If youâ€™ve tried a faster chip and adding bandwidth, and your network is still too slow, donâ€™t give up. The solution might be right under your nose.
In the movie “South Park: Bigger, Longer and Uncut,” the parents found one simple solution to all their problems: Blame Canada! Inside the enterprise there is a similar call. Whenever there is a computer problem: Blame the network.
It is tempting to answer such screams for faster connections by adding bandwidth, but often that doesn’t solve the problem. Putting in a faster CPU won’t improve a computer’s performance if the chipset and bus can’t take advantage of the higher clock speed. The same way, the upgrading the network won’t improve a user’s experience if the latency is caused by some other component in the system. What is needed is to take a close look at each of the elements, find where the real problem lies, and then correct that.
More often than not, when an end user complains about the network, the problem isn’t the network. Nevertheless, it is a good thing to first check that the network is not where the problem lies. For this you will need to use network management software (NMS) that gives you, at a glance, a view of the overall network health, and the ability to drill down with a few clicks to verify that each link along the way is performing as it should.
If you are using an NMS that has a web interface, you can even make this available to end users over the company’s intranet. This way, they can check the network status themselves before contacting the help desk, which is a great way to eliminate a lot of unnecessary calls. This is also a great way to build user confidence in your competence. When you tell them it is not a network problem, they can look and see for themselves rather than having to take your word for it.
In addition to the current status, however, you will want to also look at the historical graphs to see if there was a traffic peak at a particular time or if the network traffic has been trending higher and you should plan for additional capacity.
If you are using a wireless network there are a few other areas to check into. The user may, for example have switched to some furniture around in his office, so that a filing cabinet is now attenuating his signal and slowing down the data rate. Water also does a great job of absorbing wireless signals and the human body is more than ninety percent water. So, if the person has a couple people standing around his desk talking to him, this may slow down the connection. Another problem arises if you are using 802.11g, with a nominal 54Mbps transfer rate. If anyone on the 11g access point is using an 802.11b device, that entire segment will drop down to the 11b speed of just 11 Mbps. So, a user who is accustomed to a 54Mbps connection will suddenly find a huge drop in speed if someone with an 11b laptop or PDA drops by to visit.
But let us say that the network traffic is moving without problem. What else might be causing slow performance that makes the user think that something is wrong with the network?
First of all, there could be any number of problems with the PC itself. One common problem is the PC running low on memory. When this happens the PC uses a portion of the hard disk as virtual memory. The process of moving files in and out of the hard drive is much slower than getting them out of RAM. While some of this occurs regularly and is not a problem, if excessive it greatly hampers performance. You can often detect this by hearing the disk “thrashing” as it keeps moving data back and forth. If this is what is occurring, you have several options including having the user close some applications or adding RAM.
Another thing to check for, which may be tying up memory, the processor or the network connection is additional software loaded by the user. One network administrator, for example, reports that a remote user called in to support because he had a slow connection over the VPN. A bit of investigation turned up that this person had loaded Gator spyware onto his laptop. The firewall kept blocking Gator’s attempts to call home, but the spyware kept banging away at different ports five thousand times an hour trying to find a way out. If you have a change log as part of your inventory software, you will quickly be able to locate and uninstall any unauthorized software.
Disk fragmentation is another factor that will slow things down at both the server and workstation. Even a new computer will have thousands of fragmented files right out of the box, and this continues to deteriorate over time. Eventually this fragmentation brings files writing and retrieval to a crawl. If it is one user who is complaining it may just be on his own computer, but if several people complain about a particular application, the fragmentation may on the server.
Overloaded CPU. It is unlikely that the CPU itself is too slow unless you have a very old machine. More likely there is a problem with faulty code causing the processor to max out. In tracking down both CPU and memory utilization problems, you can use the Task Manager that comes with Windows. This however is a bit of a resource hog in and of itself, which can lead to further problems on a machine that is already running slow. A better approach is to use a remote SNMP monitoring software which only polls the device every five minutes. This is far less intrusive and in most cases gives you the data you need.
If you are getting complaints from several people start taking a look at the servers. They are subject to the same memory and CPU problems as the workstations, but are handled differently. For example, if the memory and CPU are overloaded, then you may need to set up a load balancer to take some of the work off that server, or switch some applications over to another box.
Disk fragmentation is a bigger problem on database servers than it is on workstations. Since the data is constantly being accessed and rewritten, databases can easily be split into tens of thousands of fragments in a short period of time. In addition, databases are subject internal fragmentation — holes left in the database when records are deleted — and need to be compacted regularly to restore performance.
Another thing to look for on servers is whether large reports, backups or other resource intensive processes are happening during peak hours which can be rescheduled for off hours.
One more way to lighten the load on servers is to use a dedicated appliance such as a load balancer, SSL accelerator or storage appliance to move some of the traffic off the servers and let them do what they do best.
Addressing the above problems will solve most user complaints. But what if you find that it is the network after all? Well, your NMS comes in handy here as well. You may find that the problem is only at certain times and you may need to reschedule bandwidth intensive applications such as video conferences. But if you do need to upgrade, it is far more convincing to present the CFO with a graph showing when and where the network is overloaded and how rapidly demand is expanding. It makes a far better sales tool than just passing along some vague user complaints.
Of course, if none of this works, you can always feel free to blame Canada.
About the Author: Holding a Masters Degree in Computer Information Systems, Michael Patterson has many years of experience. He served for many years as manager of the Network Operations Center at Cabletron Systems in Durham, NH. He is currently the president of Somix Technologies, a maker of network management software such as WebNM and Logalot, based in Sanford, Maine. www.somix.com