Why Utility Computing Failed (But Cloud Computing Didn't)

Since about 2006 I’ve been involved with data center IT automation. Back then I started with Cassatt, one of the first companies trying to automate infrastructure components in the data center.  Rob Gingell, the CTO, had a design principle of “service-level automation”, where the variable monitored and maintained was the service, not the server. That was a revolutionary thought.

The technology behind this was a combination of orchestrating physical and virtual devices, which automatically composed appropriate infrastructure stacks to keep the service SLA within pre-defined bounds. And it absolutely worked!  The best market description we had for this technology was “Utility Computing,” and drew from the analogy of electrical utilities: No matter what the draw (load), the supply would always be generated/retired (elasticity) to keep up with it.
 
But selling the Utility Computing model, and service-level automation technology, was hard, if not impossible. We’d frequently have successful POC’s, and demonstrate the product, but the sales inevitably stalled.  The reasons were many and varied, frequently tied to the ‘psychographic’ of the buyers.  But overall, we could point to a few frequent problems:

• Automation was scary: The word “Automation” frequently scared-off  IT administrators.  They were accustomed to complete control of their hand-crafted infrastructure, and visibility into every layer.  If they couldn’t make and see the change, they didn’t trust that the system actually worked.
• Lack of  market reference points:  Peers in the market hadn’t tried this stuff either – and there was no broad acceptance that utility computing was being adopted
• Inflexible Process: The use of ITIL and ITSM procedures were designed to govern manual IT control, and had no way to incorporate automatic approaches to (for example) configuration management.
• Organizational fear: There was usually the un-stated fear that the utility computing automation systems would obviate the need for certain jobs, if not entire IT organizations. Plus, the systems spanned multiple IT organizations, and it was never clear which existing organization should be put in charge of the new automation.
• Multiple Buyers: Because Utility Computing touched so many IT organizations, the approval process necessarily included many of them. Getting the thumbs-up from a half-dozen scared organizations was hopeless. Even if the CxO mandated utility computing, implementation was inevitably hog-tied.

Enter Virtualization

Somewhere around 2007, OS virtualization began to go mainstream. And its value proposition was simple and uncomplicated: Consolidate applications, reduce hardware sprawl. It was a no-brainer.

But just below the surface, virtualization had an interesting effect on IT managers: It began to make them more comfortable to break the binding of physical control and physical management of servers, transitioning instead to being more at ease with logical control of servers.

As consolidation initiatives penetrated data centers, additional virtualization management tools followed. And with them, more automated functions. And with each new function came IT’s incremental comfort with automating logical data center configurations.

And Then, Commercial Examples

At just about the same time, Amazon Web Services had begun to commercially offer these virtual machines in their EC2 – Elastic Compute Cloud. This could be had for the use of a credit card, and charged-for on an hourly basis. IT end-users now had simple – if sometimes only experimental – access to a truly automated, logical infrastructure. And one where all “hands-on” aspects of configuration were literally masked inside a black box.

Now the industry had its proof-point: There were times when full-up IT automation, without visibility into hardware implementation, worked and was useful.

Use of EC2 (initially) lay outside the control bounds of IT management and IT’s organizational boundaries. Developers and 1-off projects could leverage it without fear of pushback from IT – usually because IT never even knew about its use.

Once IT management acquiesced that EC2 (and similar services) was being used, they finally had reason to look more closely. And the revelations were telling: How was it that the annualized cost basis for a medium-sized server was lower than an in-house implementation could possibly hope to achieve? How come configuration and tear-down was so simple? Finally IT had to look in the mirror at the fact that this thing called cloud computing might be here to stay?

Looking Back

While it’s clear that the concept of cloud computing isn’t new, some important industry changes – more psychological and organizational than technological – had to take place before widespread adoption would happen.  And even then, it took some simple commercial implementations to prove the point. Too bad these weren't around a few years earlier during the "utility computing" era.

