Pete Manca's Blog

And so it begins...PAN Everywhere

Today we announced our products with Dell and Fujitsu-Siemens, with support for both bladed and rack mount servers. This is a significant milestone in the company’s history, as it’s the first time we’ll deliver product on other vendors’ hardware.

It’s an especially poignant moment for me, as I’ve been at Egenera since the early days and have watched the product grow from a single OS (Linux) and a single chip (Intel 2 socket, single core) to what it is today, which is a broad based management platform that runs multiple operating systems, multiple hypervisors, multiple chip architectures, and now on multiple hardware platforms.

It’s part of our strategy to deliver PAN in as many form factors as possible, broadening the addressable market and distribution channels for PAN while at the same time, delivering tremendous value for our customers.

Some may see this as an indictment of our hardware…it’s anything but. In fact, it’s more a validation of our PAN architecture. Delivering PAN on different hardware will allow our customers a choice of platform, including BladeFrame, which remains our business critical flagship hardware platform.

So this is a big day in the history of Egenera, but it’s really another beginning, as we pursue getting PAN delivered in multiple ways through multiple channels. Our mission remains the same – to simplify the data center for our customers.

PAN + VMware – the killer combo

Yesterday, we announced our support for VMware Infrastructure 3 support on the Egenera BladeFrame. What’s the big deal, as everyone supports VMware today?

The big deal is the combination of the best-in-class server virtualization from VMware combined with the best-in-class I/O virtualization from Egenera. That combination is very powerful.

Why? Because server virtualization technology is enhanced greatly by I/O virtualization. Take virtual server migration (VMware VMotion) as an example. Doing this in an environment without I/O virtualization means open-zoning SANs and/or complicated clustered file systems.  Either situation can create configuration and security nightmares. But, add a secure I/O virtualization environment and these problems go away. Storage devices can now follow the virtual server, meaning no need for open-zoned SANs or clustered file systems. Much more manageable and much more secure.

That’s just one example of where the combination is a win-win for customers. Another area is the ability to manage physical and virtual servers in the same system. With PAN, physical servers can be deployed, managed, failed-over, and easily set up for disaster recovery. The same is true with VMware Infrastructure 3 and virtual machines. Combining the two gives the customer the best of both worlds.

This is a great announcement for Egenera customers, as it adds a key technology to our portfolio and adds greater capability for managing and securing virtual machines.

Dell and Egenera Partner to Bring Dynamic Data Center to the Mainstream

Today, Egenera and Dell announced a strategic OEM agreement, where Egenera’s industry leading PAN Manager software will be delivered on Dell hardware. This deal is important for many reasons, particularly:

• It’s a marriage of complementary technologies, shared vision and of strategic importance to both companies.
• It opens a significant channel for Egenera, bringing our proven software to a far broader customer base than ever before.
• Dell can now leverage the Infrastructure Virtualization capabilities of PAN – in production for more than 6 years – to simplify their customers’ environments and gain significant competitive advantage over other tier 1 hardware vendors.
• Our customers brought us together – more choice, more flexibility, and total simplicity in the data cent

Let’s face it, when it comes to delivering cost effective, industry standard servers to the market, no one is better than Dell. And, when it comes to delivering Infrastructure Virtualization and Management solutions, Egenera is the best. The combination in the market will be extremely powerful – for each of our companies and for our customers. Together, we are simplifying IT – leading virtualization and management into its next critical phase and allowing customers to focus on what’s important to their businesses.

This is not your standard “add another partner to our website” agreement.  It’s strategic for both companies and will bring real benefits to customers.

Stay tuned for more about the products we’ll launch together in the future. This is a great day for Egenera, for Dell, and most importantly, a great day for customers.

Is Disaster Recovery the next killer app for virtualization?

There have been a slew of articles speculating that Disaster Recovery is the next killer app for Server Virtualization. The premise is that since the OS and application are abstracted from the hardware, it’s simple to move the entire stack from one location to another, and presto – DR solved!

