You don't have to be Einstein to realise that interoperability is
the way forward for storage products, and market leader EMC is
showing the way with new innovations. Nicholas Enticknap
reports.
EMC is the world's largest specialist IT storage company, selling
disc storage subsystems plus associated software. According to
analyst firm IDC, the company is number one in the storage area
network (San) and network attached storage (Nas) markets, and is
also top in the direct-attach markets for both Unix and Windows
NT/2000 servers. Gartner Dataquest figures show that EMC is the
largest supplier in the storage management software market.
EMC started life in 1979 selling add-on memory for IBM mainframes.
The success story did not begin, however, until the launch of the
Symmetrix Integrated Cached Disc Array (ICDA) in 1990.
Up to this time mainframe disc subsystems were quite different from
those used on other systems. They were larger, with massive 14-inch
platters; they provided much faster access to data; they were
engineered for maximum reliability; and they were much more
expensive. The key to EMC's success was its recognition that the
performance and reliability levels demanded by mainframe users
could be achieved using off-the-peg 5.25-inch drives.
EMC provided the necessary reliability by mirroring. It achieved
the performance required by using much larger caches than competing
products. This recipe worked, and EMC's effective selling and
marketing brought the company to market leadership by 1995.
EMC has maintained this position. First, it recognised that
mainframes were increasingly being replaced by large Unix systems,
and adapted Symmetrix for them in 1995. Second, it recognised that
disc systems were developing into subsystems, and introduced
software products to add functionality.
The first of these, Symmetrix Remote Data Facility (SRDF), for
mirroring over distance, appeared in 1994. Another major product is
Timefinder, launched in 1997. Both have sold well, as has the
Controlcenter automation product, with more than 40,000 copies
installed.
However, contrary to the claims of its marketing machine, EMC did
not pioneer open systems attach, or the addition of software
functionality. Encore led the way in multi-platform connectability
with the Infinity subsystem, introduced in early 1995. IBM
introduced software including peer-to-peer remote copy on its 3990
disc controller in the late 1980s. The first product to offer
Timefinder's capability was StorageTek's Snapshot, introduced in
1992.
EMC's success here stems not from pioneering but from its timing
and then its selling. It read the market right on each
occasion.
EMC was also early into the Nas market. Here its product is
Celerra, which is effectively Symmetrix with a different front-end
to permit file sharing. The company enhanced Celerra with the
introduction of a software product called Highroad at the end of
2000. This allows a Nas system to operate at the speed of a San
where possible.
Celerra, like Symmetrix, is a top-end product. EMC did not enjoy
anything like the same success lower down the scale until it
decided to acquire Data General and its Clariion disc subsystem
range in 1999. EMC has continued development, with the latest
Clariion being the FC4700, introduced in January 2001. However, EMC
is still a long way from the market penetration it has in its
top-end markets, with just 5% of the global mid-range market.
One successful outcome of its acquisition of Data General has been
the development of a Clariion Nas product, known as the IP4700,
which was launched in December 2000. Sales of this product,
together with Celerra, elevated EMC to number one in the Nas market
during 2001, as measured by both IDC, with 42% market share, and by
Gartner Dataquest, with 48%.
It should be said that erstwhile market leader Network Appliance
disputes these findings. "If you look at EMC's Celerra, one of the
ways they have been able to report substantial sales is that they
count the whole Symmetrix as revenue, whereas typically only a
small portion is used for Nas," says Stuart Gilks, Network
Appliance's Northern Europe SE director.
For future growth EMC is looking to its first truly innovative
offering since Symmetrix. This is Centera, a type of product that
EMC calls content-addressed storage, which was launched at the end
of April (see right). EMC's prospects for future growth depend
heavily on market acceptance of Centera. But this will take time.
"This year Centera will not account for 10% of our revenue, but it
will be significant," says EMC's president and chief executive Joe
Tucci.
The major challenge facing the company is the need to demonstrate
that its disc subsystems are truly open. "EMC has to get more into
the open market," says Robin Burke, principal analyst at Gartner
Dataquest. At present, software products such as SRDF are
proprietary: you need a Symmetrix at each end.
EMC recognises that openness is now a market requirement. "We are
now opening up our software from a management perspective to other
people's storage devices," says EMC's technology analysis director
Karl Steinhardt.
EMC launched an initiative called AutoIS (Automated Information
Storage) last October to implement this. AutoIS is based on a new
version of Controlcenter called Open Edition, which is a framework
for plugging in various automation products. A key element, called
Widesky, is middleware that will allow external suppliers of both
hardware and software to interface with EMC's products, as well as
allowing EMC software to run on non-EMC systems.
EMC continues to prosper, and will continue to figure heavily in
companies' storage plans, it seems.
The next growth engine?
EMC's new Centera system is
designed for the storage of fixed content - data items such as
cheque images, medical scans and movies which never change. EMC
says this type of content accounts for an increasingly large amount
of the data stored online, and estimates that it will represent 50%
of all data stored online by 2005.
The innovative feature of Centera is that the address of each data
object stored is calculated from the data content of that object.
This contrasts with traditional systems, where a data object's
address describes only its physical location.
Using the content to calculate the address brings several potential
benefits. For example, applications do not need to know anything
about the storage device in use. Also, it is easy to tell whether
the data stored has become corrupted, because the algorithms used
to calculate the address can be rerun at retrieval time, and should
produce the same answer.