Interviews

posted on 26 May 2008 10:34

ExaGrid is a deduplication vendor and lots of us were surprised when Dell set up a deal with it to resell its product.

B&F talked with Bill Andrews, ExaGrid's CEO, about deduplication, Data Domain and Dell. This interview took place immediately before EMC announced its deduplication backup products using software OEM'd from Quantum which is why there is no mention of the new EMC DL3D products.

B&F: Tell us a little about ExaGrid.

Bill Andrews: ExaGrid was founded in 2002 in Westboro, Mass. It originally built deduplication in primary storage. I joined in 2005 and had the software redesigned to work in the backup world.

There are five ExaGrid models with disk:  EX 1000; 2000; 3000; 4000; and 5000. The number refers to disk sizing with the 5000 being 5TB. It's all about the number of weeks of retention and we're equivalent to Quantum and Data Domain.

Customers can mix and match six of our EX models in a grid and asymmetrically. It's all NAS-based, not VTL (virtual tape library). The backup server targets a NAS (network-attached storage) share.

There's also the ExaGrid (diskless) gateway.

Customers can replicate an EX product's data from one site to another. Some 49 percent of our customers with two sites have 100 percent shut tape off. The first box has the last full backup plus weeks of byte-level deltas. The second box has just the byte-level differences. You could keep years (of data) on it.

Our customer count is going up and up. From Q4 to Q1 (2008) Data Domain's new customer count went backwards because Quantum and ExaGrid started to slow them down. We're winning 73 percent of the time against Data Domain; we're lower priced; we scale better, we're faster at backup; and we're faster at restore - better value at a better price.

Customers want an appliance that they can afford.

Data Domain gives them slower than disk performance at a higher price that slows as it grows in capacity.

We have 44 user stories on our site, more than all of our competitors combined. There are another ten in the works.

B&: Would you compare and contrast ExaGrid and Data Domain please?

Bill Andrews: Data Domain did the first deduplication at block-level, taking 8KB blocks and comparing them and tossing the repeated ones this making disk blocks as affordable as the price of tape. They do inline deduplication and so it slows down backup. Data Domain didn't look at backup; it was just storage.

We entered the market and took the view that there was more to it. Customers want low cost and fast backups, to fit into a backup window. The first thing we did was to let backup data run into the box at the speed of disks. We deduplicate post-process and we're 30-300 percent faster than Data Domain for backup.

Secondly, we keep the most recent backup in its entirety but compress it two to one. Most restores come from it. We can restore these much faster than Data Domain with its gazillion 8KB blocks.

Three; we blow Data Domain away on tape copy. Most customers make this offsite tape from their Friday night full (backup). Data Domain has to reconstitute it from their gazillion 8KB blocks.

We have the same approach as Diligent  in the Fortune 500 world; we're byte-level deduplication not block-level (and) store  100 meg segments; it's better for disk fragmentation.

If there is 1TB of data, a hash table has 125,000 times 8KB entries. If there is 8TB there are 1 billion entries in the hash table. They (Data Domain) don't have a scalable system. All they can do is add disk shelves to the head processor, which is a single point of failure.

If you add disk shelves the processor is fixed and has to do more work to dedupe the extra storage. As you scale their performance degrades.

We have servers - processor, RAM, capacity and bandwidth - in a grid. We bring more processor, memory and bandwidth with every server addition. We flatline and don't degrade performance (as we scale).

B&F: How about ExaGrid and NetApp?

Bill Andrews: We don't know much about them; we just don't see them. We see Data Domain all the time.

We sell to customers with 10 - 60TB of data (and) see Data Domain six times in a week, Quantum too. We're in the 5 to 10K employees, the mid-tier space. Maybe NetApp sells higher.

B&F: How about ExaGrid and Dell?

Bill Andrews: Dell bought EqualLogic - iSCSI SANs - and sells to SME too. You could have Exchange and SQL accessing SAN storage as well as a backup server. But you backup data over and over and it kills the SAN disks. With dedupe it only stores the changes so Dell has ExaGrid as a gateway via an iSCSI HBA to the iSCSI SAN. The backup server links to the ExaGrid which has no internal storage.

You could have the Friday night full (backup), say, 5TB compressed to 2.5TB, and the previous nine backups in nine times 100GB of byte-level deltas for ten weeks of backup data on the system.

Our dedupe ratio is about the same as Diligent and Quantum - the data that changes is the data that changes.

We do lots of co-marketing with Dell.

Comment: The Dell deal is a great win for ExaGrid and expands its market each greatly. ExaGrid is pretty focussed on Data Domain and convinced that Data Domain's approach is inferior to ExaGrid's. If Andrews is correct about Data Domain's new customer count falling because of ExaGrid and Quantum's entry into the same market that Data Domain sells into then EMC's entry should strengthen that trend.

Incidentally, Bill Andrews says there are lots of people from EMC in ExaGrid - it's just across the street.

^{[Chris Mellor.]}