SAN FRANCISCO -- Enterprises have leapt all the barriers to Linux adoption in the data center and are focusing
on practicalities, according to Linuxcare Inc. CEO Avery Lyford. Lyford, a speaker at this week's LinuxWorld Conference & Expo, said that enterprises finally began deploying Linux this year after running pilot programs last year. It's been a stark evolution since 2001, when enterprises were gun-shy about Linux. In this interview, Lyford offers some answers to this year's top question as he discusses how to make Linux work in the data center.
Today, what are the perceived and real barriers to deploying Linux in the data center?
Lyford: In 1998, Linuxcare started out by saying, 'The big obstacle to Linux adoption is support.' Then, we ended up doing a bunch of work for OEMs, like IBM and HP. Now, commercial companies have gotten on board with supporting their Linux customers and [saying] that Linux is an equal, if not more equal OS to Windows or Solaris or HP-UX or AIX. So we're past that old obstacle.
In 2003, a lot of the perceptions about barriers have been laid to rest. Now, people are dealing with practicalities. In the past, they deployed platforms that had a set of tools wrapped around it that made it into a complete offering. As we were working with customers, the thing we saw clearly lacking was a set of tools to make Linux more manageable, manageable around things [that are] important, like rollback, for instance.
By the way, they need to roll back because -- even if they've done a lot of testing -- at some point, somebody will take the wrong thing in and cause the system to go down. Then they'll desperately wish they could take back their 'enter.'
Do you see a movement in businesses to consolidate on Linux at the data center level?
Lyford: Absolutely. Businesses are consolidating all those servers they bought a few years ago.
During the [dot-com] boom, people just wanted to get things deployed as quickly as possible. So the model for the last couple of years was all about speed. How getting deployed was accomplished was not so important. So people used all sorts of, let's say, baroque architectures.
Now what's happening is people are saying: 'That's all very well and good, but we're left with an infrastructure that's very expensive to maintain.'
What's gained by consolidating on Linux?
Lyford: The data center team is responsible for sets of servers and charged with figuring out how to make them more efficient. They're finding consolidating with Linux in the data center hugely attractive because it allows them to reduce the number of platforms in the data center.
Linux is a great substitute for all of the hardware-based or hardware-bound Unixes, such as HP-UX, AIX or Solaris. Standardizing on Linux reduces the hardware and software costs and the complexity of what IT shops have to maintain.
So why is there so much talk about teaming Linux and mainframes? It seems like the two are polar opposites.
One of the things that happened in the boom, despite all the changes, was that existing databases continued to be run on mainframes. Wells Fargo, for example, still has its customer records on IBM DB2 on a mainframe. In fact, a fifth of the world's data still resides on mainframes.
I see Linux in the data center in a bi-modal way, on the mainframe and on Intel servers. There is a big spike of people using Linux on the mainframe. There's an even larger body of people using Linux on top of Intel. They're using Linux on Intel both in distributed and data center environments. In the data center, Linux on Intel and on the mainframe is being used to consolidate file and print and application servers and more.
How did Linux make the leap from being a Web server OS to a data center OS?
Lyford: Well, when people were using Web-facing applications, for example, they started doing extracts from data using Sybase transforms to WebLogic running on Solaris. So this whole string of servers could do a set of transforms. But, boiling it down, what they're doing is taking data from a mainframe and putting it on the Web. A much more elegant architecture is to run DB2 on top of Linux on a partition on the mainframe. Then, rather than ending up with disaster recovery on nine boxes, you're doing it on one physical box.
What's the most common mistake IT shops make when consolidating servers?
Lyford: Best practices arise through making mistakes. The challenge is taking advantage of and learning from other people's and your own mistakes. That's why there's a saying in the British navy that all navy regulations are written in blood.
Not paying attention to the network implications of server consolidation is one of those mistakes that you make that you swear you'll never make again.
Consolidating servers can have wide-area networking implications. You're changing physically where the servers are located, but networking architecture has those servers located differently. So you better make sure that you talk to your networking group before you change server locations. You could be changing where IP addresses are and changing wide-area network loads.
It's frequently the case that that networking group reports through a whole different management chain. So you could have done all this elaborate server consolidation planning and never talked to them.
So it's people issues and not technology issues that gum up server consolidation projects most often?
Lyford: Yes. People get too dogmatic. Don't make it a pitched battle between two IT groups. That's an organization autoimmune disorder. People come at server consolidation from a standpoint of 'it's about the data center people beating the distributed computing team.' That's a tragedy.
FOR MORE INFORMATION:
FEEDBACK: What barriers remain to widespread adoption of Linux in the enterprise data center?
Send your feedback to the SearchEnterpriseLinux.com news team.