Tech Trends

PC on a (USB) Stick

Whenever I hear about “cloud computing” (storing critical data and applications on a server on the Internet rather than on my local hard drive) I groan. I don’t want to rely on an Internet connection anytime I want to check an appointment or tweak a story.

This reliance on a network connection is one reason VDI (Virtual Desktop Infrastructure) hasn’t moved beyond niche markets like call centers, financial traders and other applications where users spend all day on strictly controlled desktop computers. In VDI, the operating system and application is hosted on a virtualized central server, which also stores the user’s data and other settings. Only the user interface runs on the local machine. While this can save a lot of money by centralizing management chores, the user is out of business if the connection to the server is cut.

MokaFive Inc., a three-year-old startup founded by a team of Stanford University researchers, has a way around this: Create a virtual version of the user’s PC, put it on a USB flash drive and let the user boot their own system (complete with their own  user settings and applications) from any compatible system whether or not it has a network connection. Their “PC on a stick” can get needed updates (ranging from new application data to operating system patches) automatically when the computer reconnects to the home server.

Well, maybe. I’ve written previously about MokaFive’s library of downloadable “LivePCs” that are created by enthusiasts to provide, say, a secure VPN client on a user’s desktop. That announcement struck me as exciting, since it taps the creativity of users to create new virtual appliances you didn’t know you needed until you saw them.  Their most recent announcement is about, essentially, capturing the user’s own PC and letting them install it elsewhere. This looks like a decent way to provide disaster recovery (as MokaFive customer Panasonic is doing.)

I’m less convinced by their claim that LivePC will “enable IT managers to centrally manage dozens to thousands of desktop computers across an organization.” LivePC lets administrators manage the LivePCs running round on all those USB flash drives, but how often does the average user need to run their virtual desktop on, say, a PC at a hotel business center rather than on their own notebook?

Central management also requires functions like discovery of remote devices, the ability to ensure changes are rolled out consistently to those devices, and an audit trail to ensure the changes were made. MokaFive may have ways around those problems, but didn’t focus on them in our briefing. To me, this looks like a better disaster recovery story than a centralized management story.

Face Book for Systems Administrators?

Imagine a social networking site where system and network administrators, database administrators and other IT "plumbers" could share not their latest book pix or movie reviews, but specific workflows and even configuration scripts for managing virtual servers, complying with security policies, or linking Oracle databases to popular ERP systems.

That's the vision -- or at least part of the vision -- for Tideway Systems, a six-year-old startup in the system and network management space.  Tideway competes against specific components within system management frameworks from vendors such as IBM Tivoli and HP.  It gathers and correlates information from OSI (Open Systems Interconnection) levels two through seven and identifies actionable information such as which databases are supporting which applications, and which application servers or storage systems are supporting specific business workflows.

The aim, of course, is to help organizations manage their IT functions more effectively and less expensively.  To promote their products, Tideway is taking a page from the social networking (and open source development) models by creating an on-line community called Configipedia. It hoping system administrators will share at least some of their best practices to spare each other time and trouble – and maybe become Tideway customers at the same time.

If any vendor has successfully used the open-source, community-development model to build critical mass for their products, I missed it. With the Linux operating system, the process was the opposite: Years of onesie-twosie adoption by enthusiasts eventually convinced larger customers to pay for “official” versions complete with paid support, which convinced for-profit vendors like Novell and Red Hat to jump into the market.

Will big company lawyers let their sys admins to turn over “best practices” (even those as lowly as system configuration tips) to their competitors via a vendor-owned Web site? Will those sys admins be scared off for fear of being locked into using Tideway’s tool in order to keep accessing its best practices Web site? Or is thus just another twist on every vendor’s online customer forums? Time will tell.

DataCore Gets Another $30 Million in Funding...

I can't take credit for it, but several days after I posted about the quality of DataCore's messaging the company announced another $30 million in VC funding. The SAN (and now, NAS) software vendor says it will use the money to fund product developoment, and to expand its gloal sales, marketing and distribution effforts. One key focus will be resellers who target Microsoft, Citrix and VMware virtualization-oriented customers.

DataCore's plans for the money make sense given several ongoing trends. 1) The push to sell to the small-medium business market through resellers, 2) The ongoing emphasis on virtualization and 3) The increasing importance of non-U.S. sales for domestic IT vendors.

