VAR's Guide To Pros And Cons Of Migrating Data To The Cloud4:00 PM EST Tue. Feb. 07, 2012
Moving a customer's data to the cloud is not an easy decision, nor is the process of doing so. Here are the pros and cons highlighted by Brian Day, director of cloud solutions and application strategy at Logicalis, a Farmington Hills, Mich.-based VAR with a multimillion-dollar cloud business.
It's simple: The cloud gives more people more access to more data from more places. And nine times out of 10, more is good. As enterprises increasingly allow customers to use their own devices, the cloud is a perfect way to tie together disparate devices without having to support them all independently, Day said.
"If you have data behind a corporate firewall, it's not easy to get to. Using a PaaS or SaaS solution makes it more available to devices," he said.
But if your data's in the cloud that means anyone can get to it, right? That's a common concern customers have, Day said. The way to approach this is to take smaller data sets into the cloud in the first place.
"Take your Web site and enable it with customer self-service stuff. You probably don't need everything in the cloud. Maybe name and address, but not Social Security numbers," he said. "Second, make sure you have it secured in a proper manner."
This is where choosing a proper service provider is important, Day said. For example, some large public cloud providers have only implemented firewalls in the past year or so, he said.
"Even then that's what you get: one firewall. We see this as a big opportunity. We can do whatever [customers] want. If they want multiple layers, we can protect those," Day said.
Moving data to the cloud gives customers access to high-end platforms, often at a lower cost than buying infrastructure. This also allows developers to focus on value-added work and not on setting up new environments, Day said.
"The old way of thinking was you need to step back and ask what's a peak load and how big a server do I need for an application, even before you're finished with the functionality of the application," Day said. "You can get away from that with cloud. You no longer have to size for peak loads. The cloud has the additional capability to expand elastically."
Customers have concerns that their IT performance will slow down by moving data and applications to the cloud, Day said. Virtual desktops help avoid any potential latency issues, he said.
"Our generic recommendation is to put the apps near the data. You don't want a client/server application with data in the cloud with the client by the desktop. Those are an important piece for us as we start to migrate clients to the cloud," Day said.
In addition, orchestration and automation tools can help solve latency issues, Day said. "It's not hard to take an existing application and put it into something like Cloud Foundry," he said.
The cloud can handle situations where realtime data is an absolute requirement and be able to share that data on a realtime basis with multiple users, Day said. A correctly designed and deployed cloud system can accomplish that. But it's not easy, he added.
"This gets into when you're putting data in the cloud, where is the data originating at? You may have something that needs to get data from a POS or other system which might not be in the cloud," Day said. "What's the frequency that those things happen? We look into things like how to set up managed file transfer so clients as they put things into our environment can get updates in a secure, controlled manner."
Organizing and accessing fresh and realtime data was a challenge in the pre-cloud world, so how can the cloud expect to accomplish that goal? Again, it's not easy but it can be done utilizing a NoSQL database with an in-memory data grid to keep all the data that has to remain fresh in memory, Day said.
"A lot of problems people have with data replication as they grow is they start with one database and then start clustering it. The cluster loses its effectiveness," Day said. "What people are doing nowadays is to slot some NoSQL data grids in front of them, things like GemFire, Memcached or MongoDB. Something along the lines of an elastic computer works much better. That's hard to do in a physical world."
When CIOs run applications in-house, their only worry is enough capacity so they typically build big, Day said. But when you move into the cloud you can right-size your IT capacity based on your needs at any particular time. You can get by with a smaller infrastructure and then tap into extra horsepower as it becomes necessary, paying only for the time you need it.
"There are plenty of tools to monitor so you know what your load is. Having a blueprint in place to say when you're above 90 percent utilization, let's snap in another [server] is important. The other thing that can be done in terms of data crunching is the cloud can split the job. You can take batch-type jobs and split them up and run them faster," he said.
A problem CIOs worry about is that virtual machines may not be able to handle a workload with vast amounts of data, Day said. "The concern is that it won't perform as well as having real data based on real iron," he said.
While the best practice five years ago was to run a real database on real hardware, that's not quite as clear right now, Day said. "When you look at a NoSQL solution, it radically does speed things up. But when you move to those types of things you may have to rearchitect the solution. There's some rework there," he said.
Companies such as VMware, however, are working on trying to interpret NoSQL databases as SQL databases. "From an application purist's perspective, that's kind of a hokey solution, but there's a lot of legacy applications that you're not going to rewrite," he said.
Again, leveraging monitoring tools allows the infrastructure to respond on-the-fly by increasing capacity, he said.
Overall, Logicalis sees more customers moving to cloud-based solutions because the technology pros outweigh the cons, Day said. The VAR doubled the number of virtual machines it manages in 2011 compared to 2010.
"People aren't buying the same old hardware in the same old way, but we need to get them computing capacity. We're getting more [cloud] opportunities in a broader way. It's also a savvier customer. They're coming and asking us for specific things. Maybe before it was they just wish they didn't have to worry about this application anymore. Now people come and say, 'We're looking for a solution, a [disaster recovery]-as-a-service.' Whether it's through a box or a virtual machine, either way we win."