isabel wang

“And now,” cried Max, “let the wild rumpus start!”

Two REALLY important things happened this past week:

1. Microsoft unveiled its utility computing platform (Ray Ozzie's plans for it, at least) and it looks a lot like Amazon Web Services. Both offer not just hardware-as-a-service, but (as Simon Wardley puts it) frameworks-as-a-service.

Ozzie says Microsoft will provide "a fabric upon which online services run. Application frameworks that support a variety of app models designed for horizontal scaling. Infrastructure that manages performance optimization." And Alex Iskold calls S3, EC2 and SQS the foundation of Amazon's Web OS, which "hides complexity behind simple APIs; turns the concepts of search, storage, lookup and management of data into web services."

Bill Boebel made me realize why FaaS > HaaS. He said he's always looking to get more done faster (who isn't??). Last year he built a backup system atop S3: "We coded the storage client, not the storage server. Initially we had planned on building both. But when S3 came out our thoughts quickly shifted to 'Screw that, lets just build the client and get this thing released'." In this case he was able to leverage Amazon's Web OS. In other cases he's hired people to build stuff from scratch - but how often would he (or anyone else) choose to if it weren't necessary?

Amazon gets it. Microsoft too. I'm not sure infrastructure vendors who sell HaaS without FaaS will be able to compete. As for those who offer standalone servers (with non-interoperable storage media and device drivers!!) as units of computing capacity? All I can say is, yikes.

2. Amazon announced the ability for EC2 users to buy/sell machine images from one another. Amazon's base price is $0.10 per instance/hour. If you listed a $0.12/hour virtual appliance on Amazon's marketplace, its billing system will automatically give you a $0.02/hour credit any time the appliance is deployed.

The interesting thing about this "reseller 2.0" arrangement is, virtual appliance publishers will help EC2 reach a broader audience without reducing Amazon's profit margin. Or promising end users any more bandwidth/storage capacity than are actually available. Or hiding behind traditional reseller hosting smoke and mirrors. Instead, each will add real AND unique value to the customer experience. As Rurik Bradbury from Intermedia.NET puts it, being a traditional hosting reseller is like having fish, but the new era we're entering will require fishing skills.

The world is changing - and it's about time! Last week Takeshi Eto (now of DiscountASP.NET) and I were reminiscing about his HostSave days. Back in 1999, HostSave's $6.95 web hosting plan included 1GB of data transfer. Today, many sub-$10 hosting plans advertise 3000x more bandwidth - which EVERY provider acknowledges they can't really deliver! In the brave new Internet infrastructure world, it'll take more than empty promises to be the King of the Wild Things. And unlike Maurice Sendak's Max, we won't find supper waiting back where we came from.

Can hosting partners make any contribution to the Microsoft cloud?

I was telling Suren Singh (who is a Hosting Solutions Manager at Microsoft) that I'm not sure it'd be wise for a web hosting provider to consider MSFT its partner. Microsoft is investing zillions on its data centers, in which server count has doubled over the past year. It's already hosting Exchange and SharePoint for Energizer and plans to sell hosted services to others "vigorously and transparently". The reassurances it gave at its partner conference earlier this month were fuzzy at best...

In retrospect, my objection was poorly phrased. What I should have said was, I'm not sure how much sense it makes for Microsoft to have a hosting partner program.

The day before yesterday, Ray Ozzie outlined the three layers of Microsoft's software + services platform in his presentation to Wall Street analysts. These include:

1. Global Foundation Services: huge data centers in which "reliability is achieved through redundancy, not the fail-safe nature of any given component"

2. Cloud Infrastructure Services (which Nicholas Carr calls the Cloud OS): virtualized utility computing fabric with application framework that supports horizontal scaling and manages load balancing/performance optimization

3. Live Platform Services: data and features - such as identity management, contact lists, user presence and advertising - that are shared across multiple Microsoft and 3rd party apps

Now, in which of these layers can a commodity hoster make ANY contribution?

