| Michael
McManus, DIGITIMES, Taipei
[Monday 9 June 2008]
LSI was founded in 1981 as an ASIC company
that did service functions, which meant basically it turned
others companies' IP into silicon. The company did that for
more than 20 years, but in the years from 2000-2005 it seems
that the company sort of lost its way. It had great
technology but it was participating in a lot of different
markets and there was no clear corporate strategy.
However, what the company has done over the
past 24-30 months has been to significantly recast its
overall composition. LSI changed everything from its product
strategy to manufacturing – changing the composition of its
board, making four to five acquisitions, two divestitures
and undergoing a multi-billion dollar merger – with these
changes for the most part being geared toward steering the
company toward a focus on storage and networking.
Just before Abhi Talwalkar Taipei 2008,
Digitimes had the opportunity to ask LSI CEO Abhijit (Abhi)
Talwalkar to comment about the company, the economy and key
issues in the storage and networking markets.
The growth of digital content and its
influence on LSI
It's no secret that there is no end in sight
for growth in digital content in the market, whether it's
consumer oriented or for businesses. And with that growth in
digital content there are increased requirements in far more
capable network infrastructures, whether it is in
corporations or in the domain of service providers and
carriers. So all the moves we have made in the past two to
three years have been to aggressively and decisively
participate in this significant opportunity.
But just to recognize an opportunity is not
enough. We wanted to do it in a way whereby we had
leadership in the market. So scale was an in important
consideration for our market focus. In the semiconductor
industry, it is difficult to maintain market leadership over
multiple product-generations without scale.
Today about 80% of our business is in storage,
where we participate in four segments including HDD silicon,
where we are one of the top players in the market. We also
have a full line of SAS (serial attached SCSI) products,
which is a market that we helped create and drive growth,
and where we are now number one worldwide. We are also
number one in terms of fiber channel silicon, the fabric
that makes up the fiber channel storage area networks. We
also participate in storage subsystems and RAID subsystems,
where we are number one worldwide as a building block
provider.
The other 20% of the company is networking
associated, where our focus is geared at the strains that
networks are facing today. Networks continue to be built out
and IP traffic continues to double at a phenomenal rate so
networks need to be far more intelligent, and we will see
networks evolve over the next several years. We know that
because we are experiencing it today.
But all of the work we've done over the past
two years has paid off for us in the last three quarters.
Our operation and financial performances have exceeded
expectations from analysts and investors. We've
significantly cut our operating expenses without impacting
our opportunities for growth and our balance sheet is very
strong. The hard work is starting to pay off.
Hype in the economy
We are all watching our various markets and
demand signals as closely as we can. But I would say that
coming into this year there was far more hype than the
reality dictated – hype created by the media and financial
analysts, and the accompanying herd mentality seen on Wall
Street that goes with that. It helped to create a big
separation between the hype and the reality.
Yes, there have been pockets of softness, but
I think certain product categories are feeling a bigger
impact than others. If you look at the US especially,
consumers (unlike the rest of the world) had not been used
to dealing with gas prices at US$4 per gallon, so spending
on discretionary items may be seeing an impact, such as in
the consumer electronics (CE) segment. People can put off
purchasing a camera, a new TV or a handset. So, I think some
of our peers in the semiconductor industry that are more
exposed to the CE market may be facing more softness.
But in the case of servers and storage systems
and even in the PC markets, demand has held up relatively
well, and whatever softness there has been in the US market
has been compensated by continued growth in Europe and Asia.
But even if considering the possibility of a downturn in the
US, we have already run into companies that have told us
that their competitive edge is based on technology – with
some companies in segments like the banking industry
indicating they spend 10-12% of their revenues on technology
deployment – and that they plan to invest heavily in
technology in the event of a downturn and then come out
stronger than their competitors when the markets picks up.
Also, there are a number of areas where
companies can't put things off. Areas such as storage, and
to some degree networking, tend to be areas where companies
will continue spending – you can go into an economic
turndown, but you are not going to stop producing data or
creating digital content, are you?
Considering the network side of the market, I
recently returned from Europe where I spoke with the big
telecom equipment companies such as the Ericssons and Nokia
Siemens. Their message was interesting. Some of them believe
that the amount of traffic they are going to see on their
networks, specifically mobile traffic, is going to grow
5-10x in the next three to five years.
