FalconStor CEO Expounds on Layers of Networked Storage Virtualization

By Elizabeth M. Ferrarini

One virtualization vendor has shown signs of taking off like the jet it is named after. Founded in 2000, FalconStor, based in Melville, New York, has delivered on a software solution that can virtualize in an any-to-any networked storage environment. In fact, the product's very first sales quarter yielded $3 million for the publicly held company. Meanwhile, FalconStor has chalked up relationships - either reseller or technical collaborators - with both major storage vendors and networking vendors. This list includes IBM, Intel, ADIC, and Cisco.

Most of FalconStor's competitors can only virtualize in fibre channel storage area networks. FalconStor's IPStor, which runs on the user's server, provides a means for either IP or fibre channel networked servers to virtualize any block-level storage pool connected to the IPStor server. Steve Duplessie, senior analyst at the Enterprise Storage Group, Milford, Massachusetts, says, "They're one of the few vendors that realizes it's not about SAN versus NAS versus iSCSI, it's about supporting users regardless of the technologies they select."

Experienced leadership which know how to stake a claim in storage innovations underscores this company's foundation. ReiJane Huai, FalconStor's CEO and one of the founders, is the former CEO of Cheyenne Software, which was sold in 1996 to Computer Associates for $1.2 billion. In the early 1990s, Huai served as the chief architect for Cheyenne's ARCServe, the first client/server storage management product. ARCServe become the catalyst for the $1 billion server backup and restore marketplace. ReiJane recently put the mettle to the pedal by answering questions about making virtualization the underpinning for new storage management architecture - one based on networking, instead of servers.

After your success with Cheyenne Software, why not just retire to the Bahamas rather than start another storage company?

First of all, if you sit in the sun to long, you get skin cancer. Starting this company came easy for the founding team since most of us worked for Cheyenne. We had a lot of past experience running a storage software company. We knew how to organize the people, identify the appropriate technology, define the right architecture for the product, and get the product developed on time.

Before we started the company in 2000, you heard a lot of talk about fibre channel being the only way to attach a storage area network. We disagreed with assumption and, instead, said the storage industry is evolving. Our vision for FalconStor is this: in addition to fibre channel, other networking technologies, such as IP-based networks, can provide efficient conduits between the storage and the server. These conduits will take us from legacy systems to the new generation of Infiniband networks.

How do you folks differ from Nishan Systems, DataCore Software, and others?

These are all good companies, but our vision is greater than theirs. First of all, we're visionaries who created the idea of server-based backup back in the early 1990's. So, we understand how to create an infrastructure for a paradigm shift to network-based storage management. If you put storage on the network, the server must be responsible for computation. To this end, storage management tasks, such as backup, can be offloaded from the servers. Thus, you need an architecture to support a decentralize model. So, we're building a super duplex club sandwich in which virtualization becomes one of the supporting layers. Other layers include replication, and backup and recovery.

What's your core competency and what makes it possible?

Our in-band virtualization enables you to aggregate the storage and create virtualize storage, which is independent of a particular vendor, an interface, or a specific operating system. The network becomes generic. It can be IP, fibre channel, or whatever you want it to be. For example, we can take EMC storage and present it, as well as take anyone else's storage and present it. That's our core competency.

The module architecture makes the product really shine by accomplishing many important IT functions. We, not only virtualize storage, but can provision the storage, and integrate it in a similar fashion with the OS. You also need the ability to administer the end points. We're the only company that has the ability to scratch the end points from storage all the way to the OS. That's why we can enforce network security, and security between the OS and the storage. Virtualization requires this capability. If you can enforce load balancing and automatic failover, you can do automatic configuration and policy-based management. We're building these key enablers into network-based storage infrastructure.

What do you classify your product as?

Our software can run on either a Linux-based machine or a Solaris-based machine to create an appliance. Think of our product as an in-band processor for storage. If you open a storage device from EMC or Hitachi, you'll see two layers. The first one is a storage processor; the second one, is banks of disks. We're separating the storage processor from the disk. The decline of computing costs and connectivity costs have become driving forces for the network-based storage model.

Since we're a vendor-agnostic software company, we can partner with any hardware and or software vendor that wants to embed our technology, or we can develop tailored add-ons for a particular vendor's product, such as Microsoft's Exchange.

Doesn't in-band virtualization create various islands of appliances, which, in turn, could create bottlenecks and load balancing problems?

No, not if you look at our statistics. We hired a third-party laboratory to benchmark the results of our latest product release. The lab. director couldn't believe the results. Our network disk is faster than the local disk.

Cheyenne Software also pioneered hierarchical storage management (HSM) Can your product handle HSM?

We have an add-on product similar to HSM. We call it Capacity on Demand. Since HSM is notorious for configuration problems, we knew how to avoid them. Based on specific policies, our product, which can run on Solaris, will migrate the compressed contents of the stale files to a scratch area, which we control. By owning the storage pool, we offer the benefit of both compression and migration, enabling the available storage to still stay below a certain threshold. You can make the disk larger by calling our backend server. The file system can reset itself without affecting the application.

Can you highlight the cost benefits provided by your product's different layers?

Our infrastructure provides several areas, such as backup and volume management, that can yield some compelling cost benefits for you.

Most people do backup today by going from server to server. Therefore, each server carries its on cost for a software license and management overhead. Now, if you can separate the storage from the server and aggregate the storage, you can reduce the cost to approximate one server entity. Our infrastructure allows any off-the-shelf backup product to perform backup on a Linux-based or Solaris-based appliance. This technique allows you to buy one copy of backup software instead of 30 copies. We can perform the job faster than server-based backup. Since we control the storage, we can feed the storage much faster than through the operating system.

With server-based storage, if you want to make the disk larger, you buy another disk and put volume management software on top of the disk. Volume management is the only way you can grow the disk capacity. On the other hand, we provide you with a virtual disk which you can grow indefinitely without any additional software.

We can also take replication and snapshot services off the server and move them onto our appliances across a platform. Again, we can deliver these services at a higher speed and with less licensing overhead costs.

Your infrastructure also addresses some specific application areas, such as Oracle. Can you describe how your Oracle agent works with Oracle Replication Services?

Our agent works in parallel with Oracle Replication Services, which performs replication at the database record level. This's how we distinguish ourselves from the other virtualization players. They claim they can take snapshots as the database runs. You can't perform a snapshot of the active database without running into a data integrity issue. This's an area a lot of companies don't make clear. Since we know the importance of backing up the database without having to shut it down, our database agents enable you to take a snapshot while the database is active.

We provide the snapshot at the network level without taking ownership of the disk. Our agent just sits on your machine and waits until you want to do a snapshot. The agent acquires the database temporarily and makes sure it's in a consistent state. Then the agent lets you know if you can proceed with the snapshot. The snapshot will have 100 percent transactional integrity. We perform replication the same way.

Do you plan to embed your technology into switches? You're going to see a lot of companies embedding our technology into their products. MTI's disk array has our technology, as well as Accton's gigabit Ethernet switch. We'll also contemplating adding another layer to our virtualization model to help you administer the SAN if you use Cisco switches.