Ozzie went on to describe Microsoft's target audiences. Individuals, he said, will enjoy connected entertainment and productivity. Companies will be able to move their IT infrastructure into the cloud at the "lowest, lowest possible cost". And developers will have limitless access to computing and storage capacity at "very, very low cost".

(Quote from CNET: "Microsoft is trying to make sure that its business terms are attractive enough to woo the next MySpace or YouTube to bet on its technology. It has spent months talking to existing partners, but also to venture capital firms and start-ups. For now, Microsoft is offering up many of its services free for up to 1 million users, while saying it wants to strike some kind of deal if a service exceeds that threshold.")

Let's think - can traditional shared/VPS/dedicated hosting providers help deliver any of these benefits, considering their track record of fostering NO connections between customers and stranding user data on non-redundant, stand-alone servers?

Ozzie did acknowledge that some companies will want to maintain on-premise servers ("the ultimate in customization and control") and others will work with partners to “take advantage of unique vertical expertise or vertical solutions”. Unfortunately, I don't think such expertise/solutions are available from 99% of the folks Suren met at HostingCon.

Ozzie's vision reminded me a lot of Amazon Web Services, for which its evangelists have built a vibrant developer ecosystem. Why, in contrast, would Microsoft want to sign up hosting partners who are not at all equipped to advance its cause?

PS - Duncan Strong from Melbourne IT made a more compelling case for Microsoft's hosting partners program than any Microsoft employee I've met. He said since MSFT product managers have P&L responsibility, they generally gear development efforts toward distribution channels that deliver the highest ROI. Up until now, the hosting channel has done better than Xbox Live, let's say. But... software + services is sooo much farther from traditional hosters' core competency than Windows Server SPLA licenses.

IDC says use of dedicated servers to totally plummet

Earlier this week IDC analyst Chris Ingles *conservatively* estimated that within 4 years, only 50% of all physical servers will run as standalone machines (versus 93% today).

Other analyst predictions for 2011 include the virtualization services market reaching $11.7 billion (says IDC, up from $5.5 billion in 2006), and the global SaaS market reaching $19.3 billion (says Gartner, up from $6.3 billion in 2006). In contrast, the server hardware market will "struggle to maintain growth rates of more than 2% annually" (IDC).

Many dedicated servers providers will say the good news is, they already offer VPS or grid hosting and Hosted Exchange! But such offerings represent only a pin-sized tip of the virtualization/SaaS iceberg. And of course, every hosting company wants to "help customers do business online". But typical web-hosting-industry definitions of customer service (timely hardware replacement, prompt ticket resolution, etc) fall far short of what Salesforce.com, for instance, delivers.

Salesforce EMEA co-president Lindsey Armstrong's presentation at Data Centres Europe really opened my eyes as to what vendor/customer relationships should be like. She said her sales reps are held responsible for not just sign-ups, but making sure users are getting value out of their purchase. Are people logging in? How much time are they spending on the system? What features are they using? Because once software becomes shelfware, you can start looking forward to cancellations. (Which reminds me, I've got a couple of unused hosting accounts that I should cancel. Customers who don't have pending support tickets aren't necessarily happy customers!)

A dedicated server sales manager I recently spoke to gasped when I suggested he should find out what customers are doing with their gear. He can't afford to! There's not enough hours in a day! But I'm wondering if he can afford NOT to, with billions of dollars on the line?

The drag-and-drop, point-and-click metaverse

Christofer Hoff sounds like such a cool guy. His interests include:

"Anything with wheels and an engine that puts me in harms way - supercharged small blocks, motorcycles and 125cc shifter karts, especially. Kentucky Bourbon, Tennessee Whiskey and fine sipping rums. Cigars -- usually the kind that curl your toes and puts hair on your palms. Paintball, small caliber firearms, power tools, jumbalaya and BBQ'ing anything that at one time had a face and was not in the same genus/species fork as I am/was."

