isabel wang

Intermedia vs Google

The newly launched Google Apps for Domains had everyone talking yesterday; that seemed to worry Intermedia quite a bit. It sent out a mock press release highlighting Google's "24/0" tech support.

Web.com's Jeff Stiebel and HostMySite's Lou Honick made similar points during their HostingCon panel a couple of months ago. Consumers should host with their companies rather than Google/Microsoft/Amazon (the "galactic clusters", as InfoWorld's Jon Udell puts it), they said, because they've got helpful staff who are there to pick up the phone.

I think Google/Microsoft/Amazon will change the web hosting landscape in two ways.

First, they'll raise the bar on reliability. The galactic clusters have rendered hardware failure irrelevant. Their customers are protected against downtime and data loss if any particular piece of equipment dies; they've got every application and every user's uploaded files elsewhere.

To stay in business, web hosting providers will have to keep up. Last Friday I got a super awesome demo from the 3tera folks on their grid operation system. They said they've given the presentation 500 times in the past couple of months; I'm not surprised.

Second, Google/Microsoft/Amazon's do-it-yourself business models will push hosting firms towards service-based differentiation. Intermedia/Web.com/HostMySite are right; 24/7 support will appeal to a significant proportion of customers. But instead of waiting in line to hear entry level techs read from a script, they'll want personalized assistance from a pro. Rackspace's service level will become industry standard, because unless a customer gets expertise he's willing to pay for, he'd probably be better off with the galactic clusters' economy of sale.

Because of these two factors, we are at the beginning of the end of the era of unmanaged dedicated servers. Very soon, a critical mass of users will realize that it makes no sense to rely on one single piece of aging hardware to keep their applications online and their data safe. They will sign up with Amazon EC2, if they're on a tight budget and/or have in-house expertise. Or they will leverage Rackspace's knowledge. They will have a harder and harder time finding good reasons to choose less reliable solutions in between.

Second Thoughts

I am re-thinking my comparison of Amazon EC2 versus low-cost dedicated servers. Many hosting providers offer $99 servers with 80 GB of storage space and 1000 GB of monthly bandwidth. I wrote last week that the same combination would cost $284 at Amazon. But my calculation was off.

The $284 price tag assumes that (a) you need always-on processing power, and (b) you are fully utilizing your disk space and bandwidth allocations. Relatively few server users meet these criteria.

For instance, if you're a developer working on a not-yet-live project, your resource usage would be minimal. More interestingly, you wouldn't need 24/7 CPU access. You could stop paying for your Amazon EC2 server instance whenever you're away from your desk. At the end of the month, you might get a $30 hosting bill - for 250 Mbps of data transfer capacity, dedicated processing power and fault-tolerant data storage.

Once your site is up and running, scalability becomes a major concern. You could lease multiple dedicated servers to anticipate traffic spikes; some data centers charge as much as $1 per GB for bandwidth usage beyond your monthly allocation. Or you could deploy additional EC2 server instances as needed and pay as you go for bandwidth.

By unbundling processing power, storage space and data transfer capacity, Amazon gives you full control over every dimension of your computing requirements, instead of forcing you to construct solutions out of inflexible building blocks. And since bandwidth usage is billed a la carte rather than in over-sold bundles, you're not stuck paying for capacity that you might not use.

So I was wrong about Amazon's competitive position relative to discount dedicated servers. It's a viable contender for many customers' business.

Of course, one group that EC2 would not appeal to is high traffic users with predictable bandwidth requirements. Why pay Amazon $0.20 when a $99 server offers $0.10/GB bandwidth - storage space and processing power included?

I wonder how much longer deals like this will last though. If pay-per-use catches on among low-traffic users, how many hosting providers will be able to sell less-than-half-price bandwidth compared with Amazon without these customers' unintended subsidy?

Amazon vs Rackspace?

Jeremy Wright was about to move the b5media blog network from WebNX to Rackspace, but after reading the Amazon EC2 news on TechCrunch, here's what he had to say: Holy @#$@.

My sentiments exactly.

EC2 stands for Elastic Compute Cloud. It's a pay-as-you-go utility computing service where you create Xen-based server instances on demand. Each instance is equivalent to a physical server with 1.7 GHz Xeon CPU, 1.75 GB RAM, 160 GB of local disk, and 250 Mbps of connectivity. You even get root access. As Amazon puts it:

EC2 enables you to increase or decrease capacity within minutes, not hours or days. You can commission one, hundreds or even thousands of server instances simultaneously. Of course, because this is all controlled with web service APIs, your application can automatically scale itself up and down depending on its needs.

In terms of pricing, Amazon charges $0.10 per instance-hour - or $72/month for an always-on server. Bandwidth ($0.20/GB) and storage space ($0.15/GB) are extra. Data transfer within the EC2 environment, or between EC2 instances and Amazon's S3 storage service, is free.

EC2 changes the web hosting landscape in three ways:

1. Earlier this week, my friend Dan wondered if it's possible to run a web hosting business without overselling. Amazon's answer is yes. EC2 gives you exactly what you pay for; no more, no less.

2. For many years now, 30 days has been the industry-standard server lease term. EC2 is 720 times more flexible.

3. Many web hosting providers offer hardware replacement SLAs, not to mention a wide range of backup-related products and services. EC2 renders all of these features moot; any data you upload is automatically distributed across multiple servers at multiple locations. You wouldn't even know about downtime on any particular piece of equipment.

So what does this mean?

If you're currently hosted on a low-end box and have predictable bandwidth and storage requirements, EC2 wouldn't make economic sense. Many data centers offer $99 servers with 80 GB storage and 1000 GB bandwidth. At Amazon, this combination would cost $284. The $185 differential does eliminate hardware failure risks, but that's probably too steep of an insurance premium for most to pay. For the time being, I think, discount dedicated server providers are safe from Amazon. (Having second thoughts about that.)

But if you're a developer looking to build the next big thing, EC2 is perfect. You could create and delete sandboxes at whim. You could stress test your application by flooding it with traffic for an afternoon. You could find out what happens if you scaled it across a 100-server cluster, then dismantle the whole thing immediately. You could support high-volume downloads night and day - forget about commit rates and 95th percentile billing.

Like Jeremy at b5, you might have seen Rackspace as an attractive hosting destination. But now you've got another option. As Jeremy puts it:

We were about to move to Rackspace mainly for uptime support. I’d imagine this would just never go down as well (in spite of Amazon.com’s recent downtime, S3 stayedup). We provision servers at 10GB of storage and 500GB of bandwidth, so we’d only pay about 180$/server, give or take.

I think Rackspace is great. I loved hosting my site there. But in this battle of Fanatical Support versus extreme scalability, I'm honestly not sure which will win.

PS- Jon Udell from InfoWorld has a demo on his blog. Wow was his reaction.

PPS- There was some speculation on GigaOm that Google and Microsoft might follow Amazon into the hosting game. And why wouldn't they?

PPPS- Just last night, I was reading this post from Tuesday by Alex Iskold on the 6 building blocks of web-computing platforms. The first 5, according to Alex, are fully formed. They include:

• Storage (Amazon's S3)
• Inter-application messaging (Amazon's Simple Queue Service)
• Information services (Amazon's product information; also, Google Maps API)
• Search (Amazon's Alexa Search Platform; also, Google Search API)
• Web services (Flickr and del.icio.us APIs)

Building block #6 is computing. As of  Tuesday, no black-box compute services were available on the web via API. Alex thought this was a challenging project that might take another few years. As it turns out, EC2 was announced less than 48 hours later.

The 6 building blocks, Alex said, can be leveraged by "anyone who's paying attention" to roll out complex, intelligent applications in a very compressed time frame. The question is, will the web hosting industry be able to keep up with this crazy pace? Or will Amazon (notice how it owns building blocks in amost every category?) become tomorrow's premier hosting destination?

Over the Virtual Garden Fence

Wen said my picture has great composition, and Cia complimented my cat Jacob on his bright green eyes. They are two of my growing number of Flickr friends. It's such a nice feeling, to get flattering comments on just-uploaded photos.

I've started looking out for new additions to my new friends' albums as well. Nia recently took some way cool abstract photos of I can't figure out what, and Adlin rode a very colorful Ferris wheel at the Jersey shore.

So I agree with Andreas Gauger, 1&1's CEO, that consumers love chatting with their cyber neighbors over the virtual garden fence. 1&1's recent study showed that men like to publish hobby material online, while women prefer sharing  personal memorabilia.

On the other hand, unlike Andreas, I think there's a BIG difference between having a web presence and maintaining a web hosting account. As a matter of fact, when I put up photos on my GoDaddy-hosted website some time ago, it didn't result in the kind of interactivity I now enjoy through Flickr.

