I have fond memories from my childhood of Rube Goldberg contraptions. I was always amazed at how he would creatively use common elements to implement these crazy machines. By using every day items for complicated contraptions, he made even the simplest process look incredibly complex and difficult. But that was the beauty of it, no one would ever use the devices in practice, but it was the whimsical and complex nature of his drawings that made them so fun to look it.
Image courtesy of rubegoldberg.com
It is the in the context of Rube Goldberg that I find myself thinking about the EMC DL3D 4000 virtual tape library. Like, Goldberg, EMC has taken an approach to VTL and deduplication that revolves around adding complexity to what should be a relatively simple process. Unfortunately, I don’t think that customers will treat the solution with the same whimsical and fun perspective as they did with Goldberg’s machines.
You may think that this is just sour grapes from an EMC competitor, but I am not the only one questioning the approach. Many industry analysts and backup administrators are confused and left scratching their heads just like this author. Why the confusion? Let me explain.
Let’s start with some background, EMC shipped their first VTL using FalconStor software in April, 2004. The product was called the CDL and was made up of VTL software from FalconStor and a CLARiiON disk array. Since that time, EMC has grown the product line to include support for the DMX series and creatively renamed the product line the EDL (EMC Disk Library). Fundamentally, there has been little change with the product with the exception of higher density drives, CLARiiON upgrades and faster performance gained from riding the curve of ever improving server I/O improvements.
The next wave in VTL is deduplication and you would think that EMC would naturally look to FalconStor. Nope, they decided to partner with Quantum for deduplication. Like Rube Goldberg, they took a device that was once relatively simple (VTL with FalconStor) and decided to make it inordinately complex by adding in an entirely separate technology; to make the platform sound more impressive, they renamed it to the DL3D 4000.
Now you have FalconStor AND Quantum technology in the same device, the DL3D 4000. (Note: There are lower end devices using only Quantum technology. Talk about confusing!). The FalconStor and Quantum technology inside the DL3D 4000 are entirely separate. They require separate storage and management and have radically different performance profiles. (FalconStor is roughly 8-10x faster than Quantum!) EMC used to advertise high availability with the EDL through FalconStor, but that claim now goes out the window since Quantum only supports a single node. The design also fundamentally impacts restore performance. Restoring data becomes a two step process. 1.) Restore from Quantum disk to FalconStor disk, 2.) Restore from FalconStor disk to the client. (In virtually all other deduplication implementations data is restored directly from the deduplicated disk.)
In the end, I have a tough time figuring this out. EMC has hundreds of talented developers, almost $5 billion in cash and have purchased companies with VTL (NearTek, undisclosed amount, September, 2006) and deduplication (Avamar, $165 million, November 2006) expertise. Yet with all of these resources and money in the bank, this is the best that they can come up with? Oh well, on the bright side, they have created a solution that Rube Goldberg would be proud of. Perhaps they should change the name to the DLRG 4000!