BuyVM are a Canadian based hosting company with servers located in San Jose, California & Buffalo, New York. Francisco took some time to sit down with us & provide what is probably the most interesting interview we’ve done to date, grab a cup of coffee or a beer for this one ^_^
Tell us a little bit about how BuyVM started?
After our move to Fremont our Frantech brand was more or less sold out within a couple months. Workloads were pretty low since there wasn’t much to break with vmware.
At the time I was talking with the owner of 123Systems, Andrew, and he kept going on about lowendbox and how the budget market is thriving and a cash cow. He had the BuyVM domain spare (he tried to sell it but surprisingly no one wanted it) so I took it off his hands.
We put together some prices for what we thought was fair (given it was unmanaged) and posted things more so as a test of the market than anything else. Within our 5 days we sold well over 100 units which is pretty good for a brand that only started a couple weeks prior.
From there we got up to 5 nodes and then made the decision that our focus should change from Frantech to BuyVM as it was not only more profitable but a lot more enjoyable too.
What’s the fascination with ponies?
The whole pony thing started when Aldryic was watching Full Metal Panic Fumoffu! and the following scene came up:
He nearly fell off his chair cackling over it on skype. From there it was just something he’d randomly yell out in skype. Over time we all picked it up and turned it into a company meme where most things are named after it in some way just to continue the joke. I’m fairly sure some law enforcement group has looked up our ARIN entries and went “PonyNET? What kind of weirdo’s work here?”
What’s the culture like at BuyVM?
We’re a few white coats short of an insane asylum. Minus Anthony, the rest of us have known each other for over 6 years now so it’s nothing but ongoing jokes, youtube videos and horrifying company trips.
Anthony joined us about 6 months before BuyVM but since he’s short a few screws himself, he fit right in.
What are common things that people use your servers for?
A lot of users enjoy our services for IRC and such but most people run production webservers, VPN’s and the likes.
Being the first to bring IPV6 into the budget VPS market as well as the first to have IPV6 RDNS support made us almost a household name for IRC users.
You guys are notorious for selling out of stock quickly & have almost a cult following, what’s the secret?
There is no secret, we simply don’t buy enough equipment. Equipment is expensive at $3000/node in many cases. Since we don’t involve investors outside of ourselves we make sure we’re always in a comfortable spot before our next round of expansion.
While many will call me a liar, I can assure you our intentions aren’t to sell out in a split second
In the release of our Buffalo location we actually allowed pre-orders/early bird orders from current clients 3 days before any public release. We plan to keep doing this for any future locations or super popular plans we release.
As you’ve grown what’s been the hardest part of your business to scale?
It’d be ordering enough gear honestly. Our network is very flat with nothing but layer 2 switches and all layer 3 being handled by our routers.
We normally try to do a big round of gear once or twice a year.
Surprisingly scaling our support is pretty straight forward. If we find that there is something being requested more than a few times in ticket, we right away start looking into ways to either automate it, offer it as a template, or just start writing a WIKI article about it.
Your 128MB plan has the highest BearScore on ServerBear, a fine balance of performance with value. What things are most important to you when brining new products/plans to the market?
We want good value for the client while at the same time still turning a solid profit on our end. We have the resources, manpower, and experience that if we wanted to offer a super cheap 64MB plan we could, but we wouldn’t be happy with the amount of VM’s we’d have to stack on a node to turn the same profit margins as our other nodes.
Having a node sitting at 6,000 running processes is not OK.
In general, we have an internal figure we keep to ourselves of how much money we should be making per AMP of power used. If a business plan comes up and doesn’t fit that model we go over if it’s worth it for us to offer it or if there’s a way to improve it. If it doesn’t, we’re not going to bother.
What’s your favorite bit of hardware that you own?
That’d probably be our E5’s in New York. They’re really quick boxes and minus the issues we had with our kernel and some CPU options, they’re really solid.
My most favorite bit of hardware that I use on a daily basis would probably be our deployment at Awknet. It handles some stupidly powerful floods with minimal headaches.
You offer huge amounts of space on your storage plans, what’s the economics of running a node like that?
Assuming you’re putting a proper raid in there and not using basic lame-o LVM’s or something, it’s a fairly costly build. For us we’re having to put 2 x $400 raid cards in there, a large 16 drive chassis as well as a 1200W PSU just to power it all. On top of that there is all of the drives.
We skimped on our first storage node using just MDADM RAID6 over 12 drives but it performed poorly (40MB/sec) and wasn’t maintained very well. After enough tickets to Anthony we went back to the drawing board on the whole platform and came up with our current setup, which uses hardware raid50 as well as E3’s for CPU’s.
You have your own control panel – Stallion. How’d that come about & how difficult is it to roll your own panel like that?
When BuyVM started we used SolusVM like everyone else. Right off the bat we had major issues with it due to their lack of IPV6 support and many other things. We got around these issues using our own scripts to add proper IP’s, enable certain features and all around getting that BuyVM feel to our customers.
When they finally released IPV6 they didn’t have RDNS working properly. We logged a bugreport with them. They kept kicking it along and giving me bad deadlines and ignoring it. At one point they actually emailed me for help to debug why most of their clients couldn’t use IPv6 in general. I offered them a deal “Give me a fix for RDNS, or the source to the page, and I’ll show you how to get it to work”. They of course refused that
After over a year of waiting I said screw it and simply took it upon myself to code what I needed to complete the system myself in an afternoon. Someone on WHT decided to snitch to them about it, at which point SolusVM suspended our license claiming this and that with their EULA. After about 30 minutes they finally released the license and that was that.