Unfortunately, it’s not that simple. While Server Virtualization goes a long way towards helping DR, there is still the issue of the physical machine management. Hypervisors require similar if not exact hardware configurations on both the local and remote systems. And applications that require DR typically use external storage and complicated network topologies. These too have to be configured identically. All this means that there is more to DR than moving a virtual machine from one server to another.

To truly solve the DR problem in a simple, cost effective manner, you need to solve all levels of complexity at the same time. Virtual machines take care of OS and application abstraction layers, making those components easy to move. Add in a Virtual I/O infrastructure that can virtualize the storage and networks, and you're almost there.

The remaining piece, and the one that is most often overlooked, is an Infrastructure Virtualization or abstraction layer that can represent the data center components – both physical and virtual - in a logical manner and in a way that can be replicated from site to site.

Remember, applications are not always 100% virtualized. Think of a 3-tier app where the web server and app server are virtualized but the database server remains physical (usually for I/O performance reasons). An Infrastructure Virtualization layer can abstract both physical and virtual servers, and allow for simple, low cost DR by replicating that environment on remote servers. It also serves as a hardware abstraction layer so dependencies on server configurations are removed.

So, before anyone buys in to the panacea that server virtualization solves DR, please look at the details and carefully understand what it can and can’t do. Then maybe you’ll see the value (PANecia…) of what an Infrastructure Virtualization solution like PAN can do.

IO Virtualization Defined

In my last blog I talked about what I see as the next "waves of virtualization" taking the industry, I talked about "IO Virtualization" as a key (maybe THE key) cornerstone, so I thought it might make sense to describe just what it is and why it's important.

Like all forms of virtualization, the physical world is masked and abstracted into a virtual representation, allowing for higher utilization or increased agility, etc. For IO, the physical world is typically defined by dedicated connections to Input/Output (IO) devices such as disks, networks, cdroms, consoles, etc. In the enterprise, these devices can themselves be already virtualized (think Storage Area Networks) but their connectivity is typically very static – e.g. a fiber channel adapter connected to a specific port on a SAN switch.

IO Virtualization abstracts the connectivity so the server itself is no longer statically configured to specific IO wiring. This adds tremendous value to the enterprise. The most obvious benefit is that change management becomes easy. No longer do techs have to fish wire and re-wire data centers because server-to-IO affinity has changed. With virtualization, the wiring stays static but the actual mapping between the server and the IO devices can be dynamic.

And this is just the beginning. IO Virtualization also allows for easy server re-purposing, as a server can be “re-wired” to attached to different volumes on a SAN device or different networks, thus changing its personality or its life cycle (e.g. moving from test networks to production networks).

IO Virtualization also enhances Server Virtualization, as it makes it easier for hypervisors to support migration – no need for clunky clustered file systems or open zoned SANs, which can expose serious security issues.

Taken even further, IO Virtualization can help with Disaster Recovery. When the server is abstracted away from its IO connectivity, it becomes very simple to move a set of servers (and their applications) to a remote site and quickly re-create the IO environment. Something that can take weeks in the physical world can literally take minutes now.

So, hopefully you can see why IO Virtualization is such a key cornerstone technology in the continued virtualization evolution. Be aware, though, that not all IO Virtualization is created equal. IO Virtualization is not done by adding new layers of unmanaged hardware into the mix (think NPIV) and IO Virtualization is not simply re-programming network and SAN switches as some would have you believe. It’s really about creating an agile abstraction layer that truly disassociates servers with their IO devices, allowing for total flexibility in the data center.

Where do we go from here?

Well, 2008 is in full swing and I’ve been remiss in my blogging. It’s been a very busy time...but no excuses. 2007 was a breakout year for virtualization, as it finally hit the mainstream. But where do we go from here?

The virtualization "wave" is just forming. And while server virtualization is at full crest, there are many more waves behind this that are taking shape and quite frankly, are more significant.