So companies are going to invest in things
such as networking and storage and I would say that LSI has
been a bit more buffered than our peers because of that.
Another thing to consider is that if you look at our
revenues, by the time our products are consumed, 70% of our
sales are coming from outside the US by virtue of our
customer base.
Taiwan and the storage market
If you look back at the experience of Taiwan
ODMs becoming involved in the PC and server industries, it's
amazing the things that have happened once Taiwan got
involved, both for the industries and for the volume
economics that were created. And it's only been a matter of
time before Taiwan became involved in the storage industry.
But the storage and storage systems markets
have a different volume economics curve compared to other IT
markets. To put it into perspective, using rough estimates
there are 1.2 billion handsets that ship every year and
Taiwan is pretty ubiquitous in that market in terms of
developing handsets. In desktops and notebooks, 250-300
million desktops are shipped each year and 90% of them come
out of Taiwan in some capacity. Then you have the server
market, where over the last seven to ten years Taiwan's
participation has dramatically escalated. My guess is that
70-80% of the world's server shipments – at least single
processor or dual processor servers – are in some shape or
form coming from ODMs in Taiwan, and about 10 million
servers ship a year.
So if you look at the different markets, the
numbers (shipment volumes) come down. And in storage
systems, you are probably in the order in the one to two
million units per year. And the market is not as horizontal
a market as the PC and server markets are. That's why it has
taken a little bit longer for Taiwan and its machinery to
get engaged.
I think you are going to see an increasing
level of participation from Taiwan in storage – definitely
in external storage systems and primarily in the volume
space. However, we are not going to see it in silicon and we
are certainly not going to see it in hard disc drives. I
think it will be more in the space of storage enclosures and
array controllers.
The overall storage systems market is a US$60
billion market. It's about the same size as the server
market, but within that total you have software services,
HDDs and systems. The systems market, once you pull software
and services out, is about US$25 billion. It's a good size
market and there is more of a profit pool there than there
is in the other segments, and that is bringing Taiwan and
some of its ODMs, as well as some of the traditional OEMs,
such as the Dells and HPs (Hewlett-Packard) to increasingly
add more focus on storage.
One example of how the storage industry looked
to engage Taiwan over the past couple of years was through
the development of the SBB (Storage Bridge Bay)
specification, which LSI and various members in the storage
community such as Dell, EMC and Intel developed to drive
standardization in the entry-level external storage market.
The specification really just defines the mechanical and
some of the electrical interface between the array and the
electronics packages. But the objective was to have a
standard so that Taiwan Inc., because it has tremendous
ability in enclosures, would be able to develop many
different standard enclosures that developers of storage
systems and storage controllers could use. This is something
that has been happening for the past one to two years and
has helped accelerate the adoption of technologies such as
iSCSI and SAS.
Migration in the HDD market to SSDs
The development of the SSD market is a highly
debated topic because there are a lot of people that have a
lot to protect and there are a lot of people that have a lot
to gain. Everyone has his or her arguments. You will
definitely see adoption of SSDs in notebooks – it's just
questions about timing and diffusion rates that are still
under debate. My personal guess is that we will still see
single-digit penetration rates over the next several years.
Although there are certain PC applications that don't need
more than 64GB or 128GB of memory, the premium is still
significant.
Another thing to consider is that an
increasing number of PCs are becoming the notebook variety,
and people are putting more of their personal content in
notebooks, such as music, photos and videos. We are also now
moving into an era of high definition content as well, so
users are going to need far more capacity than 128GB on
those notebooks. So although there are arguments for both
sides, I'll repeat that I think the penetration rate is
going to remain in the single digits over the next several
years.
Nevertheless, LSI is participating in the
market because we believe we have some unique capabilities
to bear specifically in notebooks in terms of flash
controllers technology. I don't think it's a need in the
first generation or two but notebook drives will reach the
performance requirements pretty soon. We have all the IP,
and as more performance requirements and higher density
levels of MLC flash are needed, the market is going to
require a greater level of signal processing capabilities.
And to that, we can bring our HDD expertise, which is all
about deciphering between noise and a bit on storage media.
We can bring that expertise to bear on the issues
that will arise in the SSD space, and we are in a better
position than many people in the industry because of that
expertise.
Although LSI has opportunities in the notebook
segment, we are at a particular advantage in the enterprise
because we touch every aspect in a server. We are currently
playing with the controller itself, because if you have an
SSD, the other end of it is will be a SAS controller of some
kind and typically there is a RAID stack that is involved.
