Automated data center

Last week was the 10th anniversary of the day Mark Andreessen changed computing forever by launching the Mosaic browser onto an unsuspecting world. In two months we’ll celebrate the 10th anniversary of the publication of the Lightweight Directory Access Protocol (LDAP) specification by, among others, Tim Howes. Just as 1963 (and the rise of the Beatles) is seen as the pivotal year in the cultural revolution that defined the 1960s, so too 1993 will eventually be recognized as the pivotal year for the computer revolution that made computing part of our everyday lives.

So after you’ve started a revolution, what do you do next?

Andreessen and Howes both wound up at NetScape where they made some noises but were essentially lost in the overwhelming bureaucracy that AOL Time Warner and Sun brought to the company. Fortunately they got out with some cash in hand in time to form LoudCloud before the dot-com bubble burst.

LoudCloud was in the managed server business – it hosted your e-business data center, or provided you the expertise to maintain the one you already had. Since Andreessen and Howes are both inveterate tinkerers and attract other tinkerers wherever they go, it wasn’t long before they were developing applications and services to automate the whole deployment and maintenance of the data centers. Nobody likes to keep doing the same thing over and over again especially when it could be automated.

They got so fascinated with the automation project – and got so good at it – that they recently sold off the data center (to EDS) and changed the company name to Opsware, so they could concentrate on the software itself.

The premise is simple: Opsware automates server and application operations in the data center to provide you with a more efficient operation at lower price. Among the services it automates are: provisioning, deployment, changes, security, consolidation, migration, disaster recovery, auditing and resource reallocation.

If you have a small data center – a handful of servers all running Windows 2000 with either single installations of applications and services, or identical ones – then Opsware might be overkill. But for anyone with the least hint of heterogeneity in the server room(s) they manage, then Opsware should be a consideration in your next budgeting cycle.

If it takes more than one person, working full-time, to install, maintain and monitor your data center servers and applications, then you’re definitely a candidate for Opsware.

Growing companies, in particular need something like this so that they can avoid major downtime. As the company grows so does the IT staff and each new hire, in general, knows less and gets less training than those already in-house. That leads to inefficient – and incomplete – knowledge transfer, which in turn leads to longer downtimes when something goes wrong, but also leads to more things going wrong because provisioning doesn’t get done correctly.

How long does it take you and your staff to patch 25 Windows servers? Opsware estimates 40 hours if done manually. With Opsware automation it takes 30 minutes. If you then need to roll back the patch (and that happens all too often), Opsware can do it in 15 minutes per server rather than the two hours per server a manual process requires. That means that in less than an hour you can roll out the patch, find the hidden “gotchas” and roll back the changes, probably before the majority of people show up at their desks. Or you could be down for two days. Which scenario do you prefer?

Opsware automates everything from the initial install of the operating system onto the bare metal, all the way through eventual migration of apps and services to a different platform for not only Windows, but also AIX, Solaris, HP/UX and Linux.

Ten years from now, when I talk about the datacenter automation revolution, hopefully you’ll be able to say “I was there, I was a part of that. We changed the world.”