Server virtualization was about saving money. Allowing multiple applications to be consolidated onto a single server saves capital and operational expenses. Reducing the number of servers running in the data center is a good thing, as it also saves some carbon emissions as well. But is that it? If so, that’s more like a ripple than a wave. Don’t get me wrong, reducing power, cooling, server count and consolidating apps is a good thing, but it's not the whole story. Not by a long shot.

I don’t believe that this is it. In fact, I think we are still in the beginning stages of realizing what virtualization can do. It’s really the enabling technology that fuels the ability to create new ways to solve problems that exist in today’s data center.

As with all great technology movements, a core set of technologies must be established first. Server virtualization is one for sure, but what are the others? IO Virtualization might be the next important cornerstone technology. Without solving this problem, servers continue to be static and inflexible. We might be able to utilize servers more by virtue of the hypervisor, but we can’t exploit them to their fullest extent without the flexibility to change their IO bindings dynamically. Other key virtualization technologies include file virtualization, data virtualization, and application virtualization. These are keys to making access to applications, data, and resources agile and ubiquitous.

Once the server, IO, data, and applications are virtualized, the resulting possibilities and opportunities really are endless. These cornerstones open the market for management, security, converged fabrics, and a whole host of technologies that can free up the data center and open new markets.

Expect 2008 to be another banner year for virtualization. The next wave is here.

If it Walks Like a Duck and Quacks Like a Duck...

In a recent game of one-upmanship, both HP and IBM announced the ability to virtualize IO and scale this to hundreds of blades across a hundred chassis. I can just imagine the marketing gurus at HP and IBM, pinky in the corner of their mouths Austin Powers-like, stating that they can now virtualize thousand of blades.

Drilling a little further in to these grand announcements we find that these two behemoths are up to their same tricks. Implementing proprietary hardware and no software. In fact, the hardware that they are implementing is mostly NPIV technology that is widely available today. So what’s the big deal?

To me, the big deal is what they didn’t say. These solutions lock the customer in to their proprietary blade formats. There is no support for rack servers. There is no support for other vendors' server hardware. This is a full-blown, proprietary hardware-orientated solution. And oh, by the way, they also fail to mention that for this to work each and every one of these 100 blade chassis have to have access to all of the LUNs in order to share the WWNs across chassis. That adds up to over 1000 SAN ports and even more cables. What a configuration and security nightmare!

Look, it’s great that IBM and HP have figured out that IO virtualization is a key element in creating agile infrastructure, but putting out complicated, partial solutions is worse than doing nothing at all.

As we’ve said all along here, it takes a new architecture and a new way of looking at things to create a utility computing environment. There must be synchronicity between the hardware and software and to truly be relevant in the data center, the solution must span blades, racks, and hardware vendors. The major hardware vendors have taken a step back towards proprietary hardware centric solutions. Interesting, with our new software products, only Egenera is filling that void today.

Quack, Quack.

Hypervisor Market Commoditizing Quickly

Wow, did you catch all of the announcements for new hypervisors? Oracle has one…. Sun has one… Microsoft unveiled their plans – and pricing, for Hyper-V…. heck, even Phoenix is planning on one. These, added to a market that includes VMware, Xensource, SWSoft, SuSE, Red Hat, IBM, Qumranet… well, you get the picture. This is a very crowded market.

This is typical for hot technologies. There are a few early innovators that create the technology and as they gain success, others follow, and the market quickly commoditizes. We’ve seen this a number of times before. And as with these previous technologies, the last piece (and often forgotten piece) is management.

Standards are slow to emerge in the management space and vendors rush to try to add value to justify the commodity under the covers. This is exactly what is taking place in the hypervisor market. Microsoft has set their price as $28 for Hyper-V. Xen is “free” and the gorilla of them all, VMWare, has announced a “lite” version at a much reduced price.

In the end, the hypervisor will be free and part of the server platform. The next battle will be around the management of this new virtual world, and as in the past, this battle will take many years to evolve and will go in many different directions, as the lack of standards means each vendor will provide their own flavor.

