Tech Trends

The Limits of RAID

Over time, every technology ages and starts to cause its own problems. That's happening now with RAID (Redundant Array of Independent Disks.)  For years, the various flavors of RAID have been one of the most cost-effective ways to boost disk performance and to protect data that an organization can't afford to lose.  Raid stripes data across multiple disks to either increase the speed with which data can be read from or written to the disks, or to ensure no data is lost if even one disk fails.

But as the amount of data that can be stored on the average drive rises to hundreds of gigabytes, or even a Tbyte or more, the amount of time it takes for a RAID array to recover from a disk failure can stretch into a day or more.  Customers can't wait that long if they need that data to accept orders, ship products or keep their e-mail system up.

A number of vendors in the iSCSI space are claiming workarounds to this problem.  By creating mirrors of LUNs (logical unit numbers) FalconStor's FalconStor IPStor can continue to run as a primary storage device (keeping data available) even while a failed disk is being rebuilt. The high end of Hifn Inc.'s Swarm SAN appliances use Raid 61, which creates two sets of parity data, either of which can be used to rebuild data from a failed disk.  This allows the array to remain functioning even while rebuilding not one, but two, failed desks.

My guess is that like most "legacy" technologies, RAID will never fade away, but just be enhanced and improved over time. That means some of you will get to pitch "RAID is dead" stories to editors, and others will get to pitch "RAID isn't dead after all" stories.  The good news is that either way, you can make a good, convincing case -- as long as you're very, very clear about exactly which RAID-related problems you're solving, exactly how you're solving those problems and of course why your approach is better than the competitors. And don't forget those all-important reference customers, either.

Virtualization for Mere Mortals

VMware sells an impressive suite of software for creating multiple "virtual machines" on a single physical machine.  But a wave of recent briefings around VMworld shows just how much room VMware has left for its software partners to exploit, as well as some of the challenges they face explaining their products.

First up was FalconStor software, with a version of its CDP (continuous data protection) software it claims can eliminate the need for backup and restore devices for small to medium businesses running VMware. Called the FalconStor Continuous Data Protection Virtual Appliance for VMware, this software boasts a starting price of below $8,000 and an install time of five minutes or less.  Because it instantly captures every change in the system image of each virtual server, Falconstor claims it can do a bare metal restore of the virtual environment.

Falconstor positions this as a new capability, called Continuous Data Availability (CDA). This is a nice reframing of the more commonly used CDP term which focuses on what customers really want -- to be able to get to their data, not just recover it.

A second vendor, Innowave, called next with potentially interesting technology for increasing the number of virtual machines a customer can host on a single physical server.  Their  specialized algorithms constantly learn which data blocks a specific server or application is most likely to request from a disk array, and proactively stores those blocks in system memory on the server.  Since memory is significantly faster than disk, the company says, this eliminates a key performance bottleneck and lets more virtual machines run on each physical server, reducing costs for the customer.

Finally, Left Hand Networks called to talk about their Virtual SAN Appliance, which is a version of their SAN/iQ software for managing and clustering iSCSI SANs. Their pitch is that once you've virtualized the CPU and memory of your servers into logical pools, the Virtual SAN Appliance will also virtualize the disk drives within those physical servers.  This solves the problem (which I didn't know existed) of physical disks  "stranded" on otherwise virtualized servers.  Left Hand says their Virtual SAN Appliance is a good alternative for smaller firms which can't afford a full fledged SAN, or who want to evaluate VMware implementations without buying a SAN.

If I've lost you with any of this virtualization jargon, you see the challenge reporters and editors face.  If you've got a virtualization product to pitch, build the groundwork first by reviewing the "state of the art" in current virtualization and then carefully explain how your product makes things work better.  Explanatory diagrams also help, but keep them big, easy to read and uncluttered.

When a technology is as new as a virtualization, be sure to at least cover the basics of what you're talking about -- especially when you're educating an editor or an analyst who isn't as close to your "solution" as you are.

Finally: Hardware Makes News (Security)

Stillsecure sells its network and security software modules either as preconfigured appliances or as software, running in either case on off-the-shelf Intel or AMD-based hardware.  The combination, the company says, can replace security appliances built around custom microprocessors to deliver networking functions such as routing, DHCP and DNS as well as security functions such as firewalls and content and spam filtering.  Admitting that appliances based on specialized hardware are still at the core of very complex networks, Stillsecure for now targets less demanding markets such as remote offices and SMBs.

The cost benefits of off-the-shelf hardware make a powerful story.  So does the potential for deploying these modules on virtual servers so an administrator can deploy more CPU power to, say, firewalls or routing as network conditions change.  This also taps into two other trends: a blurring of the lines between network operations and network security products, and the trend of greater integration between network operations and network security staffs.

In the case studies or white papers based on these trends, I'd like to see lots of details about the network, the amount of traffic each piece of hardware is handling, and the number of networking or security services running on each server.  Kudos to Stillsecure for admitting that custom hardware is still best for the cores of the largest networks (candor helps build credibility.)  Finally, for products targeted at SMBs, don't forget the need to also communicate to the resellers who serve these smaller customers.

Finally: Hardware Makes News (Storage)

For years, the only news in hardware was old news.  Microprocessors, buses (the internal design for routing data within a computer) and disks got faster, faster and faster.  But the real innovations came in these software that ran on this physical foundation.

That may be changing.  The latest chip and bus technologies handle so much data so quickly, and at such low cost, that they can deliver quantum leaps in not only performance, but in cost and power consumption.  And that makes for some interesting stories.

agami Systems Inc. is using AMD's 64-bit Opteron processors and Direct Connect technology (which links the processor directly to networks and storage) in what agami claims is the first network attached storage device "to break the GB/second performance barrier while reducing the total cost of ownership."  Besides the AMD technology, the agami Information Server is built around the 64-bit agamiOS NAS (network attached storage) operating system and agami's file system replication software.  As a result, agami claims its Information Server provides the "only integrated real-time file system replication" for disaster recovery and high update availability failover and offers other goodies such as lower power consumption.

Microsoft has sold its Windows storage server software to hardware OEMs for years, but it always seemed those OEMs had an uphill battle differentiating their products with the relatively narrow range of chips and bus designs available to them.  With 64-bit chips like the Opteron and new interconnect technologies like Direct Connect, companies such as agami have a lot more options and a more interesting story to tell.

Vendors with stories like these should (as agami did) be very clear about which of their technology is off-the-shelf, which is homegrown and why their combination of the two is better, cheaper, or faster than the competition.  It also helps, of course, to explain which types of applications and what size customers will benefit the most from these products.