Look, Ma, No Hardware

Adoption of storage virtualization still trails server virtualization. But storage virtualization is real and will catch on, for the same reasons that drove server virtualization: Flexibility, and lower costs for both storage products and ongoing management. 

That's where DataCore Software Corp. comes in. It has a strong, ten-year history built around a unique story of selling storage virtualization delivered through software that isn’t tied to any vendor’s hardware (supporting multiple flavors of server hardware as well as disk technology.) It offers a broad product suite ranging from SANsymphony on the high end to SANmelody on the low end. Its newest offering is SANharmony, which it claims allows network-attached (NAS) storage accessed via Windows servers to be managed by the same tools, and given the same capabilities, as the SAN storage already managed by existing DataCore software.

I like DataCore’s story of hardware independence, since customers like choice, independence and flexibility. They also look well-positioned for the upcoming convergence of server and storage provisioning. Finally, they’ve coined an eye-catching marketing term: "Virtualization 2.0,” meaning the ability to easily perform functions such as snapshots, thin provisioning and continuous data protection on “pooled” storage. That’s the kind of word-smithing more virtualization vendors could benefit from.

Microsoft Gobbles Virtualization Vendors

As Microsoft gears up for the launch of its own Hyper-V desktop hypervisor, it’s also scooping up other desktop virtualization vendors. It aims to be the leader in virtualization – the ability to split a single computing resource into multiple virtual environments – anywhere from the server to the desktop.

I’ve already written about Microsoft’s marketing of the application virtualization software it acquired from SofTricity. Earlier this month, it announced its acquisition of Kidaro, whose Managed Workspace product “allows enterprise data and applications to run within a transparent virtual machine wrapper," according to a blog maintained by Microsoft’s virtualization team. Just how this differs from the SofTricity capabilities is unclear, although the blog says “the wrapper provides ‘enterprise class’ management, deployment and a clean user experience.”

Along with security, look for management to be a key theme in upcoming desktop virtualization announcements. Microsoft, for example, is offering its newly-acquired virtualization capabilities as part of its Microsoft Desktop Optimization Pack (MDOP) services, which include diagnostics, group policy management and desktop error monitoring. This makes sense, especially in a recession, because the exploding use of notebooks, along with new and stricter security rules, makes the management of mobile PCs a huge and expensive challenge.

Virtualization Dot. Oh-Oh

Virtualization (the ability to split physical devices into multiple "virtual" devices, or create logical pools of physical devices) seems to be on my mind a lot of these days, maybe because vendors keep redefining it. Sometimes there are actual technical distinctions  which are worth keeping track of. But other times, it just seems like marketing blather, as with “Virtualization 2.0.”

If Virtualization 1.0 was all about saving money by consolidating servers and simplifying management, Virtualization 2.0 is supposedly all about "agility." The idea is that if you can quickly shuffle applications and other computing chores among “pools” of servers and storage, you can get new applications into the hands of your users more quickly and they can run out and do wonderful creative things with them.

For some reason, this set my BS detector off.  If Virtualization 1.0 has been working as well has it seems to have, why do we need a Virtualization 2.0 so soon?  Virtualization has been one of the huge success stories in the IT market over the last several years. VMware has grown from a little-known startup to a $1.6 billion subsidiary of EMC, and one of the leading platforms for software development.

Maybe the marketers touting Virtualization 2.0 fear they missed the boat on touting cost savings and feel they have to claim new capabilities to give you a reason to buy their products. Or, maybe the real situation is that virtualization is actually more complicated than we thought and people aren't really seeing cost savings from it.

In either case, vendors making the Virtualization 2.0 claim need to flesh it out a lot more. First, they must prove, with hard numbers, how virtualization helps companies roll out, update or manage applications more efficiently. Then (and this is more difficult) they need to prove how those applications have actually helped the business. 

If you're pitching a Virtualization 2.0 story to a trade pub, define what you mean by “agility” and then have specific numbers ready to prove not only how you speed application deployment and management, but how those applications help your customers make money more quickly.

Virtualization Collides With Services

Virtualization is a hot buzzword. So is service-oriented development. They can co-exist, but they’re not the same thing, regardless of what some vendors claim.