Watching this unfold, lessons *I* learned – or at least some explanations of this effect I’ve examined have been

• Psychology/Attitude shifted:  the more broadly OS virtualization was adopted, the more IT’s attitudes became accepting of automation and of logical control.
• Technology change was replaced by operational change: The new approach is more a change to the operational approach than a technology upheaval. The way users interacted with the cloud was appealing and nearly viral.
• Value was Immediate: The “new” cloud economic evidence was/is usually so compelling that it has forced IT to take a second look. This started with simple consolidation economics, but has expanded well beyond that.
• Broad availability accelerated adoption: Even only a few commercially available cloud providers helped provide immediate proof-points that the new model was here to stay. And purchasing this technology was as simple as entering a credit card number

Going forward, I would expect these 4 (perhaps more) “pressure points” will continue to help accelerate the use and adoption of internal clouds, public clouds, etc.   In future Blogs I’ll begin to look at how to further mainstream Cloud (and automation) adoption, as it serves to accelerate improvements to business’ bottom line.

Hosting/Cloud Index outperforms NASDAQ Nearly 4x

Exactly a year ago, in 2009, I started looking into whether the market for Hosting and Cloud Computing could be measured in the stock market - with my first Blog on the topic. I revisited again in January 2010, and later in February 2010.

I created a number of model portfolios based on leading public hosting companies. The list included: Digital Realty Trust; DuPont Fabros; Equinix; Internap; Iomart; Macquarie Telecom; Navisite; Rackspace; Savvis; Switch & Data; Telecity; Terremark. During 2010, this index has gained 45% to the Nasdaq's 17%

I also created a sub-group of those companies with explicit cloud computing offerings. That included Equinix; Navisite; Rackspace; Savvis; Terremark. During 2010, this sub-index has gained 57% to the Nasdaq's 17%

I'm not surprised by either outcome... Assuming the market doesn't lie because it takes into account valuations and expectations, I draw a few personal conclusions:

• More enterprises are turning to outsourcing their IT. Whether or not it involves cloud computing, I suspect enterprises find it advantageous to hand-over IT (management, or at least co-location) to businesses where this technology is core. It may also indicate the skyrocketing consumption of computing power by enterprises.
• It would appear that hosting firms with Cloud Computing offerings are being valued higher than their counterparts. I haven't looked at cashflows or balance sheets (yet) to determine whether this is actual value, or speculative value.

What's next for 2011?  I'm going to guess more of the same, if not an acceleration as more companies move to outsource non-core IT operations.  I'll also be watching consolidation of data center operators, as recently evidenced by Rackspace acquiring cloudkick, and Cologix acquiring Navisite. 'Guess I'll have to update my portfolio companies...

IO Virtualization: The “Hypervisor” for Your Infrastructure

An Explosive Technology, But Don't Treat as a Standalone Product
  
More than ever in 2010, IO Virtualization (IOV) has been showing-up in products, written about, spoken about. Because I’ve had a few years’ experience with this technology, I wanted to give a very brief explanation of the concept, and focus more on why it will be increasingly important.
 
In particular, I want to draw an analogy where you should view IOV as a critical enabling feature of future IT Management…  but not as a stand-alone product. Why? It's similar in concept to how the hypervisor is an enabler (but usually not used as a stand-alone product) of data center management services. 

This blog is related to my 2009 installment on Fabric as an IT Enabler.

What is IOV?

Today's Physical Infrastructure

IO Virtualization is an approach whereby physical IO components such as Network Interface Cards (NICs) Host Bus Adaptors (HBAs) and Keyboard/video/Mouse ports (KVM) are reproduced logically rather than physically.  In other words, a physical IO port (Ethernet, Infiniband, PCI, etc.) might logically represent itself to the O/S as different configurations.

Clearly this is convenient because it (a) eliminates multiple costly IO devices that also consume power and installation time. But it’s also convenient because IO – and it’s associated addressing such as IPs, MACs, Worldwide Names, etc. – can be instantly configured with a mouse.


The other consequence of IOV is that a single physical port means a single physical cable.  In essence, a server’s logical IO is consolidated down to a single (physical) converged network which carries data, storage and KVM traffic.   So this means that no matter how many logical IO devices you configure for a server, there is still only a single cable out the back.  So IOV yields the ideal “wire-once” server environment that’s still infinitely re-configurable.

The overall value of IOV becomes clear fast:  Fewer physical IO devices to buy, fewer cables to install, zero re-cabling, fewer physical ports to buy, and instantly re-configurable IO. 

