Over the years, we’ve been fortunate enough to receive a lot of attention for the way we leverage automation in the data center to operate efficiently, provide a better quality of service and improve the customer experience.
When we first set out to automate our business, we weren’t trying to set any records or anything like that; we were simply looking for ways to automate the work we had to do every day while building, operating and growing our company. As our capabilities advanced, we discovered that the same tools we were creating for ourselves were highly desired by our customers as well, so we created LEAP to extend to our customers the same level of automated control that our data center staff has. Overnight, it was a huge hit.
But that’s not what this post is about. Not exactly. Let me take a step back for a moment.
So some people find it surprising when I go on about all of the handwork that happens behind the scenes here to ensure everything is ready to go when a customer presses the ‘Deploy Now’ button in LEAP or on our website. As you probably guessed, we have developed a wide range of automation tools for this work. But I thought instead of talking about that, it might be enjoyable to give you a behind the scenes look at what happens here before you push that ‘Deploy Now’ button. This is the work that is hidden in the background, out of sight. The people who do it are as talented as they are invisible to our users.
Here’s a fairly sequential look at what happens before you deploy a server with SingleHop:
- Hardware Selection. Because, just like people, servers are what they eat, we only use name-brand top-of-the-line components. We don’t cut corners to try to save a few points on pricing. We strictly use Intel, Supermicro, Western Digital, Seagate, Kingston RAM, and so on. The same is true of our networking gear, which is all Cisco.
- Inventory Forecasting. A month or two before an order comes in, we are already busy preparing for it by using our internal models to forecast the number of—and most importantly—configuration of—the servers customers will statistically deploy over the next month. This means estimating how much memory, hard drive size, how many drives per server, the model of—and number of —processors, how many with RAID configurations, and other basics. This results in hundreds of permutations of configurations, which enables customers to later find servers that precisely meet their varied needs.
- Meticulous Racking and Cable Management. Racking is an art. At SingleHop, it’s a way of life. When it’s time to place a batch of new servers into one of our racks, we have what we call a rack-a-thon, but you could also call it a textbook case of teamwork and precision craftsmanship. Each server is fastened in place by hand, we cut each cable to size and we document everything — down to which APC port the server is plugged into.We also take care to keep all of the wiring in the rack nice and organized so that the server’s airflow isn’t obstructed by a jungle of hanging cables. That way, it’s also a lot easier to work on.As we work, we exactly catalog and map the location on the rack of each server, as well as inventory its exact configuration.
- Automated Audit. After racking, we run an automated configuration and hardware diagnostics tool we created to ensure each server powers up, contains the correct amount of memory, processors, and other specified components, and that all are in good working order and ready to go. The audit scans for:
- Is the hardware installed correctly?
- Does it have the specified amount of RAM?
- How many hard drives are installed?
- Is there a RAID card?
- How many hours have the drives been in-use?
- And much, much more.
- Ready-to-Go: Into the Provisioning Pool. Once the server passes our audit it is ready to deploy and at that point it’s automatically powered off to conserve resources, and assigned to what we call our “provisioning pool.” At any time we have more than 1,000 servers in the provisioning pool ready to go. Servers in the provisioning pool are ready to be powered on and deployed when a customer places an order.
- Match Making. Once the ‘Deploy Now’ button is pushed by a customer, our automated systems activate a server with the same (or better) configuration as requested. When users specify which of our data centers they want to be serviced from, the server is activated within that location.
- Provisioning. Once the match has been made, automated deployment process begins. This includes installation of operating system and updates, user-specified applications, network configurations, firewalls, and security hardening. A second automated audit is conducted. As a final step the server goes through a quality control check by a technician. Only after this does it touch the network, and the customer is sent the server’s login credentials and it becomes part of the customer’s cloud infrastructure.
The great news is that what we do is exactly what any conscientious IT professional would want to do—only we do it, so they don’t have to. Our exacting preparation also greatly reduces the elapsed time from the pushing of the ‘Deploy Now’ button, to the customer going to work with their new server.
One of the Results: Fewer Mistakes and Less Human Error
Many competitors deploy servers by hand, building each server manually, to order, in real-time. They install the operating systems by hand, configure the software by hand, and audit by hand. This introduces a lot of potential for human error. Something our process is designed to minimize. For a human being to deploy a server in the same amount of time that our automated platform can, they would have to be in a big hurry. That would exacerbate the likelihood that they make a mistake. Think about that next time you need a mission critical server deployed right away, at the last minute.
You might be wondering. How long does this take?
The entire process from when a customer pushes Deploy Now to when the welcome e-mail is sent has averaged less than 45 minutes (42 to be exact) for 2 consecutive years. Some servers are deployed as fast as 21 minutes (no control panel, standard configurations) and others take a little longer. With deployment times this fast, it sometimes is hard to forget that these are physical servers. Not VMs.
All the technologies we use were developed right here at SingleHop. Nothing was outsourced or purchased off the shelf.
What questions do you have about how data centers operate?