Linux has truly proliferated in the enterprise. In fact, I would be willing to say that most large organizations have a Linux box sitting in a corner performing some critical, but not necessarily high profile, task. Unfortunately, a lot of these machines are old Pentium 486, or even 386, machines that were pulled out of the storage closet and pressed into service.
I made my own mistakes in that regard. When I took my old workstation and slapped Linux on it, I did it more as a lark than with the goal of making it a serious piece of our network infrastructure. The original specs called for Internet access and e-mail for maybe a dozen people. An old Pentium 133 with 32MB RAM was perfect for those specs, and for six months, it worked great.
Now that same machine is expected to serve up access to 70 people. Today, it is just squeaking by. E-mail checks by 70 people about every five minutes just overwhelms the poor little guy, resulting in bad logins, long delays in mail retrieval, and denial of service. However, it works like a pack mule. This machine, which could not run Windows 95 for a full day without crashing, has run for nine months. I’ve shut it down twice; once due to an extended power outage and the second time to add a modem.
Figuring out if you need more horsepower
Here are my simple tips to help you determine whether you need more hardware than you have:
- Do you experience slowdowns when trying to log in or execute certain programs?
- Have you received a lot of bad login errors when checking your POP account?
- Does your latency spike when going through your Linux-based NAT at certain times of the day?
- Does your machine stop responding at all for brief periods of time?
Some commands to help you out
- Use top (or its x-based equivalent, gtop) for a quick rundown of active processes and CPU/RAM utilization levels. My physical RAM was tapped out, pinpointing my real performance problem. This is an excellent utility, if you want to watch it for a while.
- Use gstripchart to give you a rundown of CPU utilization, RAM utilization, Virtual Memory utilization, Load Average, and Network activity in and out. Plus, It also puts this information in a strip-chart style vaguely reminiscent of Windows NT’s performance monitor.
As you watch either of these utilities, look for common problem factors such as Load Average over 50 percent, RAM utilization in the high 90s, or extreme usage of your swap drive. Also, be on the lookout for odd items—things like a user who checks mail multiple times at once, a user that logs in multiple times at once, or a machine that refuses to drop its connection.
Tell us how you determine what needs to be upgraded on your Linux box. What are some of your favorite load-checking utilities? If you'd like to share your opinion, start a discussion below or send the editor an e-mail.