He's also the founder of three start-ups and he is currently Crossbeam Systems' Chief Security Strategist.

Barry from 3Tera forwarded this blog post, in which Christofer describes a "drag-and-drop, point-and-click Metaverse of virtualized application and data combined with on-demand infrastructure":

1. You provision virtual switches, load balancers, firewalls, servers, etc for your applications using AppLogic's visual canvas

2. You use Yahoo! Pipes to define whether/how various applications can access/consume/manipulate each other's data sources

Back in January I was very excited to hear about SoftLayer's Customer Exchange, a private meet-me-room where dedicated server owners can exchange data at gigabit speeds without incurring bandwidth charges. Now I'm envisioning a virtualized customer exchange on which application owners can create either inter- or intra-company mashups on the fly. How awesome would that be?!

Where does data belong?

Jon Udell wants a central repository for his digital assets, which include not just photos/videos/blog posts/etc, but also medical/financial/other records, and maybe social networking profiles/contacts too.

"Some are on websites that I control, and a lot more that I don’t. Others are on local hard disks that I control, and a lot more that I don’t... I’d be willing to pay for the service of consolidating all this stuff, syndicating it to wherever it’s needed, and guaranteeing its availability throughout — and indeed beyond — my lifetime."

He imagines a scenario that begins at birth: "In addition to a social security number, everyone gets a handle to a chunk of managed storage." The question is, where would this storage be located? Ilya Baimetov and I have been discussion 4 possible alternatives:

1. A "virtual personal server", as envisioned by Rod Boothby:

Your data would reside not in one single location but across a collection of services (such as Flickr, Zoho, etc) that you stitch together via interoperable APIs. The main drawback is, in order for your personal server to function properly, all underlying services must be up and running. In addition, Ilya points out that the larger your files, the less feasible it becomes to send them over the Internet for processing.

2. An enterprise-class storage system maintained by a hosting provider, as discussed in Hitachi CTO Hu Yoshida's recent blog post:

Ilya thinks it would be most efficient if you used applications that are hosted in the same data center also. But since this would require hosting providers to reach FAR beyond their accustomed role and get involved with selecting/maintaining/integrating/supporting large numbers of 3rd party apps, I'm not optimistic it will result in a coherent end user experience.

3. Multiple web apps, one storage vault:

Some time ago Fred Wilson mentioned he'd like to store his Gmail/Flickr/TypePad files on a 3rd party system so that he has more control over his data. OmniDrive CEO Nik Cubrilovic says that's exactly what he'd like to help accomplish.

"The idea is to aggregate your content from different apps and have them all in one place. Your files get cached with Omnidrive; from there you can open them up on your desktop or drag them into another folder which is linked to another app."

Of course, now you become dependent on OmniDrive's performance. Ilya also worries about the data transfer speed between OmniDrive and various apps.

4. Multiple web apps, distributed storage vault:

There's another storage provider called Cleversafe; its software splits your files into dozens of encrypted slices and stores each at a different data center. You can access your data as long as (and only when) you're able to access a majority of the slices. This simultaneously prevents data loss and unauthorized access. I'm fascinated by the possibility that when you sign up for Salesforce, Google Apps, etc in the future, you'd give each service your Cleversafe key so that any content you create is immediately encrypted/dispersed. But Ilya isn't sure it's possible to dis-assemble/re-assemble data on the fly without impacting performance.

Jon says a "hosted lifebits" service seems inevitable in the long run. Ilya thinks the shortest path between here and there is #2; he's even developed standards to help make it happen. I'm intrigued with #4 but thinking #1 is most realistic for now. You?

Web 3.0 magic

There's an article in the latest issue of Business 2.0 (sorry, not yet online) about the semantic web (aka Web 3.0). I was reminded of this passage while watching 3Tera's mega-grid demo last week (504 instances of UnixBench running across 400 CPUs with 1TB RAM, 40 TB storage and 200 Gbps of bandwidth capacity!).