I found this out recently when I was tasked to write about Microsoft Application Virtualization (MAV), which it acquired with its purchase of SoftTricity. MAV downloads only the software components needed to run an application from a central server to the user’s desktop. There, they run in a secure zone and use the physical resources of the desktop (such as the processor and memory) but don’t touch the operating system or registry.

Because the application components are restricted to this secure area, there’s much less need for compatibility testing and a much lower chance of bugs crippling the user’s machine. This is especially important in very large, complex IT environments which need to deploy multiple versions of design applications to the same user. These older versions have reference libraries of calculations and product specifications which the user needs to, for example, revise the setup of an oil refinery which was designed using the older software.

That’s a decent story for an identifiable customer base. But Microsoft had me unnecessarily confused by referring to MAV as a way to “transform your computing environment into a dynamic, services-oriented infrastructure.” To me, a service orientation implies that individual application functions are available over the network to any application. MAV is actually closer to the ten year old -Java model which downloads only the application components the client needs, and runs them in what Java calls a virtual machine and which Microsoft calls a SystemGuard.

MAV is also different from Microsoft Terminal Services, a form of thin-client computing that runs only the user interface on the client, with the bulk of the code running back on the server. Suffice it to say that there’s a lot of different ways to slice an application these days, with more coming all the time. Yes, MAV does allow you to “serve” up application components as needed to the desktop, but application components aren’t the same as application services. They must be processed and handled differently on both the client and server ends of the transactions, and sending specific code to run on a client doesn’t (to me) meet the  SOA requirement of decoupling physical and virtual resources.

Such religious issues aside, mixing the service-orientation buzzword with the virtualization buzzword cost me hours in understanding MAV’s actual value. In marketing, branding or just explaining, don’t use two buzzwords where only one really fits.

Clustered Confusion

A recent assignment gave me the opportunity to plunge deep – maybe too deeply – into the world of clustered storage. Simply put, storage clustering -- like clustered anything -- means making multiple things (like storage servers or disk drives) work together to increase either their performance, capacity or reliability.

What did I find out? First, that there is no real market called “clustered storage.”  There are plenty of vendors who, when you look beneath the surface, do clustering but that’s not how they describe themselves, nor how most customers evaluate them. They want the benefits I described above – performance, capacity or reliability – and look for storage that will provide it most easily and cheaply.

Second, I learned that even the experts have trouble figuring out the subtle theological differences between, say, a cluster, a grid and a Global File System. You can slice the market (if you can call it a true market) by storage protocol (iSCSI vs. Fibre Channel, block vs. file), storage topology (SAN vs. NAS), and application (high-performance computing vs. backup vs. transactional databases, for example.) Even EMC CEO Joe Tucci got caught when he seemed to call an upcoming product clustered storage, forcing PR handlers to insist the new offering wasn’t clustered storage at all and wouldn’t compete with EMC’s existing Celerra platform.

Again, most customers – aside from those in the high performance computing market whose job it is to think about these things – don’t really care what you call it or how the product works three layers down in the technology infrastructure. (This is especially true of those I spoke with in the fast-growing small to medium business market.)

Having said all that, there’s obviously a whole lot of value being generated for customers or clustered storage products (no matter what you call them) wouldn’t be getting so much attention. In recent months we’ve seen NEC’s entry into the grid storage market, the IPOs of 3Par and Isilon, HP’s purchase of clustered file system vendor PolyServe, Sun’s purchase of the Lustre clustered file system and IBM’s purchase of Israeli storage grid vendor XIV Ltd. 

When you pitch your clustered storage story to an analyst or editor, you’ll obviously need more technical details than a salesperson selling to the local law firm. Here’s where it pays to dive into the weeds.

1)     Explain which resources are being clustered (i.e., NAS heads, disk arrays, etc.) and why (to increase availability, performance, reliability, etc.)

2)     Explain which types of applications the clustering is best suited for (high performance computing, backup and replication, video serving, etc.)

3)     And, as always, explain which other vendors your client competes with and why your client is different. I always recommend this to clients, and am always stunned by how many vendors’ Web sites and pitches are almost word for word identical.

For all its hype and mystification, there are some exciting technology and business stories in clustered storage, as older topologies such as RAID begin to reach their limits. Spend the extra time and effort to make sure your client gets the attention they deserve.

Will BORG Replace RAID?

