Can a cloud model surprise you… in a good way?

We all hate surprises. Too often they’re bad things. But occasionally you can get surprised in a good way. That’s what happened to an IT manager I ran across this week. He achieved some remarkable results in a way he did not expect: in the cloud.

His company, a large financial services provider, needed to improve the scalability of an online service offering. Batch jobs were already running too slowly with the number of users they had. And they were expecting the user base to grow significantly. What would they do then?

A cloud service provider thought he could help, but the IT manager was skeptical. Would it be fast, secure, reliable, etc?

The cloud provider was able to satisfy the security and reliability concerns by explaining the architecture. The underpinnings relied on Xsigo virtual I/O, so they were able to provide isolated connectivity to servers. It also let them configure redundancy and fast failover throughout the infrastructure — whether a connection or an entire server failed. It helped that the provider was an established managed hosting company, so they were a known quantity from a business perspective.

But that still left the performance question. And the only way to answer that was to try it. You can talk about car engines all you want, but until you take the car to the track, you won’t really know its top speed.

So the cloud provider set up a test: a batch job with data from 100,000 users. This job that took 2 1/2 hours to run in the IT manager’s own shop (their forecasted number of users was 400,000, so you can imagine why the IT manager was anxious for a new solution). The goal was to drop that to 1 hour. That would be a success and would allow the IT manager to safely expand his user base.

When they ran the test, the surprising result was this: it took six minutes. 96% less than the original, and 83% less than the goal.

That’s good, but would it scale? They tried it again with a much larger data set and found that it scaled nearly linearly.

It’s not often we’re surprised in a really good way. And with all the fear, uncertainty, and doubt around the cloud (the pundits see all kinds of threats), it’s pretty easy to believe that any surprises are likely to be bad ones.

But if this IT manager had gone with that flow rather than looking at the potential upsides, he would have missed a solution that saved his company significant time and money.

He had a healthy skepticism… which is a form of preconception. But he got over it.

Pods are everywhere! From ubiquitous music players (iPods) to pods that threaten humanity (Invasion of the Body Snatchers), the pod concept thrives in many forms.

Now the pod is becoming a central element of many data center deployments. It is a trend to watch because pods are already revolutionizing data center management. What are these pods, and why do they matter?

A pod, in our thinking, is a standardized unit of infrastructure. A mini data center-in-a-rack that is equipped with servers, connectivity, and perhaps storage.  The exact configuration will vary depending on each organization’s requirements. But within a data center each pod will look just like the one next to it.

Why do this?

The objective here is to save time. Employing pods standardizes deployment and management tasks, which means less time needed to complete these tasks. And that translates to higher efficiency and better resource utilization. Read More

At EMC World this week, the theme was “The Journey to the Private Cloud.” Why did they choose this theme? Well you could be cynical and say that is was simply to highlight their new storage architecture for the cloud. But I believe that’s thinking too small.

Cloud has become real. One observer noted that most IT people in attendance  no longer questioned the value of the cloud. They were looking for the best strategies to get there… quickly. EMC picked the theme that simply reflected what most people are thinking about.

And there’s little question as to why. Over the course of the event, I met with hundreds of people and heard their concerns: the excessive time required to deploy infrastructure, the high levels of complexity (one person said he had 22 I/O cables running to each of his servers!), the difficulty of pinpointing issues when they occur in a complex environment, and the I/O bottlenecks that block the path to higher utilization.

One attendee from an insurance company put it bluntly: “When we have an I/O issue, it’s usually easier to add another server than to fix the issue on the old server.” A frank admission. And an obvious wake-up call that doing things the old-fashioned way does not work anymore.

VMware has of course been espousing the values of the cloud for some time now. Their server virtualization and management technologies are of course key enablers.

We would maintain that virtual I/O is a key enabler as well. It resolves all of the I/O issues that arise as you move towards a more fluid, more easily scaled infrastructure. VMware agrees, which is why they use virtual I/O in their own environment to make it that much more powerful and that much simpler to manage.

For an in-depth look at virtual I/O and the cloud, I’ve condensed the blog post series, “Open Solutions for the Private Cloud,” into this white paper. 8,000 attendees at EMC World were thinking about this topic. If you’re not yet, you should.