The SilverBack folks introduced us to Servprise, a young company headed by a young CEO. But for all its youth, the Servprise WebReboot product line addresses an old sore spot in datacenters far and wide: Namely, the need to safely reboot server hardware via remote access. Typical server rebooting solutions, even from companies as experienced as APC, generally involve power cycling. That makes for messy reboots at the OS level and unsafe power outages at the hardware level. The WebReboot solution is literally like pushing the server's power button, allowing for pillow-soft power downs.
It's funny. Sometimes the products that have the greatest impact are the most difficult to write about because they simply work. That's the case with Universal Electric's Starline Track Busway, a straightforward solution that takes an impressive leap forward in the basic task of providing electrical power to the equipment in the server room.
One of the most important vendors behind our project's success brought no hardware to the party at all. SilverBack Migrations Solutions, based in the US, is a datacenter build-out and migration consultancy staffed entirely by large-stomached ex-corporate IT and facilities geeks. You may recall that, in the main part of this story (see "Five lessons of a datacenter overhaul"), we lamented our inadequate planning and stretched-thin human resources. Both of these are common problems for a project of this scope, and they are exactly the issues that a company like SilverBack can go a long way to mitigate.
Rackwise Data Center Manager (DCM), from Visual Network Design, arrived a little late to our HIG 319 build-out. Generally, a datacenter project would have made use of the Rackwise product from the very start of the planning process. But better late than never -- the DCM software arrived in the nick of time to help us with our sudden weight limitation issue.
APC's NetBotz surveillance system can spot any hooligans invading our precious HIG 319 sanctum, but it can't keep them out. That's where OSI Security Devices' Omnilock 2000 comes in. This high-security physical access system met two key requirements for our specific situation.
It may have come in the smallest box, but Lantronix's SecureLinx Spider KVM had an impact on our project that was far greater than its physical size would suggest. The Spider is a "zero U" KVM, meaning it takes the form of a USB or PS/2 KVM dongle on one end and a dual-port Cat 5e plug on the other. The whole ensemble is light enough to hang off the back of a server, saving you the rack space normally eaten by IP KVM switches and such.
Avocent brought its industry-leading out-of-band management systems to our project, providing IP KVM for PC and Sun servers, service processor aggregation, serial terminal services, and the DSView management server. Because our new datacenter, HIG 319, functions like a multicompany colocation service, we ended up with a wide variety of equipment and at least three different flavors of service processors (Sun, Dell, HP) with three different management interfaces to juggle. Avocent's MergePoint 5224 appliance, a 24-port service processor aggregation system, gives the School of Ocean and Earth Science and Technology's IT group a single service processor management interface, while losing none of the functionality from individual dashboards.
American Power Conversion (APC) played the central role among the vendors in the HIG 319 datacenter project. The company donated a boatload (literally) of APC InfraStruxure gear and no small measure of professional services expertise, but considering the pro bono nature of the project was understandably forced to cut corners, especially regarding the free expertise. We'd recommend engaging APC's professional services arm for a real-life, budgeted build-out. You'll get the same excellent equipment, backed by a full-on, start-to-finish project consultancy that has a few decades of experience managing datacenter infrastructure projects.
What are the three most important ingredients of a successful project? Planning, planning, planning. For our datacenter makeover at the University of Hawaii's School of Ocean and Earth Science and Technology, we planned early and often, and still got bit by last-minute surprises and devilish details that cost us time and money. We'll do it a little different next time. You too can learn from our mistakes.
More servers, more racks, more UPSes, more users -- the reasons for expanding a datacenter are the same everywhere. Today's datacenter projects, however, have the additional component of modernisation. Rebuilding takes place for tighter integration, greener power usage, greater redundancy, and especially more control. Datacenter administrators would control individual dust motes in their racks if they could.