Differing Implementation Approaches

Infrastructure With
IO Virtualization

In brief, there are a few differing approaches to IO virtualization:

• Existing on-board Ethernet with new IO drivers: (e.g. Egenera)
• Converged Networking Adapters (e.g. Qlogic, Emulex)
• Appliances + high-throughput IO devices (e.g. Xsigo)
• Existing physical IO but with address hardware-based mapping/virtualization (e.g. HP VirtualConnect)

Putting IOV in Perspective

You should think of IOV using the following analogy: The way in which the hypervisor abstracts software in the application domain, IOV abstracts IO and networking in the infrastructure domain.  (However, to be clear, IOV is not a software layer as-is the hypervisor)

This analogy leads to a few more observations:

1. Where the hypervisor added software portability in the software domain IOV will do the same for the infrastructure domain.  Higher-order services like HA and consolidation were made possibly by the hypervisor.  Similarly, HA, DR and migration can be accomplished with IOV. And what’s more, a hypervisor is not required for IOV, so you can use IOV with native applications too.
2. The hypervisor used to be the focus, but now it’s merely an enabling feature embedded within higher-level IT management products. Those products leverage the hypervisor to perform tasks such as migration, fail-over and consolidation. You should view IOV similarly: it is an enabling feature that will allow for analogous IO consolidation, migration and fail-over.
3. Where hypervisor implementations and performance used to be hotly-debated, nobody really cares anymore.  Today the real *value* is not in the hypervisor, but in the management tools surrounding it.  Similarly, IOV should be judged less on how it is implemented, and more on the management tools and automation which manage it.

Forrester analyst Galen Schreck made a similar observation recently:

….Aside from benefits like reducing cabling and switch ports, I think the most interesting aspect of virtualized IO is the ability of a physical server's personality to be moved to any other server in the data center. In addition to the underlying network technology, the thing that makes this possible is integrated management of the server and data center fabric. In most cases, this won't be a stand-alone product that you acquire (though you can build your own solution from InfiniBand and PCI Express products on the market). This capability will most likely be an integrated part of whatever server and network environments you select, but now is the time to begin planning how you'll tie it in with the rest of your system management environment.

IO Virtualization in the IT Management Landscape

How might IO virtualization be used as part of the IT ecosystem in an integrated manner?

In much the same way that the hypervisor has since been embedded in tools like VMware’s vCenter, IOV can (and has been) embedded with higher-level management tools.

Taking an example I’m rather familiar with, Egenera’s PAN Manager Software surrounds IOV technology with facilities such as integrated with converged fabric networking, server boot control and storage connectivity.  When used alongside these and other services, IOV enables:

• Server High Availability– In the case of hardware failure, a server’s infrastructure state (IO addressing, storage naming, network topology and workload) can be re-instantiated on another bare-metal server. This provides a ‘universal’ style of failover that doesn’t require clustering software. And what’s more, the failed-over server workload could be a native OS, or a VM host.  IOV is agnostic to the workload!
• Disaster Recovery – expanding on the example above, if an entire domain of servers fails, the entire group of server IO states, networking states, etc. can be recovered onto another domain (assuming shared/replicated storage).  This approach to DR is elegant because it fails-over not just workloads but  the entire logical server/environment configuration as well.
• Scaling-Out – where a series of server profiles can be instantly replicated into an instant cluster. Workloads, NICs, HBAs, networking addressing and storage connections (complete with fabric-based load balancing) can all be cloned… starting with the IO and networking profiles, made possible through IOV.

In future blogs I’ll dive more deeply into how software-based IOV operates as part of the IT management ecosystem, and why it is a popular approach because of its cross-platform compatibility in a heterogeneous data center.

Converged Infrastructure, Part 3

Converged Infrastructure: What it Is, and What it Isn't

In my two earlier posts, I first took a stab at an overview of converged infrastructure and how it will change IT management, and in the second installment, I looked a bit closer at converged infrastructure's cost advantages. But one thing that I sense I neglected was to define what's meant by converged infrastructure (BTW, Cisco terms it Unified Computing). Even more important, I also feel the need to highlight what converged infrastructure is not. Plus, there are vendor instances where The Emperor Has No Clothing -- e.g. where some marketers have claimed that they suddenly have converged infrastructure where the fact remains that they are vending the same old products.