Once they did that I held a meeting with the guys and we decided there was no way in hell we could continue working with a company that’s so drunk with power that they’d suspend us and we needed our own panel.
Stallion v1 took me a few months to crack out. It’s a lot of code to think through, especially when it comes to keep the heavy lifting parts (like bandwidth accounting) from ripping manage.buyvm.net offline while processing.
V2 is well underway and is a complete rewrite using codeigniter, bootstrap and all those newfangled frameworks. We hope to have it in live beta by January or February or 20XX.
You have some really knowledgeable folks in your IRC channel, how does the community element help with day to day operations?
Our community based support rivals that of many of the “fully managed VPS” companies that are out there. There are countless members, not just staff members, that enjoy helping others and tinkering with linux.
Many have helped fill our wiki while others are the first to help guide people if they get marked as fraud or have a question about stock releases.
Our channel plays a major role in why BuyVM is the way it is.
You offer DDOS protected IP addresses in San Jose. That’s a pretty unique offering, tell us more?
Our filtering was originally brought in to help protect our billing panel during new stock releases. We had a large stock release scheduled for a Monday in April. Within an hour of the release happening we got multi gbit UDP floods pegging it offline for almost an entire day. After some quick planning we had the platform ready for sale within 72 hours and our billing panel selling stock like usual.
The whole platform has been built in house by us using commodity hardware.
The platform is most effective in dealing with UDP, ICMP, and now spoofed SYN floods. The platform has taken a lot of blood, sweat, and downtime, to get it to where it is today. I honestly can’t think of another provider, though, that provides upwards of 10Gbit of UDP/ICMP filtering as well as up to 750,000 packets per second of spoofed SYN filtering for just $3.00/month
We’re working to bring layer 7 HTTP protection in for a very affordable additional cost. While we’d love to include it in our base plans, the amount of CPU processing required for it is simply too high.
Do you think that the Low End market is becoming too saturated?
I think the current LE market is going to collapse on itself soon if this ‘rush to the bottom’ continues. The current business models companies are using relies on them banking on you using, at most, ¼ of what you pay for or they’re hitting swap hard. Anyone that claims they aren’t overselling a 2GB at $7.00/month is full of it or is OK making $10.00/month profit per node.
My current predictions are that the market will have no choice but to correct itself (either with people selling out to each other or deadpooling) once IP prices jump in the US. Europe is already ramping up with multiple companies having to move out of Hetzner, LeaseWeb, and now OVH due to new IP restrictions and costs.
What’s your view on benchmarking in general, do you think it has its place to help hosts get exposure & win customers?
Benchmarking has its place but I think it’s being taken too literal in the current market. Customers are spending tons of time moving between different hosts just because one host might give 10MB/sec faster on a DD than theirs or the node one user on is mostly empty so it scores a little higher on a unixbench.
The fact of the matter is that these are lower spec VM’s. Users are using them for VPN’s or personal sites that get < 100,000 hits a month or less. Do you need 500MB/sec writes to complete that? Cripes I hope not.
I think customers should consider how well a company has placed itself for long term service. Anyone and their dog can order an E3 1230, throw an SSD on it, install SolusVM and ‘get into the market’.
I can tell you right now that we’ve simply refunded clients in the past when they would log tickets because their node would dip a bit during nightly crons, backups, etc, and were being nothing but troublesome. If you need 500MB/sec at any given time then you shouldn’t be using a VPS.
Benchmarks can help show which hosts really give a hoot about their product and which ones are trying to get money for Halo 4, but I think more important metrics are being ignored in the process.
If a magic pony came along & gave you a $1M investment tomorrow. How do you think you would use that to scale?
Gotta keep such secrets to myself, sorry
I can say that we’d have a lot of fun with it, though.
Any plans to expand further, even outside the states?
There’s always the idea of where to expand next and Europe gets brought up a lot. It’s a tough sell though with companies like OVH offering their shirt for a Euro or two.
What do you think lies ahead in the future for BuyVM?
Given how Matt acts in #frantech? Probably a sexual harassment lawsuit.
Kidding aside we’ll likely start growing into markets outside of the LE’s as well as continuingly improve our current offerings for new & veteran clients alike.
The guys on LET miss Aldryic, does he have plans to come back?
As they should, he’s an incredibly awesome guy to be best friends with. There are stories from our trip to Buffalo back In April that should have gotten us pulled over.
It’s extremely unlikely he’ll return though. Whatever need for entertainment he had we put an end to when we purchased him and his wife brand new desktops to game on as a late wedding gift. I actually set aside some of my wage every month to buy the guys/gals games that go on sale on Steam as random gifts.
He does hang around in #Frantech all the time though so people that miss him are welcome to take part in the community and chat with him there.
In closing I wanted to thank you for the invite to do an interview It’s very humbling that you found our quirky group interesting enough to want to interview. If I missed something or you want more detail I’ll try to provide it where possible.
ServerBear makes it easy to benchmark Linux servers & compare performance metrics (Disk IO, IOPS, FIO, Network Performance & UnixBench).
Don't settle for poor server performance from your current host, run a benchmark & instantly compare results against over 1000+ plans: