<html><head></head><body><div style="color:#000; background-color:#fff; font-family:HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, Sans-Serif;font-size:14px"><div id="yui_3_16_0_1_1445580371658_2539"><span></span></div> <div id="yui_3_16_0_1_1445580371658_2748">Hi Nishal,</div><div id="yui_3_16_0_1_1445580371658_2749"><br></div><div id="yui_3_16_0_1_1445580371658_2735">I'm updating the blog post, based on your feedback. </div><div id="yui_3_16_0_1_1445580371658_2735"><br></div><div id="yui_3_16_0_1_1445580371658_2735">Thank you again for bring some light into this. </div><div id="yui_3_16_0_1_1445580371658_2735"><br></div><div id="yui_3_16_0_1_1445580371658_2735" dir="ltr">My personal opinion on this is that we need better statistics. What could be done is requiring ISPs put RIPE atlas probes at each ISP's end on at least 2 customers. Then, the website for MIXP should constantly display statistics such as latency. Showing only traffic transiting like it does right now does not with diagnosing peering problems. </div><div id="yui_3_16_0_1_1445580371658_2735" dir="ltr"><br></div><div id="yui_3_16_0_1_1445580371658_2735" dir="ltr">So, for ISP A to ISP B, we would get the latency. For ISP B to ISP A, we would get another latency. </div><div id="yui_3_16_0_1_1445580371658_2735" dir="ltr"><br></div><div id="yui_3_16_0_1_1445580371658_2735" dir="ltr"><br></div><div class="qtdSeparateBR" id="yui_3_16_0_1_1445580371658_2576"><br></div><div class="yahoo_quoted" id="yui_3_16_0_1_1445580371658_2559" style="display: block;"> <div style="font-family: HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, Sans-Serif; font-size: 14px;" id="yui_3_16_0_1_1445580371658_2558"> <div style="font-family: HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, Sans-Serif; font-size: 16px;" id="yui_3_16_0_1_1445580371658_2557"> <div dir="ltr" id="yui_3_16_0_1_1445580371658_2561"> <font size="2" face="Arial" id="yui_3_16_0_1_1445580371658_2560"> On Friday, 23 October 2015, 0:35, Nishal Goburdhan <nishal@controlfreak.co.za> wrote:<br> </font> </div> <br><br> <div class="y_msg_container" id="yui_3_16_0_1_1445580371658_2556">On 22 Oct 2015, at 22:01, Loganaden Velvindron wrote:<br clear="none"><br clear="none">> Hi All,<br clear="none">> I wrote about my adventures at attempting to fix the huge latency in <br clear="none">> Mauritius:<br clear="none">> URL:<br clear="none">> <a shape="rect" href="http://logan.hackers.mu/2015/10/mixp-latency-problem" target="_blank" id="yui_3_16_0_1_1445580371658_2838">http://logan.hackers.mu/2015/10/mixp-latency-problem</a><br clear="none"><br clear="none"><br clear="none">logan,<br clear="none"><br clear="none">i think that you have some misconception on how the MIXP (or any IXP) is <br clear="none">meant to operate.<br clear="none">this sentence is correct:<br clear="none">“The Mauritius Internet Exchange Point is where my traffic and his <br clear="none">internet traffic can meet other.”<br clear="none"><br clear="none">this sentence is incorrect:<br clear="none">“for traffic which is supposed to be managed by the Mauritius Internet <br clear="none">Exchange Point”<br clear="none"><br clear="none">…it’s likely that you haven’t worked for any length of time at a <br clear="none">network operator, or any network that participates in internet peering, <br clear="none">so let me help you understand how this all ties in together.<br clear="none"><br clear="none">an internet exchange point IXP is - by almost anyone’s definition - a <br clear="none">neutral meeting ground where network operators exchange traffic. the <br clear="none">IXP itself *does not* and *should not* participate in the traffic <br clear="none">exchange.<br clear="none">that is to say, that the IXP operator (let’s call it the MIXP <br clear="none">management, in this case) does nothing to the traffic, other than to <br clear="none">support a platform for it to be exchanged, as quickly, and cheaply, <br clear="none">across - in this case, they would be responsible for the availability of <br clear="none">the ethernet switches that are in use as the foundation of the MIXP (ie. <br clear="none"> is the peering switch up?)<br clear="none"><br clear="none">that’s it.<br clear="none"><br clear="none">the choice to *use* the IXP to move traffic between networks, is <br clear="none">fundamentally, a decision of the networks that choose to peer (or not!). <br clear="none"> a good IXP operator doesn’t enforce any sort of mandatory <br clear="none">multilateral peering policy, and in fact, there’s zero history in the <br clear="none">22years of IXPs, of this MMLP succeeding. a good IXP operator performs <br clear="none">introductions, and encourages networks to peer; but the choice is <br clear="none">really that of the network provider. then again, the onus is on *both* <br clear="none">operators to peer. if operator A advertises his network prefixes to <br clear="none">operator B, but operator B does not advertise B’s prefixes back to A, <br clear="none">we only have half of the solution. traffic would flow from B->A and <br clear="none">then A->outside world->B. (this is more common that you might imagine)<br clear="none"><br clear="none">i’m correcting you, because, i think it’s important that you <br clear="none">understand that it’s *not* the MIXP that’s playing tricks with your <br clear="none">traffic, and *not* the MIXP that needs to take action here; indeed, you <br clear="none">did the right thing: complain to your ISP. and if they don’t fix it, <br clear="none">complain louder, and then vote with your feet and wallet! and do the <br clear="none">same at your workplace. and tell your friends…<br clear="none">in your case, there was some improvement made; clearly you feel <br clear="none">there’s more that could be done, and i’m sure with the help of <br clear="none">“man traceroute” you can guide your ISP further :-)<br clear="none"><br clear="none">so this sentence “I sincerely hope that the Mauritius Internet <br clear="none">Exchange Point fixes the latency issue.” … is quite incorrect; what <br clear="none">you mean is: “i sincerely hope that the two networks resolve their <br clear="none">peering problems”. you might even find that they don’t use the <br clear="none">MIXP at all, but choose to peer elsewhere in-country (and that’s ok!)<br clear="none"><br clear="none">perhaps, you should be asking instead:<br clear="none">- what can the MIXP do to promote peering, and teach networks about why <br clear="none">this is important? i think that’s a reasonable request, and one that <br clear="none">the MIXP management team should be able to do.<br clear="none">- what tools can the MIXP use to make it easier for peering participants <br clear="none">to get started peering; also, reasonable, imho.<br clear="none">- how we can, as the open-source, non-tech affiliated community help?<br clear="none">…etc.<br clear="none"><br clear="none">i hope you understand a little better, about how truly non-intrusive, an <br clear="none">IXP is.<br clear="none"><br clear="none">best,<div class="yqt1942152830" id="yqtfd21729"><br clear="none">—n.<br clear="none"></div><br><br></div> </div> </div> </div></div></body></html>