ML 10/30/01: NTAP, EMC, industry trends:
We think it is important to put the recent Dell and software
announcements in context as EMC is executing a clear
strategy here. We have long argued that intelligence,
complexity, and, therefore, margin are migrating to
network. This is not any network but data center networks
that we think will ultimately be three SAN’s—a storage
area network, a server area network, and, ultimately, a
system area network. The long term value add in these
three SAN’s and the products to which we see margin
migrating are three software stacks—a network OS stack, a
virtualization stack, and a management stack. In turn, we
have argued, that this industry evolution will produce
secular storage hardware pricing pressure and force EMC
and most other enterprise vendors to go through multi-year
transitions.
In EMC’s case, we have argued that this means
transitioning EMC go to market and sales model away
from high-cost direct sales and building new software for
the 3 SAN’s.
Automated Information Storage—Critical Seeds Sewn
If EMC executes, the products in the Auto IS
announcement could be the backbone of the management
stack that we think will be one of the keys to enterprise
leadership.
Yesterday EMC announced four products: 1. ECC Open
Edition, 2. WideSky, 3. ECC Replication Manager, and 4.
ECC StorageScope. While we discuss the architecture in
much greater detail in a longer piece that we will shortly
publish, we want to cut to the chase as well.
Over the next 90 days, the different products will be
released and, if they deliver as promised—always the key
“if” with software, this could be a very important step for
EMC to reassert leadership. Management is one of the
critical, enabling elements to making SAN’s real. EMC
strategy is to provide holistic management of servers,
storage, switches, and software that will eventually
comprise the 3 SAN’s.
Don’t Get Bogged Down in API’s or Interoperability—
Remember What Happened with SRDF
While many rightly point out that this depends on
cooperation from the server, switch, software, and non-
EMC storage vendors, we think that this misses a
fundamental, but subtle point. Device discovery,
monitoring, and other rudimentary function can be
achieved without the cooperation of the product OEM.
And yes richer interoperability can be achieved with
reverse engineering and R&D spending. But we should
remember what happened with SRDF when EMC started
to make RAID intelligent.
SAN’s, when made manageable, can often more than
double disk and server CPU utilization and drastically
reduce the headcount required to run and manage systems.
In the joint piece that we published with McKinsey, we
found that people costs account for nearly 45% of the total
cost of ownership for direct attached storage. If EMC can
make SAN’s manageable and accelerate their adoption,
storage and server vendors will face the unhappy choice of
working with EMC to drive deeper integration and ceding
the management opportunity to EMC or of refusing and
seeing their hardware do poorly as SAN’s are deployed.
Remember that server vendors “chose” to play ball with
EMC on SRDF because their customers wanted the
functionality.
And also like SRDF, there is a noticeable hardware cost
savings. In small SAN deployments, we have seen disk
utilization jump from the 20 to 35% range to 70% and
server utilization jump from 25% to 50%. These savings
coupled with people costs can reduce the total cost of
ownership by over half. In our report, “The Storage
Report—Customer Perspectives and Industry Evolution,”
we saw the cost per managed terabyte drop from $0.80+ to
the $0.30 to $0.35.
While the move to SAN’s does not mean good things for
storage or server hardware, we think the migration will be
inevitable. That EMC has rolled out the beginning of a
management stack that should not only spur SAN adoption
but grant EMC an important control point is important.
Yes, it may not spur sales of Symmetrix and it might
reduce sales for SRDF, it also shows that EMC recognizes
that the subsystem battle is over and that its is focused on
being a leader in the next wave of enterprise computing.
What About the Competition
We have also long argued that three vendors, Brocade,
EMC, and Veritas, seem uniquely positioned to deliver the
three software stacks (network OS, management, and
virtualization) for the 3 SAN’s. While we expect the
competition to be long and drawn out, EMC’s intentions
here seem a bit more clear. The Auto IS architecture, if
successful, would effectively transform fabric management
into a niche task and would limit Brocade’s addressable
market. Similarly, it would transform volume
management and tape back up into niche tasks as well.
Next, if Auto IS takes of as the management standard,
EMC may have the opportunity to differentiate its
hardware again. EMC hardware would presumably be
better integrated and offer better performance. But
transitioning out of that business model once proved a bit
difficult.
