Today we’re pleased to announce the next chapter in our story as a scrappy infrastructure startup.
You see - it all started about 18 months ago with a single datacenter in northern New Jersey, a couple racks of Supermicro and Quanta servers, some Juniper routers, a fiber ring around NYC, and a core group of starry-eyed people who wanted to reinvent how developers could use bare metal servers.
A zillion lines of code, dozens of integrations, tens of thousands of installs and 2200 users later, we’re finally crossing off something that has been on our list since the beginning: opening a second facility. And since we were in the zone, we decided to open a third one at the same time!
On Why You’re Not a (Cool) Provider With Just One Location
Our vision was always to build a platform for production workloads, which means delivering packets to eyeballs that live all around the world.
While there are a lot of eyeballs in our home base of NYC metro (about 20.2 million pairs of them) we love the West Coast, Asia and Europe as well. Which is why it’s been somewhat awkward to talk to our super cool, target customers about our super cool, awesome platform and only have one pin on the map.
That being said, our focus has been to build for automation above all else, and filling out the services from our first facility (which, by the way, we’re insanely proud of) just made sense. But as soon as we could pull the trigger on an expanded footprint, we knew we had to do it.
From a strategic point of view, here is why:
- Geography is tough. If you don’t believe me, ask Neal Stephenson.
- The infrastructure business is dominated by huge providers.
- One location equals (kind of) a single point of failure.
- Who wants to hang out in New Jersey when you could be in CA or Amsterdam?
The Internet is Still a Series of Tubes (Latency)
The reasons for having locations around the world are many and varied, but the first for us is geographic proximity to end users. With our focus on bare metal and network performance, we want to be as close to those users as possible.
In selecting our 2nd and 3rd facilities, we looked at internet usage patterns, interconnection possibilities, asked our customers / prospects, and evaluated colocation opportunities from a multitude of angles. We were very confident about Amsterdam as our first European location (demand alone was a huge factor, but also it’s just one of those super connected, close-to-everything, logical markets) but less certain about our “west coast” option.
We looked at Dallas, various locations in the Bay Area, and the Seattle / Vancouver region. We even considered a multi-pronged strategy that would include a trio of 2nd tier markets (Chicago, Dallas and Vancouver). Ultimately, we just couldn’t say no to the siren song of Sunnyvale: with great access to the largest West Coast cities, as well as good connectivity to Asia, it proved the most compelling.
They Be Giants - All of Them
While we feel good about the splash we’ve made since our launch 8 months ago, from a market position Packet is still a niche player. Our competition for deals on a daily basis includes AWS, SoftLayer, Rackspace - and ‘smaller’ providers like Leaseweb and Internap that have substantial physical footprints, balance sheets, and sales teams.
To compete for the larger users ($50,000 - $500,000 per month in revenue) you really have to check off a lot of boxes. Great technology will only get you so far. You need multiple locations, a strong 24/7 support and facilities operation, and a substantial amount of inventory flexibility.
Part of building the case for Packet is product fit and market track record, but in the hosting world now you have to come with the physical goods as well. Case in point: a few weeks ago we started regularly seeing users deploy 150-250 physical servers at a time. Gulp. It's like word got out!
We expect even more of that this coming month, as we announce a series of partnerships whose products eat server clusters like they are a breakfast cereal. With our promise of "always available" having a diversity of locations, each stocked with strong inventory levels, gives us a powerful tool in serving these customers and never having to give an "out of stock" message.
One Location is One Too Few
We designed everything in our EWR1 facility to be highly available: out of the NYC floodplain, redundant Juniper routers and top of rack switches, bonded NICs, RAID SSD’s and dual power on everything all add up to an incredibly resilient stack from top to bottom. From the very beginning, even when we only had a few racks of servers, we implemented availability zone concepts to ensure customers who spun up tens of servers would get a variety of physical placements in the facility!
But even with all that, having only a single facility meant we still had a single point of failure and that simply doesn’t cut it. It's not likely, but a freak series of accidents or attacks could impact a single datacenter - as such we’re thrilled to expand our geographic options in service of more diversity for our clients.
Amsterdam, We’re Coming Your Way - Sunnyvale You Look Great Too
We pulled the trigger for these facilities just a few months ago, and our team has done a great job getting them up in time for the Spring conference season. It's hard to imagine that just three weeks ago we didn't even have gear in place, and today we're taking the wraps off two new locations - to say that I'm proud of our platform team is a huge understatement!
Which is why I'm so excited to head out and start spreading the word. We'll be in Berlin for CoreOS Fest (where we aim to chat up Trusted Compute and TPM chips!), Amsterdam for HashiconfEU, and then out to Portland for Monitorama. Hope to see you there!
Note: As a thank you for all clients who have waited so many months for us to go live in these markets, we’re offering servers in AMS1 and SJC1 at 50% the normal price as we go through our early access. Spin ‘em up!
UPDATE: Please read our KB article for more details about what our "beta" designation means for these sites.