A Managed Hosting Contrarian

Let's start telling it like it is! (Unlikely.)

by Alex Goldman ISP-Planet Managing Editor [March 20, 2008]

The team at New York City-based managed hosting provider Voxel recently published a broadside called "Top 5 Lies in the Managed Hosting Industry."

The lies are:

1. "Our data center does not experience downtime ever. Everything is redundant."
   
   
2. "We have unlimited bandwidth. You don't have any limits."
   
   
3. "More hardware will solve the problem."
   
   
4. "Everything is covered in your managed service."
   
   
5. "We are a Tier 1 bandwidth provider (or use only Tier 1 bandwidth) and that's better."

ISP-Planet is the voice of the independent ISP, the subversive competitor taking on the establishment., so I've got to make the call.

I reach Raj Dutt, Voxel CEO. First question: why this press release? "Our PR group, Group SJR, has been coming up with different top five lists. When they came up with the 'Top 5 Lies in Managed Hosting,' I liked it. 'Managed hosting' is such a murky term. There's no standard definition, so 'buyer beware.' It can mean so little and so much."

No downtime? No way!
Take, for example lie number one, no downtime. Any individual data center, Dutt says, will experience downtime at some point, and it doesn't require a natural disaster that makes the national news (though that will often take down a data center). Someone can flick the wrong switch and take down a data center. Or you can have simultaneous failures in your main and redundant power.

Cases in point: Rackspace in 2007 (and Rackspace is one of the best, Dutt says) and 365 Main earlier last year.

One reaction, "please remember that data centers are single points of failure too." Another, also from the O'Reilly Radar blog, "And ironically enough, this morning 365 Main (together with Red Envelope) put out a press release announcing 2 years of 100% uptime; one may also note that they have now removed the press release from their site, as http://www.365main.com/press_releases/
pr_7_24_07_red_envelope.html returns a file not found."

Often, it's human error, but not always. "We saw terror first hand in New York City," says Dutt. "We were in 25 Broadway, across the street from the World Trade Center. It was a very nice facility, Telehouse does a good job of running it, but nobody could plan for the air intakes getting clogged with debris."

One solution is geographic redundancy. If you've got servers in three or more data centers, and you've got 1 1/3 the capacity you need in each, then you're covered if one goes down. "Rather than have a disaster recovery (DR) site for takeover if the primary site fails, we advocate an active-active setup," says Dutt. "It's not healthy for the industry to cling to concepts that don't have merit and make claims that cannot be achieved. When someone does build a data center that will never go down, great! But for now, it's the top lie!"

Unlimited bandwidth? Untrue!
"Capacity limits are real," says Dutt. "we deal with them every day, especially in the Content Delivery Network (CDN) part of the business. The industry needs to be more up front about what's happening. All of a sudden, we have customers who actually want to use all of their bandwidth. It's not just Bit Torrent. A few days ago there was a video event for Opera Winfrey that reached 220 Gbps peak traffic. That's what's happening! The numbers are shocking. Only two or three companies in the world have the capability to deliver that. Opera could have gone to any provider claiming unlimited bandwidth, but had to go to Akamai or Limelight."

More hardware? Same problem!
"Even we've been perhaps guilty of this," admits Dutt. "If the app is not scaling, it's easy to say, 'you have ten web servers, why not add ten more?' But that's disingenuous. The industry should admit that it's treating the symptom, not the cause. Hosting providers that are doing this lack either the expertise or the desire to profile the app."

It's not easy, but webhosts can help. "There's a lot you can do, tweaking the systems and the services layer. Say you've got lots of queries on the database that are taking too long. You can optimize indexes, cache queries—and doing so provides you the opportunity to be more visible, demonstrate your expertise, and add value to the customers."

It's a change everyone has to make. Adapt or die. "If you believe you're just in the business of moving boxes, here's your wake up call: nobody has the scale to compete with Google or Amazon S3. Managed hosting is more than making sure that boxes are pingable. Your customer should not need an administrator and should be able to run their business with minimal developer resources."

Everything is covered? Not!
"This is about the difference between managed hosting and proactive managed hosting. You can hire people with all the certifications [list of certifications omitted here], answer the phones when problems, fix problems when they happen—all of this is good, but it is not proactive. Proactive managed hosts will notice a web server is running out of memory and do application level monitoring. Reactive hosts will cover everything—after it breaks."

Tier 1? So what!
"Tier 1 is about politics and size," says Dutt. "I've seen the threads on ISP-Bandwidth. Tier 1 providers have size and scope, but the idea that a tier 1 provider will give you better performance or support is not true. Good luck getting more bandwidth from AT&T for your app with only one week notice! The attitude is, 'we don't have to care. We're the phone company.' We had a customer who needed a 10 GigE port within a day or two. No tier 1 provider will do that. We had them provisioned that afternoon and Equinix completed the cross connect the next day."

Isn't it also true that the smaller providers are more local. "In many cases, the tier 2 providers are regional. We're that way in New York, with our dark fiber metro rings."

—End