It’s been a while since my last post, and I’d like to take a few moments to bring everybody up to speed on some significant developments on the Voxel network.

First, I’m pleased to announce the expansion of the Voxel network to Seattle, WA. We are peering at the Seattle Internet Exchange (colloquially, the “SIX”), a not-for-profit organization committed to interconnecting organizations which run the gamut from small research networks and access providers to popular content providers and telecommunications giants. This is an important gain for us; it provides us strong reachability to viewer “eyeballs” in the northeast region and western Canada, long neglected by larger providers as a point of interconnection.

Not to overshadow this announcement, Voxel has also gone live with its Singapore (SIN1) facility, our first deployment in Asia. We are providing our full suite of services, including virtualized and dedicated server hosting and content delivery, out of this location.

Some notes from the field:

Why Singapore?

Raj, our CEO, posted to our blog last summer, explaining our reasons for opening an office in Singapore:

http://www.voxel.net/blog/2008/08/singapore-inc-here-we-come/

With an office in place, opening a datacenter in Asia was next on our plate.  Singapore appealed to us for a number of reasons:

1. It was “centrally” located, providing reasonable path latencies and transport options into Japan, Hong Kong, Malaysia, India, and mainland China, to name a few locations.
2. We were able to find a datacenter with plenty of available space and power, and reasonable connectivity options out.
3. Having an office in close proximity makes operations easy, even as we develop infrastructure and systems to keep our staff away as much as possible.

Facilities

We decided to go with Equinix as our facilities provider, occupying a large caged suite in their flagship “SG1″ facility. We searched long and hard, and found Equinix to be the best solution as a neutral facility in the country, with vendor and peering options blowing away the competition. At the risk of sounding like a press release, the Equinix Asia folk are true professionals, continually setting a high bar and proceeding to meet or exceed expectations. Thanks to the tireless efforts of their sales, project management, and network teams, we were able to turn up with no major snags, and following a fairly aggressive schedule. I’d highly recommend them to any organization looking to expand into the Asian market and unable to have all their needs met by Voxel directly.

(One anecdote here: you can e-mail Equinix’s Asia support center at any time, day or night, with odd requests such as “hey, could you please grab a laptop, console up this newly-arrived device, and give it an IP and local login?” and receive a call ten minutes later asking if you’re able to log in. Remote-hands organizations elsewhere would be well-served to follow this lead!)

Servers and Support

Since our early days, we’ve been strong believers in a “lights-out” server environment. We were among the first American server providers to provide customers with out-of-band serial console access as a standard feature, along with more-popular remote reboot capabilities.  Behind the scenes, we’ve invested a significant amount of human capital in tools development and OS install automation.

Our Singapore facility is no exception to this philosophy, and the benefits to date have been huge. Our Singapore systems and support staffers are able to spend the bulk of their time in the comfort of the office supporting customers, with only a weekly trip to the datacenter to swap out failed components and install new hardware.

Network

We deploy large masses of infrastructure in new and exotic locations on monthly basis with relative ease. Fortunately for us, there were no real surprises to speak of on the above. Perhaps the biggest eye-opener for us was the bandwidth and peering ecosystem in the Asian market, as compared to Europe and North America…

For starters, the climate for peering in Asia is marked by steep extremes. On one hand, many of the larger telecommunications providers are highly protective of their home market. Connecting their networks with other networks, locally, equates to enabling the competition, and they’ll have none of that. Bearing the cost burden of hauling traffic to Europe or the United States to peer is no problem, nor is sending down paid-for transit circuits in the United States (round-trips through Los Angeles to get across town in Singapore are not at all uncommon!), yet peering in Singapore is a big no-no. In the end, the customers lose.  On the other hand, we’ve got another class of access and hosting shops sick and tired of this mentality, and cognoscent of the goodness that is Settlement-Free Interconnection, or SFI for short. For connecting to them, our connection into the Equinix peering switch, in addition to private cross-connects in the  Singapore facilty, has proven extremely fruitful.

Purchasing IP transit, or paid-for bandwidth, has proven equally fun thus far. We enjoy vendor uniformity; unfortunately many of the top networks in Europe and America have Asian presences that are laughable at best; conversely, many providers one would seldom consider domestically have proper infrastructure and strong inter-provider capacity in their home markets. And “mixing and matching” transit providers between continents has required working additional mapping intelligence into our Content Delivery Network, to make sure end-users are always routed to the appropriate server farm.  Developing an inter-provider routing policy to ensure that traffic to Asian networks doesn’t leave the continent was a daunting task as well.

Another key challenge faced is the cost of IP transit in Singapore, which is far greater than in the United States and Europe.  We pride ourselves in maintaining transparency with our customers.  We like to say that “a bit is a bit” — whether traffic is served by your server or our CDN, irrespective of location, you can expect to pay the same price per unit of data served.  We’d like nothing more than to deliver an easy-to-understand bandwidth pricing model that’s consistent throughout our entire product set and and network footprint.

At the end of the day, I think we’ve succeeded in lining up the appropriate infrascture to provide our customers with the strong connectivity and attention to detail they’ve come to expect from Voxel.  We look forward to servicing new and existing customers in this exciting new market!

What’s next?  Lots!  We’ve just hit the ground running, and we’re already eying additional locations in Asia to extend the network to for purposes of peering.  Back at home, we’re evaluating vendors and designs for our “next generation” of server infrastructure (think: ridiculously dense 10 Gigabit Ethernet), and are on the cusp of rolling out native (”dual stack”) IPv6 connectivity to our customers.  More on the latter in a blog entry to follow next month!

Posted in General Posts | 1 Comment »

This entry was posted on Friday, June 26th, 2009 at 11:11 pm and is filed under General Posts. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.

One Response to “State of the (Voxel) Network Address”

1. Richard Says:
   August 21st, 2009 at 4:30 am

   Cool looking forward to playing with IP6 stuff!

Leave a Reply