Applying Fanatical Support where it counts

On the same day VMWare's share price valued the company at $21 billion+ (that's more than 20x its annualized revenue of $1.03 billion), and Citrix acquired XenSource for $500 million, Bryan Betts from Techworld.com quotes Rackspace that virtualization is "not ready for the big time".

Bryan's assertion is based on results of Rackspace's recent virtualization survey (see press release and survey results PDF). He claims 87% of 3,000 customers would not be willing to share a physical server with other companies. Which is wrong.

Rackspace got a ~12% response rate, which means ~360 rather than 3,000 customers participated in the survey. 13% said yes to server sharing, 36% maybe, 51% no. So 183 respondents and not 2610 supported Bryan's conclusion.

What's even more misleading than Bryan's misinterpretation is the way Rackspace equates "using virtualization through a 3rd party" with "sharing a physical server with others". Customers, of course, said such an arrangement made them nervous about performance (46%) and scaling issues (10%).

Have the folks in San Antonio not heard of Amazon's S3/EC2, or Microsoft's utility computing plans? Ironically, Jeff Bezos and Ray Ozzie see their platforms as solutions to, NOT causes of, performance/scaling challenges. There's a world of difference between within-a-box and beyond-the-box virtualization, which gives customers on-demand access to multi-server resource pools.

So I agree with Bryan that hosting providers offering virtual servers could be "barking up the wrong tree". Partitioning standalone boxes into multiple virtualized environments is so 2004! Sadly, Rackspace's Nicholas Keller tells Bryan this is exactly what he's planning to do - by the end of 2007. Nick acknowledges he's "a bit behind the curve", but under-estimates the extent to which Rackspace is missing out.

Amazon and Microsoft made me realize that Internet infrastructure solutions should be - will be - delivered in 4 layers:

(a) Data centers/physical servers/virtualizataion software
(b) Utility computing fabric comprised of large pools of servers across multiple facilities
(c) Application frameworks, such as Amazon's web services APIs
(d) Shared services, such as identity management and social networking

Rackspace absolutely excels at (a), and Nick sees that the next phase of virtualization will be about infrastructure management, or (b). I think his plans to apply Rackspace's signature Fanatical Support at this level is a mistake. It's man vs machine, dude - time to think SLAuto and not just SLA!

More importantly, Rackspace is already losing business to Amazon because of its inability to deliver on (c). And there's no place like (c) for a professional services organization like Rackspace's to add value!!

SoftLayer, by the way, is all about automating (a) and (b) so that it can focus on (c), where it will need to either leave behind its "unmanaged hosting" mentality or build a 3rd party professional services ecosystem. The latter can't be done without (d).

PS - Lew Moorman nails it in his latest Racklabs post. He said VMWare's long term value will depend on how they handle the race to virtualization ubiquity. Now they're in the driver's seat, but they'd better re-evaluate their high prices before it's too late. Um, doesn't the same apply to Rackspace? Shouldn't Lew and his team move Rackspace's value proposition out of the infrastructure layer in a hurry before utility computing ubiquity crushes their pricing structure?

PPS - Allan Leinwald says even telcos should reach beyond the infrastructure layer (BT is certainly working on that). He wonders if Verizon or AT&T should buy a Web 2.0 property like Facebook; it'd give them lots of friends. James Urquhart responds that plumbers are plumbers. While Google/Amazon/Microsoft/eBay/SalesForce will make a buck off the rest of us with the full (a) through (d) stack, telcos most likely won't be able to complete - but at least they own the underlying network. Rackspace, SoftLayer and other hosting players will have to think hard about where they fit into this picture...

Comments

[John Rath]

What about Cache Fly?
Check out their latency shootout to show how they best Rackspace.
http://cachefly.com/research_latency_shootout.html

Maybe VMWare will buy them instead?

[Isabel Wang]

I've always been intrigued with CacheFly, but don't quite understand how they're positioning themselves. They call themselves a competitor to Rackspace/ThePlanet/Hostway and the no-longer-existent Interland, but they only offer 300MB to 2GB of storage. This makes their solution not quite complete. Would it make more sense for their service to be marketed as complementary to traditional hosting? Amazon's S3?

