem new, but as you look at payroll processes in the last 20 - 25 years, companies have taken their payroll processes outside the company because expertise in all the tax laws of the company, was not a competency they wanted to build. That’s the piece of external computing that today is increasingly in the cloud. What we’re seeing is that IT companies look at other pieces of processes and say, “Do I really want to be an expert on this? Am I doing something which is fairly a standard in the industry where someone else is an expert on that packaging of the process and could add value to me?”
Things like payroll are fairly simple to take out of the company, but if you look at some of the custom software packages for the workflow of the core business, I think you’d like to deploy it yourself. Cloud computing is most likely to pick up in processes that have standardized expectations.
A major issue with cloud computing is compliance, especially archiving using storage virtualization. How do you tackle this?
The one thing I hear from customers when they think about cloud and shared infrastructure is that they want to create the highest level of sharing in order to derive maximum efficiency. This sharing comes from hosting sets—which could have Coke and Pepsi on the same machine. Now the question is: “How do I create a level of isolation where both Coke and Pepsi are comfortable?” In our storage devices, we have the ability in one device, to show itself as many devices.
So, we’ve never had an isolation issue with all kinds of compliance testing and deployments in a wide variety of hosting environments. This is the same case with server virtualization, where one has a single physical server and multiple virtual machines. You can have one physical storage device that can present itself as many storage devices—this feature is called ‘MultiStore.’
An enterprise would have different departments but each of them would like to feel they have their own infrastructure. They are OK if it’s virtual, as long as they get the performance, the right service levels, and they trust the level of isolation.
We have many partners in the hosting space who have customers taking their core processes out on the cloud and they have thousands of such deployments.
Bandwidth is still an issue in India, both in terms of the cost of bandwidth as well as reliability of connectivity. How do you circumvent this bottleneck?
It depends on which city and how local the interactions are, and what kind of workloads are going up and down. If I look at enabling call center packages where all my work is going straight on to the wire bandwidth, that might be difficult to make it work.
We’re going to see more cloud services around DR and archival since regardless of the cost of bandwidth, distance is required. If you take your data to another facility or to a cloud provider, you’ll have to pay for the bandwidth anyway. So you can just write the network out of consideration in areas where distance is a necessary part of it. I think DR and archival would have cloud deployments at a faster pace.
There were a lot of discussions around file area networks (FANs) about two years ago, but the concept just disappeared. Even NetApp used to talk about it. What happened to that technology?
Customers look for two things: first they need a deeply integrated capability for multiple storage systems to present themselves as one—that’s at one end of the spectrum. At the other end of the spectrum, customers want an accelerated outcome against the cloud system over a distance, so they want to have a cache and that product is called ‘Flex Cache.’
The earlier things that you were seeing around FANs were neither as integrated nor as suited for distributed topology. Today’s technologies are so integrated that people stopped talking about FANs.
The FAN that Brocade offered at that time, was a set of agents that created a pseudo file system across a set of servers and hosts. However, having a cache that is not in the host but in the network makes it completely automated and you wouldn’t even know that you’re using it. It’s the transparency that you have in the system. The technologies of that time period were interesting as they could solve some of the problems but at the same time everyone had to learn about it, manage it and what people wanted was the next level of automation.
FAN was the term Brocade took to market and it wasn’t very widely accepted; we were partners with them around that product so we talked about it.
Latency could be a big issue with customers that want to share data with their offices spread across geographies using distributed topology. So in order to solve this problem, one of the most interesting developments we’ve made is the ability to put solid state accelerators into the systems so that the multi-performance could be much higher than the standard storage. What we’re doing is serving the read locally, so bandwidth won’t be an issue because we’re using a cache large enough to take care of the local volumes, and since we’ve made the read so fast, the write would have longer latency and the workload becomes normal.
What kind of investments are you making into cloud computing?
We’re not going into the cloud business by building our own data centers but we enable our partners to build cloud infrastructure for their customers.
We look at shared storage infrastructure as our primary market target and that’s the key building block of the cloud. It’s also a key building block of the next-generation data center. The ability to share many things in the same device is an enabler even in the SMB environment. That’s the fundamental technology where we’ve focused 95 percent of our engineering efforts—enabling shared infrastructure for a variety of customers.
The NetApp Dynamic data center has already been launched and we have a whole range of other products for data management which are specific enablers for the cloud.
What are the latest developments that you’ve made with the launch of ONTAP8 around cloud computing?
While we recently announced ONTAP8—which has some next-generation pieces of cloud, our cloud initiatives are much wider than that. If you compare our position and initiatives in cloud to our competitors, you’ll see one major difference—we already have a large installed customer base that is using the cloud over very large-scale deployments. The next-generation features have to do with automation and integration.
|