Why splitting hairs in defining terms? Because true converged infrastructure / unified computing has architectural, operational, and capital cost advantages over traditional IT approaches. (AKA Don't buy the used car just because the paint is nice)


Defining terms - in the public domain
Obviously, it can't hurt to see how the vendors self-describe the offerings... here goes:
 
Cisco's Definition (via webopedia)

"...simplifies traditional architectures and dramatically reduce the number of devices that must be purchased, cabled, configured, powered, cooled, and secured in the data center.  The Cisco Unified Computing System is a next-generation data center platform that unites compute, network, storage access, and virtualization into a cohesive system..."

Egenera's Definition

"A technology where CPU allocation, data I/O, storage I/O, network configurations, and storage connections are all logically defined and configured in software. This approach allows IT operators to rapidly re-purpose CPUs without having to physically reconfigure each of the I/O components and associated network by hand—and without needing a hypervisor."

HP's Definition

"HP Converged Infrastructure is built on a next-generation IT architecture – based on standards – that combines virtualized compute, storage and networks with facilities into a single shared-services environment optimized for any workload."

Defining terms - by using attributes
Empirically, converged infrastructure needs to have two main attributes (to live up to its name): It should reduce the quantity and complexity of physical IT infrastructure, and it should reduce the quantity and complexity of IT operations management tools. So let's be specific:

Ability to reduce quantity and complexity of physical infrastructure:

• virtualize I/O, reducing physical I/O components (e.g. eliminate NICs and HBAs)
• leverage converged networking, reducing physical cabling and eliminating re-cabling
• reduce overall quantity of servers, (e.g. ability to use free pools of servers to re-purpose for scaling, failure, disaster recovery, etc.)

Ability to reduce quantity and complexity of operations/management tools:

• be agnostic with respect to the software payload (e.g. O/S independent)
• fewer point-products, less paging between tool windows (BTW, this is possible because so much of the infrastructure become virtual and therefore more easily logically manipulated)
• reduce/eliminate the silos of visualizing & managing physical vs virtual servers, physical networks vs virtual networks
• simplified higher-level services, such as providing fail-over, scaling-out, replication, disaster recovery, etc.

To sum-up so far, if you're shopping for this stuff, you need to
a) Look for the ability to virtualize infrastructure as well as software
b) Look for fewer point products and less windowing
c) Look for more services (e.g. HA, DR) baked-into the product.

Beware.... when the Emperor Has No Clothes...
In closing, I'll also share my pet peeve: When vendors whitewash their products to fit the latest trend. I'll not name-names, but beware of the following stuff labeled "converged infrastructure":

• If the vendor says "Heterogeneous Automation" - that's different. For example, it could easily be scripted run-book automation.  This doesn't reduce physical complexity in the least.
• If the vendor says "Product Bundle, single SKU" - Same as above. "Shrink wrapped" does not equal "converged"
• If the vendor says "Pre-Integrated" - This may simplify installation, but does not guarantee physical simplicity nor operational simplicity

 Thanks for reading the series so far.  I'm pondering a fourth-and-final installment on where this whole virtualization and converged infrastructure thing is taking us - a look at possible future directions.
 

Postcards from the IT Financial Management Association

This week marks the third time I have been invited to speak at the ITFMA World of IT Financial Management conference.  This is a really amazing/unique conference, created nearly single-handedly by Terry Quinlan, their Executive Director. Quick overview:

The IT Financial Management Association (ITFMA) was established in 1988 and founded the IT Financial Management profession at that time. ITFMA is the only association dedicated to this profession and provides a comprehensive education program on the principles and practices used to financially manage Information Technology (IT) organizations. ITFMA is the national leader in the education of IT financial management professionals and the only recognized provider of certification in the various financial disciplines of IT financial management.

The attendees are largely non-technical, but are comprised of financial managers, controllers, project managers and purchasing managers all in the IT field mainly with F1000 companies.

And what sets this conference apart for me is the fact that 90% of the topics of conversation are non-technical. It's not about the speeds-and-feeds, but rather about the project management, cost accounting, charge-back, managerial and regulatory issues facing IT.  It gave me pause that, while technologists focus on keeping the electrons moving, there are also folk who keep the paper and the money moving.