In my mind, consumers need web presence enablers (which include traditional web hosting providers as well as MySpace-type social networks and Flickr/YouTube-type content-sharing services) to deliver the following four value propositions:

1. Infrastructure (storage space and bandwidth for putting data online)
2. Applications (software that facilitates data collection/organization/presentation)
3. Service (advice/assistance from sales/support staff when needed)
4. Interactivity (an audience for the consumer's content)

But whereas Flickr and other Web 2.0 services focus on #4 (which consumers are most likely to appreciate), traditional web hosting providers have persisted in marketing their services on the basis of #1 (which consumers are least likely to understand). This just doesn't make any sense to me.

1&1, GoDaddy and dozens of other hosting providers each serve enormous numbers of site owners. Instead of one-upping competitors with lower prices and larger bandwidth allocations, why hasn't anyone considered building a platform on which customers can interact, share hobbies/memorabilia, and perhaps even do business with other another? That's what I'd like to know...

It's Too Much Trouble

I just saw the Illusionist, which I liked a lot. On my way home, I thought of Robert Young's recent post on how movie studios and theater chains could encourage consumers to watch more films.

Whenever anyone goes to a movie theatre, they end up with a ticket stub. I would propose that you create a program/system to make those ticket stubs into currency… Let’s assume I went to see The Pirates of the Caribbean on opening day. I leave with a ticket stub that has been date/time-stamped. Then sometime during the following week, I run into my friend Sarah at Starbucks... I pull out my wallet and give her my ticket stub. Sarah decides to go see the movie the following weekend, and since exactly one week has passed since my ticket stub was stamped, she is entitled to receive a $1.00 discount. Now, had she waited two weeks to see the movie, she would have received a $2.00 discount.

I thought this was a really cool idea, but now I'm realizing it's not feasible at all. I don't have my ticket stub any more; I probably lost it even before getting to my seat. While I'd recommend the movie highly, I'm no good with keeing track of tiny slips of paper.

So instead, maybe theater owners should involve consumers in their word of mouth marketing right after the movie ends. Ask viewers to turn their cell phones back on and text message a promotional code to any potentially interested friends. The code would give recipients a small discount on tickets for the same movie. The monetary incentive, I think, is less important than personal recommendation from someone who's just seen the film.

Going one step further, maybe viewers could also be encouraged to join a referral network. Promotional codes that participants send out would be tagged with their user IDs. Each time someone redeems a member's promo code, the member would receive a small discount on his next ticket purchase. And more interestingly, consumers would be able to track movie recommendations they've given and received.

Does this sound like it'd work? My neighbor Irakli says that being a good salesperson is all about knowing when to ask for the business. Maybe the same applies to product promotion - being an effective marketer might require knowing when (and how) to ask for referrals.

Ordinary Hosting Customers

I came up with a great idea yesterday. I think every management-team member at every dedicated server provider should be given one of his or her company's lowest end machines - along with the responsibility of securing/configuring/maintaining it. Every time any server malfunctions in any way - whether from hardware failure, software glitches, networking issues, etc - its owner would pay a substantial penalty fee.

Why? Because it'd make for a compelling marketing campaign: if your CEO is willing to bet cold, hard cash from his own wallet on your equipment's reliability, wouldn't your sales team have an easier job convincing prospective customers to have faith?

It'd also help you develop a dead-on-relevant product strategy. What better way for your product managers to learn about customers' preferences than by living their daily experience?

You could even earn your company some good karma by donating all the penalty fees from your staff to charity.

Since I started this blog, I've been told many, many times that my thinking is not entirely consistent with reality. It's cool that Microsoft and Google run their web apps on super enormous clusters, but individual dedicated servers work just fine for ordinary hosting customers.

The thing is, unlike most of the folks I've heard this from, I've actually been an ordinary hosting customer. A long time ago, I ran a pay-per-click advertising directory. I bought incoming traffic at certain prices, with the hope that visitors would produce enough revenue generating clicks to cover my advertising and hosting costs - not to mention next week's payroll. The system worked, unless there was any kind of downtime, in which case I worried about our ability to keep the office lights on.

Given my experience, I have to say that if I were to start another web-based business, I wouldn't consider any hosting solution that doesn't offer beyond the box redundancy. Fortunately, there's an increasing number of options to choose from: TextDrive's Container Hosting, Rackspace's Mosso, ServePath's UtilityServe, just to name a few. I think of Concentric's clustered hosting every time I see all those Hello Kitty products at the mall (Sanrio is hosted on that platform). I doubt I'd be able to afford DataReturn's InfiniStructure, but that 200-core Opteron processing array sure does look awesome.

