What would you say if I told you that you could get a faster transfer rate and lower latency from a provider with a single 100 Megabit uplink than a provider with multiple GigE’s?

It sounds broken; each GigE is 1,000 Megabits, how could a single 100 Megabit link be faster? It all comes down to a term referred to in the hosting industry as utilization.

Utilization translates to how ‘hot’ each of a provider’s links are running. Not just uplinks but internal links between routers and switches as well. The very best way to make this immediately apparent is the 10 Gig - 14.4 Modem scenario.

If you were to take a 10 Gig Ethernet (~10,000 Megabits!) link, plug it in to a router, then use a 14.4 modem to dial in to that router and receive internet access, you would still only run at 14.4 speeds. Common sense, right?

That said, if you were to connect a  DS-3 (~45 Megabits) link to a router and then connect a 3 Megabit DSL line to it for internet access, you would be running at 3 Megabits, much faster than a 14.4 modem.

When spelled out with these figures, sure, its obvious. But lets think about this on a larger scale.

Back to Provider A and Provider B from above. Let me add some important information to the comparison of these two providers:

Provider A:

• Uplink: 100 Megabit
• Current Utilization: 5 Megabit

Provider B:

• Uplink: 2 x GigE
• Current Utilization: 1,997 Megabit

Provider B is running ‘hot’ here, as they are only 3 Megabits from filling their uplinks to capacity. What this means is, regardless of their total amount of bandwidth, only 3 Megabit is still available for your use. They could hand you a gigabit uplink to your server, but the fastest speed you will ever receive is 3 Megabit; that’s all that is available.

Provider A is running VERY cold in this example, with only 5 Megabits of usage on a full 100 meg Ethernet uplink. This means you could pull up to 95 Megabits of bandwidth before hitting any bottlenecks, obviously a much better choice.

This all might seem obvious, but frankly it is quite common in the hosting industry. Some providers feel that so long as customers don’t complain, there’s no reason to add additional bandwidth. Some don’t even add bandwidth after complaints! As soon as the provider adds bandwidth, costs are increased. Now the money coming in from customers no longer covers the bandwidth costs.

So, next time you see 40 GIGABITS OF CAPACITY in marketing speak, make sure you check to see how much of that ‘capacity’ is in use

The Adtaq network is now powered by a Cisco Systems Catalyst 6509 Carrier Switch, equipped with a Supervisor 720 3BXL. This switch has 9 slots available for modular upgrades with 3 presently occupied; the Supervisor itself, a 16 port GBIC blade, and a 48 port 10/100 blade.

The device supports 10 Gig blades as well, allowing us to scale well beyond our current size without skipping a beat (bit?).

The scalability of this device is staggering. This is the same configuration which companies such as ATT, Verizon and Qwest use to deliver their bandwidth to customers; we are certainly in good company.

Customers, enjoy the continued high performance! Everyone else, what are you waiting for?!

I started this company barely knowing what a router was. Heck, I didn’t need to… I sold web hosting!

The company grew and my market evolved. People wanted to host entire machines, not just web-sites. I knew a switch took one uplink and turned it in to many, so, there we go, problem solved!

With a nice little (very little) collection of customers, imagine my surprise when for no reason whatsoever my customers were unable to access their sites. I ran to the datacenter, fearing the worst… Cabinet is on fire, rabid dogs overtook Westin security and chewed through power cables, someone opened the window because it was hot, etc. When I arrived, everything was fine. The servers were up, my switch was up; why is this happening?

The answer of course is that my upstream provider had failed. Not only had they failed, but apparently their lead network engineer had just quit and he was the only one with the enable (think ‘Administrator’ or ‘root’) password to their routers, so, I was dead in the water until they tracked him down and asked him if he could pretty please provide the enable password so they could actually log in to fix the issue.

I remember trying to think of a way to communicate this to my customers without sounding like an idiot. I also remember not coming up with one. It was that day that I vowed this would never happen again; time to learn what a router is.

While we were certainly a small company at the time, we did have an immediate need to remedy this situation. Budget did not allow for simply ‘throwing money at the problem’ so we had to get resourceful. After much searching we ended up with two Cisco 3745 routers coupled to two Cisco 3750 switches in a redundant, fully meshed configuration. Once that was in place, you can bet that the very next thing we did was to procure another provider so that an outage like this would never happen again.

Well, fast forward to today… To this point, the old 3745s and 3750s have run things famously; powering two datacenters worth of equipment, BGP sessions to large customers, peering sessions with just about everyone in the Seattle Internet Exchange, and months (actually over a year, I think) of consistent stability.

I will be forever proud of what we accomplished with this gear, but there always comes a time to let go.

I have a sort of ‘unspoken’ methodology I follow both in life and in business; if you aren’t going for the best, then you are just wasting <time|money|effort> that would be better spent just getting what you really want in the first place.

Yes, it would be ‘cheaper’ for Adtaq to run Netgear switches… Customers might not notice, everything might be fine… But, would it be ‘the best’? When they overheat and lock-up, would it still have been a good decision? When the lack of VLANs create broadcast traffic that drives customer’s bandwidth bills up artificially higher than they should be, would it still look like a good decision?

We waited until the time was right and then went all out.

Tonight, March 7th, 2008, marks our network’s graduation.

Stay tuned for details.

Adtaq has operated from Seattle’s Westin Building since 2003 and will continue to maintain a presence in this building for its core routing equipment and large bandwidth customers. “The Issaquah location will be much better suited for the majority of our customers,” said Tony Faoro, Adtaq’s Founder. “The connectivity is the same, the facility is just as nice as our Westin location, and the price is much lower. For us, it’s a no brainer as we aim to provide the best service to our customers at the lowest price.”

A colocation datacenter is designed to keep web and email servers up and running at all times. Computer room air-conditioners run 24 hours a day to address the heat today’s high tech computer systems produce while UPS battery backup systems and diesel generators ensure that power is always available to the facility.

For more information visit <a href=”http://www.adtaq.com”>www.adtaq.com</a> or call 1-866-591-COLO.

For more on Hurricane Electric, visit http://www.he.net.

This additional connectivity from Mzima raises Adtaq’s network capacity to 4 gigabits! Adtaq is continually evaluating network conditions to ensure your application will consistently deliver content to end users at the greatest speeds possible.

For more information on Mzima Networks, please visit http://www.mzima.net.