On particularly illustrative conversation I had in mind -- with an IT financial manager from the State of Oregon, who oversees the state's shared/hosted IT infrastructure.  They were promised by a large national consulting company that through consolidation of equipment and data centers, the state would save tons of $$ and reduce the managerial headcount as well. As it was described to me, the technical consolidation was largely a success, but the consultant failed to accurately account for the business and managerial staffs associated with the IT. And over time, while the square feet of data center shrank, the overall IT staffing continued to grow. Lest we commit the sin of assuming that all of IT is technologists.

Overall, the ITFMA is a "must-attend" -- especially now that IT is going through such large changes as data center consolidation, virtualization, automation and cloud computing. All of these have non-linear impacts on IT finances, and all can cause disruptive effects on topics like capital forecasting, project management, expense vs  investment projections, etc. Not to mention the newer issues caused by cloud computing such as data ownership, security, operations control, etc.

The event is a relative bargain to attend, and Terry always finds classic, historic venues for the conferences.

Converged Infrastructure Part 2.

Part 2. Converged Infrastructure’s Cost Advantages

In my first installment about converged Infrastructure, I  gave an outline of what it is, and how it will change the way in which IT infrastructure is managed.

In this installment, I’ll go a bit deeper and explain the source of capital and operational improvements converged Infrastructure offers – and why it’s such a compelling opportunity to pursue.

But first, the most important distinction to make between converged infrastructure and “the old way of doing business” is that management – as well as the technology – is also converged.  Consider how many point-products you currently use for infrastructure management (i.e. other than managing your software stack). 

This diagram at right  has resonated with customers and analysts alike. It highlights, albeit in a stylized fashion, just how many point-products an average-sized IT department is using.  This results in clear impact in

• Operational complexity – coordinating tool use, procedures, interdependencies and fault-tracking
• Operational cost – the raw expense it costs to acquire and then annually maintain them
• Capital cost – if you count all of the separate hardware components they’re trying to manage

That last bullet, the thing about hardware components, is also something to drill down into.  Because every physical infrastructure component in the “old” way of doing things has a cost.  And I mean I/O components like NICs and HBAs, not to mention switches, load balancers and cables.

What might be possible if you could virtualize all of the physical infrastructure components, and then have a single tool to manipulate them logically?

Well, then you’d be able to throw-out roughly 80% of the physical components (and associated costs) and reduce the operational complexity roughly the same amount.

In the same way that the software domain has been virtualized by the hypervisor, the infrastructure world can be virtualized with I/O virtualization and converged networking. And, once the I/O and network are now virtualized, they can be composed/recomposed on demand.  This eliminates a large number of components needed for infrastructure provisioning, scaling, and even failover/clustering (more on this later).  And, if you can now logically re-define server and infrastructure profiles, you can also create simplified Disaster recovery tools too.

In all, we can go from roughly a dozen point-products down to just 2-3 (see diagram above).  Now: What’s the impact on costs?

On the capital cost side, since I/O is consolidated, it literally means fewer NICs and elimination of most HBAs since they can be virtualized too.  Consolidating I/O also implies converged transport, meaning fewer cables (typically only 1 per server, 2 if teamed/redundant). And a converged transport also allows for fewer switches needed on the network.  Also remember that with few moving (physical) parts, you also have to purchase few software tools and licenses. See diagram below.

On the operational cost side, there are the benefits of simpler management, less on-the-floor maintenance, and even less power consumption. With fewer physical components and a more virtual infrastructure, entire server configurations can be created more simply, often with only a single management tool. That means creating and assigning NICs, HBAs, ports, addresses and world-wide names. It means creating segregated VLAN networks, creating and assigning data and storage switches. And it means automatically creating and assigning boot LUNs. The server configuration is just what you’re used to – except it’s defined in software. And all from a single unified management console.   The result: Buying, integrating and maintaining less software.

Referencing the diagram at right, here's what this looks like on a physical level is fewer components: Costly NIC and HBA cards are virtualized, with their physical transport now consolidated over Ethernet ports, and switches/cables now replaced by a logically-configured switch.

Ever wonder why converged infrastructure is developing such a following? It’s because physical simplicity breeds operational efficiency. And that means much less sustained cost and effort. And an easier time at your job.

Next installment: What Converged Infrastructure is not.

Converged Infrastructure. Part 1

Since joining Egenera, I've been championing what's now being termed Converged Infrastructure (aka unified computing). It's an exciting and important part of IT management, demonstrated by the fact that all major vendors are offering some form of the technology. But it sometimes takes a while for folks (my analyst friends included) to get their heads around understanding it.  So I'm going to take a stab at a multi-part Primer on the topic.
 