This architecture would also nicely position EMC to begin
delivering virtualization capabilities for the 3 SAN’s. And
owning two of the three key software stacks for the next
wave that will be dominated by the 3 SAN’s ain’t such a
bad thing.
Market Size
People cost is currently about 45% of total storage cost of
ownership for DAS environments. While no software
product will eliminate this, it is important to note that
people cost exceeds hardware costs. In theory, the market
for the three software stacks ought to exceed the $45
billion storage hardware market. More important,
however, is what the 3 SAN’s will look like. These
platforms would comprise storage, servers, networking
gear, and the software that runs them. The market for
managing them must be quite large and robust.
-----------------------
Field Check Feedback:
• Pricing remains brutal and will likely start being
quoted in $’s per Gigabyte instead of pennies per
megabyte.
• The additional capacity, processors, cache, and ports
in the new Symmetrixes are helping EMC’s price
performance but it is still ugly. Our contacts have
reported buy-ins going for mid-teens per megabyte
and guaranteed upgrades selling in the $0.11 to $0.10
range.
• IBM pricing of Shark remains very aggressive—have
seen as low as $0.07 per megabyte.
• EMC will price as low as necessary on hardware, but
the software picture is more mixed. Three CIO’s did
report getting bundled and, therefore, discounted
pricing on software. But, a high level consultant who
advises customers on the pricing of roughly 40 deals
per week disagrees—EMC has refused to discount
software.
• Most continue to believe that EMC maintains its lead
in software, but some have argued that HDS True
Copy and HARC have begun to narrow the gap with
Time Finder and SRDF. The debate seems lopsided in
EMC’s favor at this point.
• Roughly half of the 40 EMC deals are EMC versus
HDS. There were some HDS versus Shark inquiries
earlier but the HP Compaq merger announcement
killed them. While he hasn’t seen EMC lose deals to
Sun/HDS, he’s getting feedback from buyers that
suggest that it may be happening.
• In dedicated Sun environments, many of those we
contacted suspect that Sun will be able to shut out
EMC and protect its installed base. Similarly, outside
of the Solaris installed base, they see little chance of
Sun/HDS gaining traction. Admittedly, this was
before the EMC Dell deal.
• It also appears that Sun/HDS may be able to sell the
code that HP developed as part of its OEM
arrangement with HDS. Since HDS licensed HP’s
microcode, Sun will be able to resell it.
• While HDS is compelling for Sun, Sun’s long wait
and other failed attempts at storage seem to have hurt
is credibility. The sales force has to argue, “we got it
right this time, not like the other 6 efforts we tried to
sell you.”
• We also heard from one sales contact in EMC that the
company did not aggressively close business last
quarter since it new the quarter was already shot.
-----------------------
Network Appliance
Our Stance On The Stock:
• Dominates NAS, but we see four risks for
commoditization:
1. Subsystem margin pressure
2. Lack of enterprise penetration
3. Poorly positioned to deliver the three software
stacks and needs a stronger data center presence.
4. Poor distribution and market reach
• Its TCO proposition could attract interest in the short
term.
Field Check Feedback:
• Heard that NTAP took some business from EMC in a
major European telecom account.
• NTAP plans to introduce richer capabilities to support
and manage block and file data formats.
• NTAP’s NT consolidation business continues to do
well and we heard of one large financial services deal.
• The fight with Microsoft over NAS support with
Exchange continues. While Microsoft clearly wants
to fight any other OS, like NTAP’s WAFL, IBM
Global Services does help customers use NTAP filers
to do consolidation without certification or
compatibility.
• Continue to hear about some nice improvement in the
database market, but the sell there still remains
missionary.
Impact:
While these data points don’t impact our long-term
concerns about commoditization, distribution, and
positioning in the enterprise, we are warming to the story
near term. We are close to anniversarying the weak
demand environment and the easier compares. Moreover,
this environment is clearly motivating some accounts to
look at filers as lower cost alternatives to DAS and other
storage implementations. NTAP may finally be in a
position to make numbers and not have to reduce guidance
going forward. |
