Unfortunately, Christensen continues to ignore the fact that EMC remains the standard for disruption in the storage industry and that Michael Ruettgers has burned the boats behind his troops TWICE to address a bigger opportunity.
He did this with original Symmetrix which allowed EMC to go from less than 1% market share of an $8B mainframe storage market in 1990 controlled by IBM with 85% market share (and 80% gross margins) to 41% market share of a $4B mainframe storage market in 1995 -- IBM's market share had shrunk to 31%.
He did this again with the open system Symmetrix which allowed EMC to go from 0% market share in 1994 to 35% market share in 2000, from $1.9B in sales in 1995 to around $9B in sales in 2000.
IP-based networked storage may be the next opportunity and nobody is better prepared for this opportunity than EMC. EMC's first software product is SRDF, which was released in 1994. It remains the popular standard for syncrhonous and asynchronous mirroring with the SRDF over IP version setting a milestone when it allowed a customer to accomplish transatlantic remote mirroring using its advanced IP network. Network Appliance is still unable to provide syncrhronous mirroring and its asynchronous mirroring technology is remarkably primitive in comparision to EMC's SRDF.
Touts like to promote NAS and Gigabit Ethernet as the next disruptive technology that will bring EMC down. Unfortunately, a closer look at the facts will show a different picture.
Fast Ethernet (100 Mbps) remains the corporate network technology of choice. 1.9M Gigabit Ethernet (1000 Mbps) ports out of a total 54M Ethernet ports were sold in 1999. 4.0M Gigabit Ethernet ports were sold in 2000 and 19M Gigabit Ethernet ports are expected to be sold in 2003. The standard for 10 Gigabit Ethernet is expected to be ratified in early 2002 with some pre-standard products expected to ship in late 2001. Unlike Fast Ethernet and Gigabit Ethernet, 10 Gbps Ethernet will utilize switched Ethernet switching and require optical cabling.
Currently, a Gigabit Ethernet port is 6x more expensive than a Fast Ethernet port and a 10 Gbps ethernet port is expected to be at least 10x more expensive than a Gigabit Ethernet port, which means that it's going to be at least 60x more expensive than a Fast Etherent port. In practical terms, that means that in the same amount of time that will take a 1 TB database to grow to 16 TB (100% a year for 5 years), most corporate networks will still be making the transition from Fast Ethernet to Gigabit Ethernet.
Keep in mind that Fast Ethernet and Gigabit Ethernet are highly probabilistic networks that typically operate at less than 30% of wirespeed so any networked storage solution that depends entirely on this network will have a limited market. It took the NAS market, for example, more than 15 years to develop into an $822M market in 1999. EMC does not expect that NAS market to account for more than 10% of its total addressable market in 2005.
NTAP's WAFL is not disruptive at all. EMC's IP4700 is already cheaper and superior to NTAP's filers because it achieves HA (high availability) in one box while you have to cluster two NTAP filers to achieve high availability. The IP4700 uses EMC's Snapview technology which provides up to 92 read-only snapshots compared to NTAP's Snapshot technology which provides up to 31 read-only snapshots. Celerra plugs off a Symmetrix which utilitizes TimeFinder technology which provides up to 8 read and write full copies. I don't think anyone will seriously compare NTAP's asychronous mirroring technology with EMC's SRDF which provides synchronous and asynchronous mirroring over at least 8 network protocols.
One of the most profound advantages that EMC has over NTAP, and the rest of the industry for that matter, is that it invested early in a proprietary quality assurance process that insures that its products are the most reliable. Not only does EMC maintain close engineering links with their key suppliers , but they subject components and the final system configurations to rigorous stress testing that takes at least 3 weeks. Nobody else comes close. NTAP, for example, has a business model that generally depends on increasing the turnover rate from the current 11x. EMC is already very profitable at 4x because 80% of revenues come from customers who appreciate the quality of its products.
That ultra-reliable platform also allows them to quickly add features and maintain their lead over the competition. EMC's Enginuity Operating System, for example, now contains at least 2.2 million lines of codes and allows up to 144 parallel operations concurrently inside a Symmetrix. With the increased popularity of clustered servers, look for EMC to build quickly on Enginuity to develop a SAN operating system (with an embedded clustered file system) that will fully exploit the rich bandwidth of a SAN to provide rich block level and file level services targeted at the fact that storage management accounts for up to 12x the cost of storage hardware. |