So I'm thinking if more execs had personal funds riding on a single piece of aging hardware, maybe there'd be an industry-wide re-evaluation of what works for ordinary hosting customers. Some of those 4-year-old Celerons in your bargain basement will inevitably go bust in the coming year. You can say this is a risk consumers just have to accept if they're looking for cheap servers. Or you can make a cluster out of this gear and deliver a more robust solution at an even lower price. But don't choose now. Hand out those Celerons at your next meeting, and let your team decide in a few months.

Inefficiency is Not Beautiful

I've been wondering whether there's an opportunity for taking advantage of the way web hosting providers oversell bandwidth. A $100 dedicated server often includes a 1000+ GB/month allowance. In contrast, GNAX, a colocation provider, charges the same amount for a 1Mbps commit rate, which supports 316.4 GB/month of data transfer in each direction (upload and download).

GNAX's pricing does not include server hardware, software licenses, cabinet space, power or tech support.

The beauty of the dedicated servers model, I've been told, is that high traffic customers (I'll call them "H") are subsidized by those with more modest bandwidth requirements ("L"). This allows web hosting providers to win H's business by advertising low prices, while relying on L's under-consumption to maintain a profitable level of average usage.

It's good to be H - unless their bandwidth needs exceed the 1000 GB limit. In such cases, they are subject to excess usage fees of $0.50+ per GB, or offered the perplexing option of upgrading their equipment (to a more expensive server, or a multi-server system; both would increase bandwidth allocation) - even if there's absolutely no need for additional processing power or storage space.

So here's what I'm thinking: wouldn't it make sense for H and L to set up a bandwidth coop? Rather than allowing web hosting providers to profit from their under-usage, L could sell their leftover bandwidth to H at rock bottom prices. This produces cost savings for both sets of customers - while cornering web hosts into the impossible position of having to honor bandwidth offers to the letter.

There are of course challenges associated with setting up such a system, but where there's a will, there's a way. Or as a Chinese saying goes, rule benders tend to make 3x more progress than rule makers. The demand for affordable bandwidth can only rise over time. Some day soon, someone will come up with a way to leverage L's unused bandwidth. At which point, the dedicated servers model will no longer be quite as beautiful.

The World is Moving So Fast Now

On July 26, Jon Udell talks about the rise of "open infrastructure", or P2P file storage and distribution. Within 3 weeks, TechCrunch says that AllPeers will soon launch its Firefox extension for P2P file transfers (holy cow! 347 comments on that post!), and GigaOm reports that Pando's drag and drop file sharing app has reached 1 million installs.

On July 29, the New York Times writes that voice over WiFi is hot. Scarcely two weeks later, Business Week announces that T-mobile will soon give customers the ability to make both cellular and WiFi calls.

Just last night, I thought 3tera's AppLogic grid operating system sounded new and exciting. 24 hours later, I learn from Bert Armijo, 3tera's VP of Sales and Marketing, that ServePath has already used its software to roll out an utility computing service.

I feel like the world is moving sooo fast these days. But maybe I'm just very slow at keeping up with cool stuff? Note to self: must read more.

In other news, I came across this post by Flickr founder Caterina Fake via Fred Wilson's blog.  Business development, they say, is now moving at a much faster pace as well, because instead of long negotiations,  partnerships can be forged through companies' access to each other's open APIs. (What I want to know is, how come Caterina has a Gmail rather than Yahoo! email address?)

Dedicated Servers are Like Answering Machines

A few weeks ago, I came across Nicolas Carr's blog post on how software kills hardware:

Consider the telephone answering machine. It began as a bulky analogue box running spools of tape. It turned into a small digital box, often incorporated into a phone. And finally it disappeared altogether, turning into pure software running out somewhere on a phone company's network. Once you bought an answering machine. Now you buy an answering service.

I've re-read it several times since; it's made me think that sooner or later, we're bound to reach the same point with web hosting.

Today, web hosting customers lease dedicated servers (or fractions thereof). As with answering machines, each of these boxes offer a finite amount of storage space. And each can only handle so many concurrent sessions. Tomorrow, users will expect on-demand hosting resources with seamless scalability. Heck, forget about tomorrow. Amazon.com's storage utility has been up and running for several months, and SmugMug CEO Don MacAskill calls it the Holy Grail.

