Embrace Composable Infrastructure to Stay Ahead of Customer Needs—and Atop Your Market
In today’s competitive market, you need to move quickly to give your customers what they need. If you can’t turn on a dime when their expectations change, your competitors will step in and take your place. This is especially true in ecommerce, and has led to increasingly shorter development cycles.
These days, daily builds are the norm on sites such as ours. Scout24 operates Switzerland’s leading online marketplaces. If you’ve got something to rent or sell, you list it on one of our portals. ImmoScout24 is for real estate; AutoScout24 is for cars; and we recently spun off MotoScout24, a site devoted to motorcycles.
We also operate a free online classifieds site called Anibis. If you no longer need your espresso machine or your smartphone, just snap a picture, put it on our portal, and then sell it. Our sites attract about a million visitors a day, and are amongst the most trafficked in Switzerland.
Scout24 employs some 250 people at our office here in Bern, but only 13 of us work in IT. Nine of us handle operations, and four of us deal with back office stuff like printers, email, and file sharing. As the company’s technology manager, I’m responsible for the technology stack for both of these teams.
Right now, the operations team manages roughly 600 servers. Most of these are virtual machines, but we also maintain two or three physical servers, so we can backport into our VMs should an issue arise.
Don’t Wait for End-of-Life to Replace Your IT Infrastructure
A couple of years ago, Scout24’s management asked the operations team to start looking into replacing our aging infrastructure. At the time, we used HPE C7000 blades that were seven years old. Although these servers were not reaching EoL until 2021, the executive team wanted us to get ahead of the curve by adopting a new platform.
This was motivated in part by our move to agile and DevOps workflows. Our development cycle had shrunk to a day, and we needed hardware that could keep up with our coders. Among other things, this meant diving deeper into microservices, and giving our developers the ability to provision servers themselves.
At the time, developers had to open a support ticket every time they needed to spin up a new VM or create a Docker container. For example, they might have asked for a dual-core I7, with 4GB of RAM, running Windows 2012 to test the day’s build. Unfortunately, our small IT team would have taken up to 48 hours to fulfill their request.
This was holding everyone back. Our developers couldn’t move fast enough to deploy new features and bug fixes, and my operations team had to take time away from maintaining infrastructure and helping other business units.
A Large-Scale Composable Solution
I kept an open mind as I looked at potential solutions. My only certainty was that I wanted blades because they are so easy to manage, and cabling them is a snap. After looking at several other suppliers, I turned my attention to HPE, our primary infrastructure partner at the time.
As I’d expected, HPE was extremely helpful during the evaluation process. They asked about the limitations we faced, and about the capabilities we wanted to add. These conversations helped me to further flesh out our system requirements. After careful consideration, we chose HPE Synergy, which was a brand new platform at the time.
We were one of the first companies in Europe to roll out HPE Synergy on a large scale. We went with four frames, totalling 36 servers, with 10 terabytes of RAM, and 1,500 CPU cores. It took about half a year to migrate from our C7000s, but we are now running 480 virtual servers on this configuration.
Automated Self-Service VM Provisioning
As I mentioned, our biggest stumbling block was the inability of our developers to provision servers themselves. HPE Synergy’s composable infrastructure—coupled with HPE Helion Cloud Suite’s automation features—allowed us to finally change that.
We now have an automated self-service portal where developers can configure VMs and Docker containers as needed. They input their hardware and software requirements, and HPE Helion compiles the necessary packages within an hour. But it’s not just developers using these server provisioning features; in fact, I’ve used them myself to set up test environments.
No longer having to provision servers for our developers has freed my team to focus on managing and maintaining Scout24’s network. At the same time, our HPE Synergy blades are so reliable—and stable—that routine maintenance time has been reduced by 50%. As a result, we can finally tackle all the minor issues that have fallen by the wayside.
I’m talking about things like removing unused cables; deleting old firewall profiles, VLANs, and databases; and updating obsolete IT policies. None of these are a priority, but all need to be resolved before they become an issue. We now have more time to go back and clean everything up.
Dev Blades and Occasional Servers
Another big boost to our IT capability comes from our adoption of HPE Synergy Image Streamer, a provisioning appliance that allows us to manage physical servers as if they were virtual servers. I’m using it to set up new dev blades and to test VMware ESXi and Windows Server environments.
With HPE Synergy Image Streamer, I can create profiles and set up and reconfigure servers in minutes. For example, if there’s an issue with VMware ESXi 6 or 7, I can roll back to version 5 in no time. Speed is critical because Scout24’s sites experience constant traffic. There are no peaks and valleys, and so we have to set up blades quickly at lunchtime, or overnight.
HPE Image Streamer also makes it easy to set up occasional servers on the fly. For example, I have to configure 10 machines as database servers to import forms at the end of every month. When I’m done, I can reconfigure them as multifunction servers until the next time I need them for this specific task.
Our next big project will be a disaster recovery site. We have a barebones DR plan that’s been in place for 5 years, and it’s about time we beefed it up a little. I’m planning and sizing a second data center with active hardware that will serve as our data recovery site, and I trust HPE Synergy to deliver speed and reliability.
Staying Ahead of Customers and Competitors
In the end, it’s all about our customers. Competition in the online marketplace arena is intense. There is always someone lurking around the corner who is trying to make the online experience faster and smoother. We have to outsmart them, and the only way to achieve this is to focus on what our users need.
We are 100% committed to providing the best possible experience for online buyers and sellers, and the best way to do this by empowering our developers and our operations team.
Visitors to our sites don’t need to know about HPE Synergy, our daily builds, or our disaster recovery data center. They probably don’t notice the subtle tweaks we are constantly making, but that’s okay. It’s our job to stay out of the way.