Say you want to organize a dinner at an upcoming conference. Today you might go through your address book and ping folks by email to see who's attending. Then you go back and forth with the group on the time and place. In the semantic web, your software agent could cull the conference attendees and make a list of potential invitees, negotiate the date and time with everyone else's agents via a calendar database, pick a restaurant from another database based on availability...

This pretty much sums up the difference between physical and virtual infrastructure. Let's say you want to build a server cluster. Yesterday you might have assembled and installed software on individual servers, connected them to a switch... Now you could simply sketch out the desired components on your AppLogic canvas and press go. Your "software agent" would immediately head over to the data center and do all the work - within minutes. As Jimmy Guterman puts it, sometimes technology feels like magic.

In contrast, earlier today I had lunch with Hillary and Frank Stiff from Cheval Capital and David Snead. (We are getting ready for a HostingCon workshop about metrics that impact web hosting success; please join us!) We joked that when asked how their companies are exceptional, most web hosting providers offer the identical response: they've got great support! (Unrelated note - sales guru Jeff Thull says widely-shared value propositions make  a company sound more equal than different from competitors. It gives everyone "expected credibility" rather than the "exceptional credibility" one needs to rise above the noise.)

The problem is, "great support" might be the wrong basis for a vendor-customer relationship. "Great support" implies that you're focused on getting tricky technology to work properly - instead of exploring ways to do cool things with magical technology.

"Great support" proponents often remind me that no automation tool will ever diminish the importance of customer service - customers crave that personal touch! And I somewhat agree. But I think we might be entering an era in which "great support" starts giving way to "great advice". Instead of troubleshooting technology problems, winning web hosting vendors (and their software agents!) will soon be helping individual customers concoct magical solutions for achieving their particular business goals.

BT vs Verizon

BusinessWeek says that telecom is back. During the dotcom bust, the industry lost $2 trillion of market value in 2 years, but thanks to "watch-my-cat videos, iTunes and Internet phoning" profits this year are expected to reach $72 billion - an all time high that beats the 1998 record of $65 billion.

As an example, the article mentions Verizon as a phone company that's succeeded in reinventing itself. (When I called to cancel my land line a few months ago, I was surprised to hear the phone rep say, "thanks for calling your broadband entertainment company".) Five years ago, Verizon employed just 100 software developers; now it's home to over 1000. They've built services that will soon allow consumers to "share photos, videos, and other media among their cell phones, PCs, and TVs".

As CTO Mark Wegleitner tells CNET, Verizon currently has a hybrid network with both IP and traditional cable-like infrastructure. The company plans to migrate to an all-IP network over the next 3-4 years.

But across the pond, British Telecom CTO Matt Bross has been working on an all-IP network for ages. And the $18 billion Web21C will be the foundation for *much* more than an entertainment company.

* A few months ago BT launched BT WorkSpace, a collaboration app, and BT TradeSpace, a small business website builder/social network. Dennis Howlett is not a fan, but he hasn't spent the past decade in the web hosting industry. WorkSpace/TradeSpace help technically inexperienced SMBs get up and running the way no amount of oversold bandwidth and webspace can.

* BT also has a Salesforce AppExchange-like SaaS marketplace in the works.

* In March, BT released the Web21C SDK, which allows 3rd party developers to integrate BT's telco functionality (voice calls, conference calls, SMS, subscriber presence, etc) into their apps.

* Last week, BT acquired Osmosoft, parent company to the open source TiddlyWiki, and appointed CEO Jeremy Rustin head of its Open Source Innovation group.

When I told Senior Strategy Consultant Venkat Raju that I'm very impressed with these developments, he said I've only seen the tip of the iceberg. For BT intends to compete not just with other telcos - or even traditional IT service providers like IBM. It's building a mashup platform not unlike Amazon's and a suite of SaaS solutions that rivals Google Apps. In comparison, I wonder if Verizon should be setting its sights higher than photo/video sharing?