Also, who did you think VMWare should buy instead of CacheFly? Or did you mean someone else could be in the market to acquire CacheFly??

[Ted Smith]

I have to applaud Rackspace for making this data public. I think most companies would keep this sort of data close to their chest, but I also found two of the conclusions alarming.

1. "Seventy-one percent of survey respondents prefer to host mission-critical applications on a virtualization platform in conjunction with a third-party hosting provider."

2. "Sixty-four percent of respondents who had not yet used virtualization prefer to test it through a hosting provider."
You have to remember "who" was sampled in this survey. The respondents are Rackspace customers, which means, they are ALREADY using a 3rd party for key IT infrastructure --> Rackspace. They SHOULD be willing to host virtualization infrastructure at a 3rd party.

I'm concerned that "29% percent of survey respondents would prefer to host mission-critical applications on a virtualization platform in-house." Isn’t that saying that 29% of their customers wouldn’t host their mission-critical applications on a virtualization platform at Rackspace?

What does that mean for the rest of us in the hosting space? It would appear that hosting expertise does not necessarily translate to virtualization expertise. Which means all of us in the hosting space have our work cut out for us.

[Jeff Huckaby]

Cachefly is an entry-point into Level 3's CDN. They specifically target the entry level web sites. With Level 3's service you still need to maintain a web server such that they can pull the content to their delivery system. Last I checked, Cachefly essentially provides both the backend and the CDN in one. This provides a SMB with easy access to a CDN.

-------------

VMWare's flavor of virtualization may not be well suited to the hosting industry.

Their appeal is in the enterprise. I recently had a discussion with an IT manager of a large metal fabricator. He said they explored SaaS for certain items and are moving certain tasks to SaaS. But for him, moving wholesale to SaaS or utility computing was like taking a step backwards to thin-clients and mainframes (which they still use in some areas).

The problem with SaaS and grid computing, as he sees it, is reliability. No matter how well you plan, the system will fail. If you have 1000's of employees not able to work because the grid or SaaS is down, that is very very expensive.

He says with PCs on every desktop, someone can still work on that auto-cad drawing, write that memo, or prepare that report even if the network is down. So while your productivity is hampered you are not down to zero.

They were recently looking at VMWare to virtualize their desktop environments not their servers. Servers would likely get swooped in so they had a unified framework but the initial target is managing about 800 different workstations with complex application bases.

[Isabel Wang]

Ted - Keep in mind that 74% already have hands-on experience with virtualization and 57% already use it for internal infrastructure.

So it's possible that only 9.4% (36% out of the 26% who are new to virtualization) are uninterested in using Rackspace for testing, and just 12.5% (29% out of 43% who don't have existing deployments) would not like to host on Rackspace's virtualization platform.

In other words, there seems to be a high level of trust in Rackspace's service. Unfortunately, for the majority who've already tried virtualization on their own, it might not make sense to re-test/re-deploy at Rackspace.

[Isabel Wang]

Jeff - did you read IBM's announcement on how it's consolidating bunches of servers into virtualized mainframes? Your friend is right; we're going back in time :)

SaaS-outage related productivity loss is definitely a huge concern. Hence all the interest in web apps with offline access. I'm trying out a to-do-list service called Remember the Milk, which uses Google Gears. Was so impressed that it actually works while disconnected.

[John Rath]

If not CacheFly, I'm not sure who. I'm not 100% sold that VM's have a play in the hosting market. I think they do...just not sure to what extent. What about the Google model? Forget virtualizing and others, maybe the commodity server and lots of them is the way to go. With the shopping spree that Google is on it is tempting to say that they would buy a CDN, but realistically what they have built (assuming it is proprietary) is pretty darn good.
I did think the following announcement was at least something to ponder:
http://enewschannels.com/2007/08/17/enc1790_213656.php

-John