When Sue Munro was being interviewed for the role of Management Information Chief at State of Wisconsin a year ago, she asked the organization which tasks would land on her plate if she got the job.
There were some upgrades that needed completing, she was told; and then there was the little matter of “a huge licensing puzzle to solve”.
At the time, Wisconsin was at the tail end of a large consolidation project that had been happening for some time, with multiple state agencies coming onto the department’s administration enterprise platform.
As part of this consolidation, the State planned to overhaul the way it provisioned databases, moving to Oracle Exadata Cloud at Customer.
So the goal was to get consolidated and then optimize the network. There were discussions over the menagerie of licenses and they were contemplating just rolling it into one big CSI.
Despite this approach setting off alarm bells for Munro, she joined the organization, her first public sector role after a lifetime in the corporate world. She then embarked on a couple of months of analysis and discovery into the State’s IT infrastructure, and discovered the scale of the challenge.
At the time, Wisconsin had a portfolio of 23 Oracle CSIs owned by six different state agencies deployed across several environments, mostly VMware, deploying four different storage solutions and four different flavors of Linux or Unix running on 22 physical servers.
Munro needed to make sense of all this, so the next stage was evaluating all of the CSIs. Munro found there were components and products in many of them that the State no longer had any use for or that weren't important anymore.
But if you're familiar with how Oracle licensing works, you can't just drop off part of your CSI. It's an all or nothing deal. So when you renew it, it's all or nothing. But is it cost effective to purchase a new license when you've got a big investment? Generally it makes sense to continue - unless there's only a very small part of it left that's useful to you - generally you're going to keep renewing that.
The issue here, as Munro pointed out, is that there's a three percent uplift every year for license renewals.
So over time, just in this simple illustration of mix of products that we had and you might also have, you can see how the cost over time to support that same entitlement and usefulness goes up and your cost per processor goes up.
So there's a tipping point where this isn't as useful as it was when you first purchased it. Initially you had your return on investment, and that's where your fixed value is out of your purchase of license, your capital investment, but at some point you've crossed the line where you have a point of diminishing return, where your usefulness of that CSI and what you purchased, just isn’t giving you the horse power or value that you really need.
Based on this realization, Munro identified various CSIs that weren't required anymore based on the processors she had running, the physical servers and the limits of products needed. So it was an easy decision to not renew those.
That, right off the bat, saved the State of Wisconsin $200,000 in annual support renewals. Just doing that kind of comprehensive analysis, figuring out what do I really need to operate today, what am I going to need to operate in the next fiscal year.
But that was the first pass, just the low-hanging fruit. The CSIs that I could simply drop and not renew.
Munro then moved onto looking at how the State was currently using its available VMware resources.
By adjusting some memory, dropping some databases that didn't need to run, doing some VM consolidation, I was able to pull out a couple of physical servers and start to shrink that VMware environment in anticipation of going to a 100 percent Exadata platform in two years. That again led me to a few more CSIs that I didn't have to renew coming into fiscal 19.
As a consequence of the State’s discussions around its Oracle licensing strategy, it came to learn about the subscription model and decided to deploy Exadata Cloud at Customer.
There's isn't a software investment, it's a more of a utility model. You pay for what you use and when you're done using it or if you decide it doesn't work out, you can get out of that licensing agreement the same as with any other cloud subscription. You can also flex it if you have surge demand or find that you need to grow over time.
Are you ready?
But the Exadata rollout brought new challenges for Munro. She noted that a key factor with this model and the way customers like State of Wisconsin provision the infrastructure is that Oracle still owns the equipment, it just happens to be sitting in the customer’s data center.
The key difference here is Oracle does not ship the equipment until you prove that you're ready to accept it, plug it into power with the networks available. So there's a process you have to go through, the site readiness template to fill in, provide all the IP addresses. You need to have the firewall rules established. Everything that you need for the support gateway to interact with the Oracle Cloud operations.
Once you believe that you're ready, Oracle will send a field engineer to your data center to check and verify that you have the cabling ports on your switches and the power receptacles ready, so that when that equipment arrives it’s ready to go. The reason is, again, Oracle owns the equipment and so they aren’t going to risk that the equipment arrives at your data center and then it sits for two months while you get electricians there and your network team gets the reports ready.
It's money and opportunity lost for Oracle because you don't start paying your subscription until the service is activated.
The template is a fairly complex and lengthy document, so Munro recommended planning for about a month of work with the Oracle team to get all the details completed, and another eight to 12 weeks for the software installation environment validation until you get to the point you when you can sign in and start using the cloud platform.
Shipping cycles for Exadata machines are also highly variable between when you place an order and when you can actually take delivery of the product, Munro advised.
We placed the order in May, but we couldn't take delivery of the system any sooner than July 1st; we had to wait until the beginning of the fiscal year. So in May, when we thought we could potentially proceed as soon as we have the site readiness form and the power provision in place, the shipping time was a couple of weeks. So if you ordered it today, and you're ready, they would ship it next week.
When we get to July and said, okay, we're ready, we’ve got the power ready, when will you ship it? They’re like - the middle of September. So that's where our account team came in handy in helping us on how to accelerate that plan.
So plan for that in your project plan, and keep in close contact with your account team to understand what the current shipping times are for equipment.