PS - BT's mashup platform/SaaS solutions will be powered by 3Tera's AppLogic! :)

The data center of the future?

I am fascinated with Sun's Blackbox, which packs hundreds of servers worth of computing resources into a 20 foot container. What it doesn't come with is a tool for managing all this horsepower through one single command line interface. The good news is, this very tool is available from 3Tera! There's a 400 CPU grid in their test lab at the moment. It's got 1 TB RAM, 40 TB storage and almost 200 Gbps of bandwidth capacity. And yes, it's being managed through one CLI.

3Tera's engineers will be running some benchmarks on the grid over the next few weeks. You can watch their progress through this live video feed.

In the not too far future, many of today's hosting providers might be running lights-out data centers that look like so. Hundreds of thousands (or perhaps millions) of CPUs' worth of processing power would be at your disposal - while you sit at a  Starbucks with your laptop.

Music-as-a-Service?

Fred Wilson says neither artists nor fans get the maximum benefit when people buy music that they can listen to as often as they want. He compares his consumption of the Rolling Stones' Exile on Main Street and Elvis' 50 Greatest Love Songs. If he has listened to the former thousands of times and the latter hardly ever, shouldn't the record labels be compensated accordingly? At $0.01 per listen, Fred would have paid $180 for Exile (over 20 years, so day-to-day costs are negligible) and $1 for Elvis. Under this model, the music industry would generate more revenue, and consumers would pay only for the benefit they receive.

Fred's post reminded me of Salesforce.com EMEA co-president Lindsey Armstrong's talk at Data Centres Europe. Lindsey said the difference between SaaS and traditional software is, developers make a living not by convincing end users to complete one-time purchases, but through their ongoing use of each app. If buyers stop benefiting from a product, its seller stops getting paid; this keeps the two parties' interests aligned.

Amazon's new pay per PUT/LIST/GET request pricing for S3 follows the same logic. You get what you pay for; you pay for what you get. This includes storage space, bandwidth and  necessary computing resources for processing your requests.

Utility pricing makes sense to me, which is why I'm excited to see that Rackspace's R&D group is "devoted to solving the issues of the future in the utility computing world" (check out their newly public blog here). The Planet, too, is looking into "running separate applications in an utility computing environment". It's about time! :)

Note from My Mom: Did You Know that Amazon's in the Hosting Biz?

My mom - who's a literature professor in Taiwan - sent me a link to this article (sorry; it's in Chinese). It's about Amazon's S3 and EC2. Now that Amazon's selling hosting, she asked, is there room left for anyone else in the market? After all, they're so far ahead in terms of name recognition. Consumers from all over the world buy stuff from their site. If she had to choose, she'd trust Amazon over a hosting-only provider who doesn't run a huge retail business on its own technology.

Now that's an interesting perspective. In contrast, my web hosting friends like to talk about core competency. They're in the business of running data centers; Amazon isn't. They're experts on maintaining networks; SalesForce and MySpace aren't.

But in this day and age, can hosting still be considered a stand-alone core competency? Where's the secret sauce, after all? Any company can buy state of the art gear and make hard to refuse job offers to experienced  network engineers?

Microsoft and Google, for instance, reportedly have $900 million and $750 million budgets for their rumored Texas/South Carolina facilities. Which hosting provider can afford that kind of capex? And on a smaller scale, Zoho's not only built a suite of web office apps but the grid infrastructure that its hosted software runs on.

It seems hosting has become more of a basic skill, a common denominator among companies that sell products or offer services through the Internet. Sure - pure play hosting providers acquired the knowhow first, but being first hardly guarantees everlasting preeminence? Jay Adelson, for instance, took the expertise he gained at Equinix to Digg. He's not the first to have crossed the web host/outside world boundary, and he won't be the last.

So maybe hosting companies need to redefine what they're in the business of, particularly since  in many cases their business models involve competing with Google for Google traffic and competing against free with monthly fees?