Tech Trends

Veritas Cuts Through the Tech Jargon

Bob Scheier — Mon, 14 Jul 2008 07:22:00 -0700

Should manage storage in a virtualized environment at the block or file level? Normally, that’s the kind of distinction that can (and maybe should) get lost several layers down in a technical evaluation.

But Veritas did a good job of making it a differentiator in a briefing about their Veritas Virtual Infrastructure (VxVI,) due for delivery later this year.

VxVI (now there’s a market-friendly acronym for you) uses the open-source Xen hypervisor to deliver to the virtual world some of the capabilities Veritas so painstakingly developed over many years in the storage world. Among the challenges it tackles is making sure that, as virtual servers (which are not linked permanently to a specific physical server) retain their connections to the right kind of storage devices as the virtual machines are moved from physical server to physical server.

Aimed squarely at enterprise data centers with hundreds of servers, VxVI goes right up against VMware and its VMFS (Virtual Machine File System). Veritas claims that managing data at the level of blocks, or volumes, rather than at the level of files (as VMware does) makes it far easier to assign and maintain connections among specific virtual machines and specific levels of storage. The volume-based approach, Veritas claims, also makes it easier to provide other storage goodies such as multipathing, which automatically finds a new route for data if, say, a switch fails, and mirroring.

It also laid out very clearly, in chart form the benefits it claims for its block-based approach.

In lesser hands, the question of block vs. file could have sounded as obscure as whether to have red or white wine with pasta, or the benefits of variable-valve timing in an engine. But by linking volume-based storage provisioning to clear customer benefits, Veritas made a convincing case.

ILM Isn’t Dead

Bob Scheier — Thu, 03 Jul 2008 10:23:15 -0700

Some buzzwords die a quick and painless death; others mature to describe actual market trends and others linger in rehab for years.

ILM (information lifecycle management) is in that in-between stage, deried by some as a myth but still flogged regularly by the likes of IBM, Oracle, HP, and Sun. ILM means creating various tiers of storage, some cheap and slow (like tape) others fast and expensive (like Fibre Channel disk). As data becomes older or less valuable, you move it from the more expensive to the less expensive tiers, saving money will ensuring you can access your most important data as needed.

Those who bad-mouth ILM say it’s too hard to do in real life, but often the problem isn’t managing the multiple storage tiers, but figuring out which data should be on which tier – more of a business than a technical issue. From what I see in my reporting, the rationale for ILM is too compelling for the idea to die, and I see many companies implementing it on at the least a tactical basis.

One example is a recent story on virtual tape libraries (software that makes a bunch of inexpensive disks look like a tape library, so you get the speed and reliability of disks without having to replace your backup software or retrain your staff.) I expected to find that customers were using VTLs to replace tape because of the high physical handling costs and unreliability of tape.

What I found is that they’re using VTLs to create a new tier of storage, below the Fibre Channel disk that is still required for business-critical transactional systems, but above the data that is so old or seldom used that it can safely be sent off to tape. Production data might be moved to the VTL at the end of every day, where it is stored for anywhere from a day to several weeks. Only then is it moved on to tape, where it is stored for either archival or disaster response reasons.

A number of vendors are using technology such as massive arrays of idle disks (which power down disk drives when they’re not in use) to bring the cost of disk-based VTL closer to that of tape. Still others are introducing technology that fine tunes the process of deciding when to move data among tiers. Asigra, for example, combines MAID with a discovery tool that (according to the company) scans each server to adjust requests for data and input for data based on the customer environment. This tool helps administrators define different tiers of data and how many versions of that data should be kept in which tier according to parameters such as how often they are used.

Whatever you call it, and whatever technologies you use to implement it, the idea of putting data with different values on different levels of storage makes too much sense not to do. Vendors that market such technology shouldn’t be afraid to play up the “tiered storage” or ILM theme, as long as they explain clearly how they make it feasible.

Pass Me The Bottle, Igor

Bob Scheier — Wed, 18 Jun 2008 08:29:59 -0700

Being a mere mortal and not an IT professional with thousands of servers and Tbytes of disk lying around, I don’t often get to actually play around with the software vendors pitch to me. That’s why I was intrigued when Check Point Software Technologies Ltd. –famous for their ZoneAlarm firewall -- briefed me on ZoneAlarm ForceField, which virtualizes the Web browser to isolate it from the rest of your PC or notebook and keep the browser from sneaking malware onto your system.

The kind folks at CheckPoint pointed me to their free trial and I was off and running, feeling like a real cutting edge type. But soon thereafter, like a character in a horror movie who wishes he hadn’t drunk the Secret Potion, I began getting mysterious system slowdowns and crashes. Since my system had been incredibly buggy anyway, I uninstalled and reinstalled ForceField several times to see if it was the culprit. When the problems returned with each installation, I had had enough and returned to the mundane world of applications running right in Windows.