About a year ago I stumbled on a company called CleverSafe, which had a promising technology for dividing data into chunks and distributing it to commodity servers across the Internet. Its secret sauce was the mathematical algorithms that reconstitute the data as needed, even if one or more of the servers failed. Not only did this dramatically reduce the total amount of data that needed to be stored, CleverSafe claimed, it improved security because it’s impossible to rebuild the data from any one node.

Electronics giant NEC has now taken the basic concept and run with it as HydraStor, which also splits data among general-purpose servers but manages it as a single easily scalable, redundant, self-managing and self-healing pool of file-based storage. One intriguing point is that neither of these companies rely on RAID to ensure recoverability from the failure or one or more disk drives. (This is another sign that the RAID architecture may be reaching some limits, especially when it comes to the complexity of setting up RAID arrays in the first place and the time it takes them to restore data.)

Like RAID, HydraStor does store parity data from which individual data chunks could be recovered if a drive failed, but the parity data is stored on accelerator or storage nodes rather than on a RAID array. The advantage? If you want or need fast recovery from a failure, you simply add more accelerator nodes to give you more processing oomph to rebuild the lost data, rather than having to rely on the same RAID controller you need to process requests from your business applications.

In a recent chat, John Matze (one of the authors of the iSCSI protocol and now the vice president of business development at IP SAN vendor Hifn, compared the emerging storage model to the Borg in Star Trek, a race of cyborgs with a collective consciousness which can easily adapt to, and learn from, the failure of any one individual. As bandwidth becomes cheaper and individual storage nodes become more intelligent, he argues, we’ll see the rise of more storage models like that of the Google File System which run on inexpensive commodity hardware and take the occasional (or more than occasional) failure in stride.

Will the Borg’s signature phrase – “Resistance is futile” – hold true as this new model of cluster/grid/distributed storage runs up against RAID? Stay tuned.

Rise of the Virtual Machines

If you thought you knew what virtualization was all about – VMware letting data center admins run thousands of virtual machines on hundreds of physical servers -- browse on over to Moka5 to get your mind blown.

Moka5 is one of a number of Web sites offering free, downloadable virtual environments for your client PC. By virtual environment I mean the whole shebang – operating system, applications, supporting database – running as a virtual machine alongside your regular OS and applications. Its library of “LivePCs” includes everything from environments for creating secure VPNs to a Christian version of the Ubuntu open-source operating system, complete with customized parental controls and the WhatWouldJesusDownload toolbar.

And lest we forget servers, there’s a whole slew of downloadable server-based “virtual appliances” downloadable – again, for free -- from sites such as Jumpbox. Let’s say you’re a start-up software company and need a real low-cost (read: free) development management tool. Download one from the Web and run it as a virtual server on the lead developer’s notebook. Need a free customer relationship management system? Download one and run it in a virtual machine on your receptionist’s or CFO’s desktop, and just make sure they leave it running when they go home.

These start-up vendors (who hope to make a living selling higher-end versions and support for their appliances) point out that beefing up a host PC to run a virtual appliance is a lot less expensive than buying dedicated server and a high-priced commercial application. And as if all this wasn’t enough, you can store many of these “virtual appliances” on a CD-ROM or even on a USB drive. This lets you use the virtual appliance off-line (although you need to hook to the Internet to get patches or updates) and to instantly load the virtual appliance on another system.

So this is all pure goodness, right? Well, no. First, you get what you pay for, and a free virtual appliance may – or may not – be properly configured to keep your data and the host PC secure. It may or may not deliver adequate  performance, may or may comply with regulations (such as those governing customer privacy) that could sink your business, and may or not be able to share data with other enterprise applications. The last thing a large organization wants are dozens or hundreds of unknown, unmanaged virtual appliances lurking on users’ hard drives or their desk drawers, just waiting to launch onto the network and do mischief.

So what does all this mean for tech marketers? First, when you talk up the benefits of virtualization you also need to talk about how to manage the resulting sprawl of virtual machines and virtual appliances. Second, when you talk about virtualization, extend the conversation beyond the first-level benefits such as cost savings to talk about agility – the fact that virtualization can quickly deliver new application functionality so users can get more work done for the business. (Microsoft is among those thumping this theme with the Softgrid Application Virtualization) technology is acquired from Softricity last year.)

Above all, get past thinking about virtualization as only being important on the server, only coming from VMware and only being about hardware consolidation.