EMIX - are they letting down ISPs in the region?

This morning for about 14 hours, EMIX took it upon themselves to do the same thing to some of its other customers, including Qualitynet and KEMS in Kuwait. They both use the Etisalat owned EMIX backbone. There were a bunch of pretty upset users in Kuwait all morning.

A possible reason for this (purely speculation) could be the installation or reconfiguring of a system known as ACL (Cisco's Access Control List - a basic firewall type application) on their main router, which could have switched the routers from All Ports ON by default to All Ports OFF by default.

Doing this on an ISP level is common. As a Backbone provider, it shouldn't happen at all.

The Internet link from Kuwait to Dubai is run by EMIX, as is the one from Dubai to the USA.

For example, KEMS have a Satellite link directly from Kuwait to the US as a backup only, since the latency is over 1 second. They don't need EMIX for a relatively cumbersome satellite link, but for the Fiber-Optic STM-1 Link, which is faster, has larger bandwidth, and is more reliable.

EMIX - Latency problems affecting everyone.

Latency WITHIN Emix is often anything from 400 to 800ms - which is totally unacceptable.

US backbones don't exceed 100ms.

These levels are "shockingly bad." They have a FIBER backbone, not satellite.

Basically it points to terrible configuration in their routers

A typical US user from East Coast connecting to Microsoft.com at Redmond has the same number of hops as one has from the UAE. Their latency would be 120-140ms, or if it was a bad ISP, 170-180 ms.

So what is this "latency" we are talking about? It is basically a delay; the time taken within Emix routers for packets entering and exiting the router.

In one port ; processed ; then out the other.

What this means for users who rely on EMIX, which includes UAE and Kuwait, and probably many others in the region,  is:

1. They can't run any application that requires low latency. For instance, Net2Phone asks ISPs who host their gateways that MAXIMUM latency to Net2Phone's main server be under 250ms
2. The cannot run any multi-user, real-time games. those need latency below 500ms
3. Any voice application (including real-time broadcasts) will have poor quality, echoes, etc.
4. The same applies to video applications that also have sound.
5. Any real-time, or near-real time application will also have problems.

EMIX - who is to blame?

One might safely say they would be "protocol related" For example, if using ATM with Frame Relay, there are over 600 parameters to be set.

One wonders if the delays are so large due to the huge list of allow/deny/reroute rules in them? These things would slow down the system, but surely not this much?

One often sees signs of bad configuration in their system from evidence of loops in their network.

Perhaps there are a multitude of bad routes with lots of redirects happening between routers that we cannot see.

It is not easy to get the shortest path routines set up in a network that has (probably) evolved in an unplanned manner.

One would think that with Cisco systems in place, they would maintain the Routers. However Cisco only  provides technical support. They donot run things.

'We know everything'

So, even if Cisco were to tell them to
"do this" or to "do that"
one could well expect an arrogant reaction to do exactly the opposite.

Really, one need only talk to ISPs in the region to know what levels of abhorrence have been reached in their dealings with Emix.

If Emix has a problem, they refuse to acknowledge that the problem is theirs.

They will pass the buck.

They put the burden of proof on their customer ISPs, who then have to PROVE that it is an Emix problem before they will grudgingly take any action.

Emix almost always start by blaming their customers that it is their own setup, fault, etc. that is the problem.

Sound familiar?

EMIX - so what is the solution?

Etisalat's way of solving the problem.

A much better solution.