<html><head><title>Re: [smokeping-users] Graph throughput distribution via SNMP</title>
<META http-equiv=Content-Type content="text/html; charset=iso-8859-15">
</head>
<body>
<br><br>
<br>
<span style=" font-family:'courier new'; font-size: 9pt; color: #800000;"><b>DSJ> Dear list,<br>
<br>
DSJ> I love Smokeping not only because it puts emphasis on (near)<br>
DSJ> end-to-end performance, but also because it provides cues (and<br>
DSJ> metrics) to its variance: this website has a median response time<br>
DSJ> of 150ms but also a std deviation of 100ms. It is easy to show<br>
DSJ> and compare the effect of caching and load balancing with this insight.<br>
<br>
DSJ> But there is a more basic domain of networking where I see<br>
DSJ> engineers oblivious to variability, link throughput. <br>
<br>
DSJ> More often than not, the references are 5min averages. We saw a<br>
DSJ> large company accepted for budget reason to replace its 1Gbps MAN<br>
DSJ> access by 100Mbps MPLS WAN connexions because of the average load<br>
DSJ> was around 60-70 Mbps. The users reactions were quick to come!<br>
<br>
DSJ> That's why I was wondering if a Smokeping SNMP probe was (or<br>
DSJ> could) be diverted to make several measures of throughput during<br>
DSJ> one step. For example, 10 measures - 15s each apart - fit largely in the std step of 300s.<br>
<br>
DSJ> In Smokeping source code, I see the use of SNMP-session module,<br>
DSJ> but I am no Perl Expert. And I didn't find reference to something<br>
DSJ> as I imagined in the doc or mail list archives. <br>
<br>
DSJ> Feasible? Already done? <br>
<br>
DSJ> As a recent user of Smokeping, I look forward your feedback. <br>
<br>
</b><span style=" color: #000000;">I happened to re-read this, as I just glanced at it the first time. [And I see I misunderstood your intent and question. !oops!]<br>
<br>
I see you're wanting to query an SNMP OID to see what the level of throughput is at a specific moment in time. [Rather than attempting to determine actual connection capacity in real throughput.]<br>
<br>
Is there a reason you're not using [or wanting to use] MRTG to do so?<br>
One complication is that if you're using SMTP naked across the internet, that's probably a substantial security issue. You'll want to make sure your SNMP traffic is handled via SSH/VPN/Tunnel etc.<br>
<br>
But, no, I'm not aware of any ability of smokeping to handle SNMP queries like what you're asking for, and I can't imagine anyone writing one when MRTG already does this, and it's even written by Tobi, and IIRC was around long before smokeping.<br>
<br>
Nagios may be able to query SNMP, but it won't keep long-term stats like MRTG or Smokeping do.<br>
<br>
<br>
-Greg<br>
<br>
</body></html>