My other concern was that I was never really sure how much ForceField was protecting me from. It did show me, on request, how much extraneous data it kept off my system (90Mytes in several days) as well as how many exploits it had blocked. But given the performance problems it seemed to cause, that was small comfort.

Suffice it to say I’m staying out of the lab for now, at least where my everyday work system is concerned. My advice for desktop virtualization vendors is first, do no harm to my system, and second, make a really convincing case of what you’re saving me from.

Server Virtualization: Worth the Hassle?

Bob Scheier — Sat, 31 May 2008 18:18:02 -0700

I've been doing a lot of reporting about data center consolidation lately, which means I've been talking to a lot of IT managers who are virtualizing servers. Almost all of them report seeing big savings in the capital expense of buying physical servers as well as the operating expenses of heating and cooling them.

When I asked if virtualization has reduced the costs of managing their servers, their answers are a mixed bag. In some cases, it seems, server virtualization actually makes management more complex, and thus might not be right for all customers.

I got an interesting perspective on this recently from Dave McCrory, chief technology officer at Hyper9, a 25 person startup in Austin, Texas which is scheduled to come out of stealth mode this month with a virtualization management product. He says he hears from customers that server virtualization doesn't make things unmanageably complex as long as the environment remains fairly static once the customer has, for example, virtualized 100 physical servers down to 30 physical servers.

Where management costs rise to troublesome levels, he says, is when the customer continually makes a lot of changes in the virtualized environment -- say, adding or removing servers, or moving applications among virtual servers as business needs change. (Not surprisingly, that's exactly the challenge that Hyper9 hopes to meet.)

For years, everyone from industry analysts to every major hardware and software vendor have been preaching about creating true, dynamic IT services, where business users can "dial up" processing, network and storage capacity as business needs change. Virtualization helps make that possible, but I get a headache imagining how to track virtual servers accessing virtualized storage over virtualized local and wide area networks. Just for starters, how do you make sure critical data and applications are secure when the underlying “plumbing” is in one big virtual pool? And how do you quickly find and fix problems when something goes wrong?

Regardless of whether Hyper9 has the right answer, they’re certainly asking the right question. Look for plenty of competition in this space as customers move beyond static, one-time virtualization and try for the Holy Grail of a true dynamic IT infrastructure.

Bare-Naked Metal

Bob Scheier — Mon, 26 May 2008 13:27:56 -0700

“Bare-Naked” is one of those tautologies which makes no sense (being completely redundant) but it sure gets the idea across.

The computing equivalent is “bare metal” and refers to the physical hardware that make up a computer. In a typical off-the-shelf PC, the “bare metal” talks to the BIOS (Basic Input Output System) that provides basic communications among components such as the microprocessor, disk drives, and system memory. The BIOS in turn talks to the operating system (such as Windows) and the operating system talks to the applications.

“Bare metal,” then, can be thought of as the bottom, foundation layer of a computer, whether that computer be your notebook computer or a corporate server. Get ready to see some virtualization vendors argue that the “bare metal” is the best place to run their hypervisor software that coordinates multiple virtual machines, rather than running the hypervisor higher up the stack atop the host operating system.

MokaFive, for example, is using BareMetal™ as a trademark for a customized version of Linux running directly on the PC hardware, “making it optimal for security-critical applications,” the company says. Look for other vendors to suggest running multiple VMs on the same physical desktop, with one VM dedicated to running security applications that protect the other VMs.

It makes intuitive sense that the closer you run to the hardware, the better you can secure it. But vendors still need to prove very precisely why and how “bare metal” virtualization is better than, say, virtualizing a plain old operating system that has been hardened by switching off unused ports and services. And with security apps running so close to the hardware, they’ll also need to prove how they’ll restore a system when a hacker – or simply an uninformed user – messes up the software running on the bare metal. If hackers can break into the code that runs elevators and thermostats, they can do it for a hypervisor.

So if you’re in the virtualization space and want to play up the “bare-metal” be prepared to flesh out your message.

PC on a (USB) Stick

Bob Scheier — Mon, 12 May 2008 07:29:16 -0700

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?

Bob Scheier — Sun, 04 May 2008 12:16:51 -0700

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...

Bob Scheier — Fri, 25 Apr 2008 12:43:25 -0700

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

Bob Scheier — Tue, 22 Apr 2008 13:02:18 -0700

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

Bob Scheier — Mon, 31 Mar 2008 12:48:56 -0700

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.