Advise on MPLS Providor


I'm wondering if anyone can speak about their experiences with the following vendors. We are in the discovery phase for searching for a vendor to provide a private MPLS network supplying throttled service up to DS3 point-to-point for 2 offices. 3 much smaller offices would be connected to the cloud over T1.

We've been speaking to Verizon, Qwest and ATT.

We're interested in the managed router services.

Each providor tells us that the other vendors do not actually own their own private network. The also indicate that the other vendors can't possibly give us the SLA that they can provide. In some cases it's 38 ms for latency. We've heard 75 ms coast to coast is more reasonable.

We would maintain our existing internet connections, and we would not be using the providor's gateway for remote access to the network and web services, etc.

We would build this with the possibility of adding VoIP later, so we're interest in the COS as well.

Does anyone have specific experiences with these providors in an MPLS network configuration? What about their managed services? Anyone doing VoIP with these providors using COS? Any gotchas we should be looking out for?

Any thoughts would be helpful.



Reply to
Loading thread data ...

While I haven't shopped for that particular service from these providers, I think its funny that they each can't identify the other top-5 Tier-1 providers in their market. All of them own their own network, and can provide SLA at whatever level coast-to-coast on their network. (The other two I'd put in the top-5 are Level-3 and Sprint, couldn't give you an ordering though, although wouldn't surprise me to have Level-3 not own all its own fiber).

All of them are pretty decent overall.

The main downside I have with AT&T is with their ticketing system. It sucks bigtime. If there's a problem, its a pain to get through. Otherwise, their network has been reliable in the past, although I'm not using them now for IPv4 transit.

VZ is pretty solid all around.

The only problems I've had with Qwest wouldn't affect you for this. Mostly overloaded peering points (hard to say who's at fault for that, could very well be their peers) getting onto other networks. Otherwise Qwest has been pretty solid too.

Not sure if that helps you at all.

Reply to
Doug McIntyre

That does help. I've heard similar things about ATT's support system. At this point the decision might come down to "value added" options. We use Qwest for our internet connection. There's been no downtime, and I like their customer service.

Does anyone have specific experience with ATT, Verizon or Qwest as MPLS providers?

Reply to

I have some experience with ATT. I also were involved in a couple of RFP for MPLS networks. While I see a lot of time being spent on comparing SLA and trying to make it apples-to-apples, I'd make an argument that SLA doesn't really matter. Also VoIP is most likely to work just fine on MPLS from any of those carriers (as long as you order priority class of service).

What will affect performance of your new MPLS based WAN is traffic flow matrix and the way particular carrier does egress de-queuing. Since you mentioned that you are going to have 2 offices connected to MPLS via DS3 links and the rest via T1, my guess is you have regular hub-and-spokes setup. In that scenario egress de-queuing mechanism is most important thing - it will affect user experience in all your remote offices. Ones with DS3 links most likely will be fine.

That being said - ATT is using 4 FIFO queues on egress (could be 6 right now, they were talking about it). Don't be confused when they say they do CBWFQ - while those 4 queues are weighted and such, they are still FIFO. One is used for voice traffic, another two for latency sensitive data and last one for the rest. Problem - there is no WFQ inside of their best effort class. So huge data transfer between systems with good TCP stack has potential to kill everything else there. Two classes for latency sensitive work as advertised if you can figure out how to avoid oversubscribing them - not easy task since you have any-to-any connectivity. As a result - every time link utilization hit 80%+ latency goes up and every delay sensitive protocol suffers.

Regards, Andrey.

Reply to
Andrey Tarasov

Thanks ... This is very helpful information. Yes, we're straightforward hub and spoke. We won't be pushing voice traffic for a while, but we will be doing limited video conferencing. Also, the users will be oracle reads and writes from a fat client.

Reply to
sillz Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.