Customers should look for a management system that is heterogeneous and provides value over and above just hypervisor management. Physical machine management is still a very important part of the data center, as are things like high availability, disaster recovery, charge back, and other management services that create real business value for the data center. The integration of physical and virtual machine management, along with higher value management tools, in a heterogeneous way is the nirvana that data center managers are looking for.

Egenera enters the software market

Today Egenera announced we are entering the software market. We are taking the incredible power of our Processing Area Network technology and applying it to other vendor’s hardware. This is a big step for Egenera, though it’s more of a natural evolution. We are NOT getting out of the hardware business and will continue to innovate and extend our hardware platform. However, it’s natural to take the jewel of our IP – PAN Manager, and apply it to as many platforms as we can so we can address a broader section of the market.

PAN Manager has been developed over 7+ years and has become the leading Data Center Virtualization software on the market. This software, coupled with our mission-critical, fabric-based, hardware platform makes for an incredibly robust and secure platform for mission critical applications. Taking PAN and extending it to support other hardware allows us to bring the same value proposition to commodity hardware.

Data Center Virtualization, or Virtualization 2.0 as discussed in my last blog, is about simplifying the data center. This can’t be done taking a narrow hardware view. Porting PAN to commodity hardware will allow PAN to have a dramatic impact on the data center.

Specifically, today we announced that we will be making PAN Manager available on select partners' hardware platforms. We're doing this for two reasons:

1. To give customers further flexibility and choice in how they deploy this powerful software and
2. To increase our global distribution channels and addressable market

We'll continue to innovate and advance our BladeFrame family of solutions in tandem with this new software business. We believe that the BladeFrame continues to offer the highest-performance solution designed for customers' most business and mission critical applications.

In the coming weeks and months, we'll be making further announcements about which partners we're going to market with and when the solutions will be available.

We look forward to leading the industry into the next phase of virtualization.

Virtualization Market Matures

Even with the incredible traction virtualization is making in the market it’s still easy to get confused as to what virtualization means in every case. Is it server virtualization provided by hypervisors like Xen or ESX? Is it storage virtualization? Network virtualization? Broadly, it's all of the above.

One thing that's clear is that the market for virtualization is growing and maturing, and one advantage of this evolution is that the industry analysts are starting to categorize and differentiate between virtualization technologies and virtualization vendors. This helps us sure, but it also helps customers understand where they are in their own deployments vs. where they can go in the future.

Recently, the pundits have begun making distinctions between the different types or phases of virtualization. For example, IDC has identified two current levels of virtualization; Virtualization 1.0 and 2.0+. Virtualization 1.0 is defined as server virtualization – using hypervisors to partition resources – while Virtualization 2.0+ is defined as the next generation of virtualization technology – focused on virtualizing data center infrastructure beyond just the server.

Virtualization 1.0 is targeted at reducing capex through consolidation, something that hypervisors do extremely well. Virtualization 2.0+ focuses on reducing opex by adding capabilities around infrastructure virtualization, management tools to simplify management, and higher value tools like Disaster Recovery and hardware failover. Ultimately, Virtualization 2.0+ transforms data center infrastructure into flexible, changeable assets that can be deployed, moved and managed seamlessly.

As I said, the categorization that's happening now is important. It allows customers to wade through the confusing virtualization landscape and choose products that they can actually benefit from and complement what they already have installed. We know that virtual machine sprawl is becoming real and with this sprawl comes a new set of challenges – managing the infrastructure that connects the sprawling VMs.

This is where Egenera PAN Manager comes in. We are the clear leaders in the Virtualization 2.0+ market and have been solving infrastructure complexity through virtualization for 7+ years. We call this Data Center Virtualization and it’s an important component in reducing data center complexity - for both virtual and physical assets. As the virtualization market matures and moves towards 2.0 and beyond, solutions like PAN will be increasingly important to the data center, as server virtualization alone can’t solve the complexity issues. In some cases, it may add to them.