How to Best Test Throughput (or Load Test)

[kc6108]

As the title suggests, I would like to get some discussions started regarding how to best load test a home network setup. I have used programs such as iPerf, NetStress, and my favorite, LAN Speed Test (using optional LST Server).

I really like how IxChariot is used here for reviewing routers, etc., but I don't have the slightest clue on how to download the software let alone how to use it.

I would like a new Forum section started on how to load test (and test throughput) using some of the best programs available.

I'm sorry if this has already been discussed. I searched but couldn't find anything.

 

[thiggins]

What are you trying to accomplish with the "load test".

IxChariot isn't free.

 

[tipstir]

LST free (lan speed test) works the best. Just run from the PC you want on your network.

Basically PC A xxx.xxx.xxx.xxx connected to PC B xxx.xxx.xxx.xxxx then select the folder on PC B to use. Select the size of the file you want to test. I use 1024MB then site back and watch the number write and read. Your results will appear on the bottom.

This program can be used on Wired and Wireless LAN for testing on PC side. If you got Android Smartphone or Tablet you can download Xtreme WiFi and test that way too. That's mostly in KBPS though. But the higher the number the better the throughput is. I've been working on re-programming tablets which are stuck at very low sluggish throughput. Now they almost match my 35MBps down over WiFi.

But there are so many test programs there was one called;

netcat
features others but they all test your network. Some use
client and server programs.

 

[kc6108]

What are you trying to accomplish with the "load test".

IxChariot isn't free.

Yeah, I know IxChariot isn't free. You use it a lot so I mentioned it. I though maybe you knew of a way to get to trial or striped down version for us to use.

I mainly use the tools to test throughput on all of our devices which includes tablets (iPads, Xooms, etc.), PCs (towers and laptops), phone (iPhone, etc). I not only use this tools to test for max throughput, but also for troubleshooting. I make 90% of all the cables I use. Sometimes some of runs don't end out working very well (all though all the wires are crimped correctly) for one reason or another.

There are lots of reasons to discuss all the tools I mentioned, as well as, their strengths and weaknesses, but I do have some reasons of my own.

We run our own small business out of our house. It is a rather large home with three levels. We use a Cisco SA540 (no wireless capabilities) as our gateway router and three E4200 v1's as APs (use the *bridge* mode in the GUI). We even utilize the guest network functionality for guests, etc. to use.

We have lots of devices (40+) configured with reserved IPs (by MAC address) versus using static IPs in the devices.

I am telling you all of this because I high throughput, even under load (even by one or more device) is important to us. We move/copy/backups large amounts of data/files/etc all the time. I would just like to get everyone's input on how to best test all of our network setups so we can tinker/tweak/etc to make sure we are running optimally.

I hope this make sense. I just thought it would make for some good discussion. I have searched around and haven't found a forum that addresses throughput/load testing/etc in much detail.

 

[kc6108]

FYI, we have found that our SA540 (and E4200's of course) does a great job of handling our network load. Even when we have 15+ active devices we don't notice much, if any, slow down.

We have had as many as 40 active devices (including guests on the Internet) and our setup works rather smoothly. There are times that things slow down a bit, and rebooting our routers are NOT an option. We have as many as 15 devices remotely connected (via IPSec and SSL VPN) and reboots are scheduled. We only reboot once a month (on a Saturday night) and only for preventive maintenance. We have gone as long as three months without rebooting due to business reasons.

Because we are a *small shop* we utilize home user grade routers (well except for the SA540, it is considered a small business device). We use the SA540 as UTM device because that's what it's designed for. We have purchased three year licenses for IPS (intrusion protection system) and Trend Micro's ProtectLink Web (cloud service that save guards us from illegitimate websites, etc). Between the SPI firewall, IPS, and ProtectLink we feel *comfortable*. I use that term loosely...

The IPS and ProtectLink functionality comes at a price. They do slow down our network but it is worth it to us.

I am just trying to give some background so some of the more knowledge network folks (which does NOT include me) can help us out in terms to testing/validating our network setups are providing high/optimal throughput.

Maybe I'm alone here, but I think there are folks that interested in testing their network setups to determine just how much their setups can handle. In my case, I want to know just how far I can push my setup before I need to upgrade and/or change things up.

 

[tipstir]

If it working fine why are you trying to push it. Unless you feel it could be better? Newer technology will always be around the corner. Network Hardware does fail not 100%. If you invest more into might get better equipment. Most sits sigh on with Cisco on lease bases for continuous support. They get the latest hardware but you still have to purchase out of pocket.

As for E4200 some of those routers were good and some were duff (bad) E4200V2 (aka EA4500) gives you better CPU double the RAM. I am sure CISCO coming out with more RAM versions for 2013. To me these two models still over heat and CISCO need to put in temp sensor fan on the CPU.

As for testing you can hire outside company to come in to test your network and tell you how it's really doing if you don't want to do this yourself. Time consuming at best.

 

[kc6108]

I would (and could) only perform a load test during a scheduled maintenance (some Saturday night) window. We cannot afford to perform such a test at any other time. Our network is extremely stable, and we depend on that stability.

Actually I am not looking to upgrade. I am just want to verify that the times that our network slows down, that it isn't trying to tell me something (like it's time to upgrade, or I have a bad cat5e cable somewhere, etc.).

Right now our LAN throughput is capped out at roughly ~350 Mbps which works fine for the most part. It would be nice to be able to achieve 700+ Mbps, but I think our cabling and NICs are more to blame for the slower speeds than the routers. That's just a guess on my part.

We do a lot of data movement between devices. Sometimes the data movement is time critical. You know... time is money sometimes!

Thanks for the replies so far guys.

 

[thiggins]

If you are just checking for weak spots, simple drag and drops of large files can work. Or tools like those mentioned in How Fast Is Your Network? Five Ways To Measure Network Speed

But first start with an accurate map of your network. I'll bet just by inspection and knowing general traffic patterns you can see where the bottlenecks are. You might be able to address them by simply grouping high-traffic devices on the same switch.

Cabling could be holding you back. All long runs should be CAT5e or 6. Any system that doesn't use PCIe to connect the network and CPU buses will be limited to below 500 Mbps. Don't worry about jumbo frames. They won't help much with modern CPU architectures.

 

[kc6108]

Thanks for responding again Tim. I was hoping to get your input.

I have actually read that article before. It is where I first heard about NetStress.

I forgot to mention in my above posts that I have *several* 5- and 8- port D-Link unmanaged switches. They are the older DGS-2205 models, not the newer DGS-1005G models. Are you familiar with them? Better yet, have you reviewed (or actually used) one? Could they be my bottleneck?

If they are the bottleneck, what *cheap* switches could I replace them with to make it work the expense?

Personally I don't think they are the bottleneck because we get the same throughput with our E4200 v1's (remember they are setup as APs). I am talking just hooking two PCs to a single D-Link switch or E4200 yield ~350 Mbps. That's why I am thinking the NIC's are more to blame.

Thoughts?

 

[thiggins]

You don't need to worry about exceeding a Gigabit switch's capacity until you get above 24 ports. Even then, I think those models are "wire speed" between all ports.

Bottlenecks form when there is too much traffic on uplink ports. For example, if you have certain Gigabit clients that often transfer large files to/from a server/NAS and could be doing it simultaneously, then you might be overloading a single Gigabit uplink port.

If you are seeing only 350 Mbps between two clients on the same switch, that points more to a NIC or PC architecture/horsepower issue. The switch is not going to be the limiter.