(Fun fact: SmugMug's users have uploaded 500 million photos, which take up 300 TB of space. Don says that Amazon's pay-per-use rate of $0.15 per GB - redundancy included - has saved him buckets of cash.)

Anyway, through Vlad Miloushev's comment on Nick's post, I learned about 3tera. Its software promises to turn commodity servers into a shared grid on which applications can scale from a fraction of a server up to the whole grid. The requirements for geting started are minimal: a gigabit switch and two or more P4 servers with 1+ GHz processors. What data center doesn't have an idle stash of such gear??

I also followed Dan Ciruli comment ("hardware eventually breaks; software eventually works") to DigiPede, which makes distributed computing software for Windows.

Shouldn't more hosting providers be investigating solutions like these?

Because you can soup up a dedicated server with faster processors, bigger hard drives, more bandwidth; it'd still be tomorrow's answering machine.

You can slash prices; many customers would still end up paying for more hosting resources than they need.

You can install remote reboot switches, promise instant hardware replacement and offer automated OS reloads; these are supposed to be appealing features because they minimize downtime, but what site owners really, really want is a hosting infrastructure that automatically handles hardware failures without losing data or disrupting applications. And that's exactly what 3tera says its software does.

Of course, many answering machine manufacturers have continued selling their products long after software-based voicemail's mass adoption. Even now, you've got several options to choose from on Radioshack's website.

But this week everyone's talking about T-mobile's upcoming UMA service, which will support metered calls through its regular cellular network as well as flat-rate calls via WiFi. BusinessWeek, in particular, predicts that UMA will open the floodgates on landline-to-wireless migration: why would consumers pay for separate home phones when they can make high quality, unmetered calls through a device they already own? And once landlines go out of style, people wouldn't even have the option of using answering machines...

By the way, is UMA bad news for Vonage or what?

Could This Work for Web Hosting?

A couple of days ago, my friend Patrick emailed me this New York Times article.  Slide.com, it says, is among the pioneers of a new ecommerce trend. Up until now, online shopping has required visiting different vendors' websites to scope out items you'd like to buy. Slide.com wants to save us time by bringing product info right to our desktops.

You can sign up on Slide.com to see slideshows of shoes, jewelry, electronics... Within the next few weeks, the service will start pulling data from 50+ retailers, including Old Navy, CompUSA and Zappos. CEO Max Levchin (a PayPal founder) says that "well over half" of his user base have made a purchase. Slide.com gets a 5%-15% commission on each transaction.

Patrick LOVES this approach. He thinks it makes as much sense as using an RSS reader - versus visiting dozens of news sites and blogs individually. He says it's not just about efficiency, but also the comprehensiveness of information you receive. What shoe-obsessed girl wouldn't be psyched to have the latest fashions scrolling across her screen? Art collectors, too, might enjoy dynamically updated slideshows of new works from favorite galleries. Patrick himself subscribes to an RSS feed of Flickr photos tagged "Apple", if some of the shiny toys he sees came with a "click to order" link, they'd totally have a buyer.

So I'm wondering if push-based advertising might work for web hosting. Would prospective customers sign up for feeds of available servers from different data centers? Let's say I'm a developer planning to deploy my new site, or a sysadmin looking to expand my existing infrastructure - wouldn't it be convenient to turn server shopping into a background process? No more opening 7 different browser tabs to compare hardware types and bandwidth allocations; this tedious chore would be replaced with on-demand, at-a-glance access to all the relevant information.

Better yet, it'd be awesome if the system could incorporate reviews, link to recent Web Hosting Talk discussion threads, and offer Amazon-like recommendations based on aggregated data.

And last but not least, the feeds could be distributed via an affiliate program widget that participants could put on their websites, blogs, MySpace profiles, etc.

Doesn't that sound waaay cooler than the same old "showcases" that TheWHIR, Web Hosting Talk and TopHosts have all had since the beginning of time??

Update: Oh my goodness!! Check out this Techcrunch writeup on Dapper:

Dapper provides a point and click GUI to extract data from any web site that can then be worked with and displayed via XML, HTML, RSS, email alerts, Google Maps, Google Gadgets, a javascript image loop or JSON.

I was thinking about the business development challenges associated with procuring data for a web hosting services feed. But as a matter of fact, all the prices, specs etc can just be auto-extracted from different hosting companies' websites. That's way cool!