up vote 0 down vote favorite
1
share [g+] share [fb]

We host a number of different servers. Windows 2003 server, linux etc. And we have a number of clients who connect into our systems using RDP, several through wireless networks that we provide. The wireless networks generally run well and we don't have many problems with them.

Several of these clients also have Linksys VOIP phones that connect into our asterisk server and provide phone services. Most of these run over the wireless links.

My question is, what tools would people recommend to monitor network performance and assist in improving general network reliability. Free ones that work on Linux & windows would be of benefit to us.

Also, I'm after network design strategies to improve performance/reliability as well.

link|improve this question
63% accept rate
feedback

3 Answers

up vote 1 down vote

As far as designing for good performance is concerned:

  • Use "modern" routing protocols where you're routing (OSPF is ideal, IS-IS if you have the needs, EIGRP is easy to deploy, but OSPF forces you to design the network up-front), this lessens the proportion of traffic taken up by routing.
  • Use VRRP or HSRP, with dual gateways, for each LAN that you want some protection on (this is why I recommend using a routing protocol instead of static routes). This lessens the risk of a single device taking your network out.
  • As far as possible, split switch-ports into "will have switches/hubs on them" and "will have single devices", this allows you to disable Spanning Tree on the non-hub ports. Ideally combined with auto-detection of BPDUs to block the port if a switch is detected.
  • Measure. Without measuring throughput and traffic levels, you cannot say what works and doesn't work to improve your network.
  • If applicable, deploy QoS to treat different types of traffic differently (allow a proportion of bandwidth for RTP, forward this preferentially; guarantee a proportion for VoIP signalling)
  • As far as possible, enable WRED on any port that may be carrying TCP, this tends to increase your overall throughput, by avoiding TCP global synchronization.
  • If at all possible, split your voice and data LANs (either physically or by using VLANs).
link|improve this answer
feedback
up vote 0 down vote

For network analysis you could use ntop or wireshark to monitor your main switch.

My understanding of wireless is that once you have over 100 connections or so, planning can be tricky. Especially if you want to allow seamless movement from one access point to the next. You have to be pretty aware of the range of each access point (using netstumbler, or an equivalent), adjusting antenna size and type or unit positioning, ensuring overlapping units are not using the same channel.

That's pretty basic stuff, but perhaps it can get you started in the right direction.

link|improve this answer
I forgot to mention that the wireless networks are not consumer grade units. They are outdoor units with a range up to 10k. Generally we get very good performance with these. It's the ability to monitor usage and outage that I'm interested in. – Matt Nov 24 '09 at 1:57
Generally having a shorter, more defined range with predictable overlap areas will be better for multiple wireless units - unless you give them each their own ssid and channel. In this case, since each AP covers your entire building, that might be your better option, as roaming between APs would not be necessary. You can monitor for outages (and send alerts) using nagios, and the other tools I mentioned wil help you monitor usage. – Brent Nov 24 '09 at 14:44
feedback
up vote 0 down vote

To monitor links performance:

To monitor and analyse packets:

  • nTop
  • Snort <- This is IDS, but based on your rules it can find malicious behaviour.
link|improve this answer
feedback

Your Answer

 
or
required, but never shown
discard

Not the answer you're looking for? Browse other questions tagged or ask your own question.