Part 1: What is Converged Infrastructure, and how it will change data center management

Converged Infrastructure and Unified Computing are both terms referring to technology where the complete server profile, including I/O (NICs, HBAs, KVM), networking (VLANs, IP load balancing, etc.), and storage connectivity (LUN mapping, switch control) are all abstracted and defined/configured in software. The result is a pooling of physical servers, network resources and storage resources that can be assigned on-demand.

This approach lets IT operators rapidly repurpose servers – or entire environments – without having to physically reconfigure I/O components by hand—and without the requirement of hypervisors.  It massively reduces the quantity and expense of the physical I/O and networking components as well as the time required to configure them. A converged infrastructure approach offers an elegant, simple-to-manage approach to data center infrastructure administration. 

From an architectural perspective, this approach may also be referred to as a compute fabric or Processing Area Network. Because the physical CPU state (i.e. naming and configuration of I/O, networking and storage naming) is completely abstracted away, the CPUs become stateless and therefore can be reassigned extremely easily creating a “fabric” of components, analogous to how SANs assign logical storage LUNs.  And, through I/O virtualization, both data and storage transports can also be converged, further simplifying the physical network infrastructure down to a single wire.

 The result is a “wire-once” set of pooled bare-metal CPUs and network resources that can be assigned on demand, defining their logical configurations and network connections instantly.

BTW, there is another nice resource -- a white paper commissioned by HP (!) executed by Michelle Bailey at IDC. In it she defines what is a converged system:

"The term converged system refers to a new set of enterprise products that package server, storage, and networking architectures together as a single unit and utilize built-in service-oriented management tools for the purpose of driving efficiencies in time to deployment and simplifying ongoing operations. Within a converged system, each of the compute, storage, and network devices are aware of each other and are tuned for higher performance than if constructed in a purely modular architecture. While a converged system may be constructed of modular components that can be swapped in and out as scaling requires, ultimately the entire system is integrated at either the hardware layer or the software layer.

Converged Infrastructure and Software Virtualization

A Converged Infrastructure is different from—but analogous to—hypervisor-based server virtualization.  Think of hypervisors as operating “above” the CPU, abstracting software (applications and O/S) from the CPU; think of a Converged Infrastructure as operating “below” the CPU, abstracting network and storage connections. However, note that converged Infrastructure doesn't operate via a software layer the way that a hypervisor does. And converged Infrastructure is possible whether or not server virtualization is present.

Converged Infrastructure and server virtualization can complement each other producing significant cost and operational benefits. For example, consider a physical host failure where the entire machine, network and storage configuration needs to be replicated on a new physical server. Using Converged Infrastructure, IT Ops can quickly replace the physical server using a spare “bare metal” server.  A new host can be created on the fly, all the way down to the same NIC, HBA and networking configurations of the original server.

A Converged Infrastructure can re-create a physical server (or virtual host) as well as its networking and storage configuration on any “cold” bare-metal server.  And in addition, it can re-create an entire environment of servers using bare-metal infrastructure at a different location as well. Thus it is particularly well-suited to provide both high-availability (HA) as well as Disaster Recovery (DR) in mixed physical/virtual environments – eliminating the need for complex clustering solutions. And in doing so, a single Converged Infrastructure system can replace numerous point-products for physical/virtual server management, network management, I/O management, configuration management, HA and DR.

Converged Infrastructure - Simplifying Management for “The other half” of the Data Center

In the manner that server virtualization has grown to become the dominant data center management approach for software, converged infrastructure is poised to become the dominant management approach for “the other 50%” of the data center – its infrastructure.

However adoption will take place gradually, for a few reasons:

• IT can only absorb so much at once. Most often, converged infrastructure is consumed after IT has come up the maturity curve after having cut their teeth on OS virtualization. Once that initiative is under way, IT then begins looking for other sources of cost take-out.... and the data center infrastructure is the logical next step.
• Converged infrastructure is still relatively new. While the market considers OS virtualization to be relatively mature, converging infrastructure is less-well understood.

But there is one universal approach that can overcome these hesitations -- money.  So, in my next installment, I'll do a deeper dive into the really fantastic economics and cost take-out opportunities of converging infrastructure...