These are markets where we are the number one player
worldwide. In the server environment, our business touches
upon the entire data path, and we have the software
capabilities as well. So in terms of interoperability
validation or driving features of security, we have all the
pieces that our silicon competitors just don't have.
In terms of SSDs in the enterprise over the
next few years, we also expect the same single digit
penetration rate but there are going to be applications in
the enterprise where people are going to be willing to pay
just about any premium, because you are getting the 5-10x
for the I/O performance.
Tough for OEMs
It's not clear who is going to ultimately win
in the SSD market in terms of providing completed SSD
devices. If you fast forward three years, perhaps there will
be 50-75 million SSDs that ship, compared to an HDD market
of 600 million units. Who is going to be a predominant
supplier of those SSDs? HDD vendors such as Seagate have
certainly been vocal about participating in the flash memory
market. Perhaps memory makers such as Intel, Micron and
Samsung will dominate. Even OEMs, which have expressed an
attitude of "why not participate" since they can get flash
controllers from LSI and flash from the memory makers, can
do their own SSDs if they want.
In my personal opinion, I think OEMs might
have a tough time gaining traction for enterprise
applications. In notebooks it will all be integrated so they
are going to play a bigger role, but in the enterprise
customers will be buying the drives discretely and then
adding them to systems, and I don't know if users are going
to be happy about being locked into only being able to buy
SSDs from their system OEM. They prefer the model they have
today, where there are three to four different brands of
drives to choose from. If there is a situation where you
can't take a Dell SSD and put it into an HP server or vice
versa, end users are not going to stand for that, so that
may be a barrier for the OEMs to participate in the
enterprise.
Network security and traffic management
In addition to networks seeing an increased
amount of traffic, more and more real time services are
being offered over networks, with the best example being
video. For standard traffic, you can drop bits and the data
can eventually get there intact and the integrity is
maintained. But for real time applications, you can't do
that, so networks have to be able to discriminate more
intelligently as to how much bandwidth to allocate to
certain applications that are going through the networks so
they get higher priority. QoS (quality of service)
capabilities have to go up.
To address this, LSI has a full lineup of
network processors that are mid-range to high-end network
processors which have a lot of intelligence to do advanced
traffic management and advanced traffic control and various
forms of QoS capabilities they lend to the applications. In
addition we became more involved with deep packet inspection
late last year, when we purchased a company called Tarari.
The whole objective deep packet inspection is
that at line speed you can crack open the IP packet and look
inside. You can search specifically for worms and viruses or
you can crack open that packet to look for what kind of
information is in that packet. If it's a video packet maybe
we need to let the application layer 4-7 know about it so it
can direct that traffic and/or provide a certain amount of
bandwidth for it.
Currently, a lot of software companies that
target these applications simply leverage x86
microprocessors. One of the reasons we acquired Tarari is
that in a very small piece of silicon (one square
millimeter) it does what 12-14 Xeon microprocessors can do
at real time speeds and much more cost effectively.
Security on networks is going to be an
increasing issue for the market to contend with because you
are seeing an increasing number of connections on the
Internet. One key driver of that growth is the mobile
market, where you have carriers that have historically had
closed systems, but those systems are now opening up to more
IP services.
Beyond security, vendors are looking to
develop appliances for content based billing appliances or
working on appliances that allow IT managers to discriminate
about how to allocate bandwidth on the network, and we are
working with the market to develop such solutions.
Today we use Tarari as a coprocessor but we
have plans to take its capabilities and move them into our
network processors. In fact we have plans to take its
capabilities and integrate them into our DSPs that are
geared toward voice and video processing in network
equipment, as well as putting them into our storage
processors, as certain storage environments are trending
toward becoming IP based networks.
Developments of next-generation SAS
Although only implemented starting a few years
ago, SAS is ubiquitous today across servers and is growing
in terms of its adoption in storage systems. The current
generation of SAS supports a transfer speed of 3Gb/s but the
industry is now working on SAS at speeds of 6Gb/s, but there
are lots of different silicon that has to be there. HDD
vendors have to be there with a SAS 6 interface for their
drives in volume and the RAID software has to be hardened
and developed. All this has to come into play before you
start seeing any kind of volume ramps. We expect that to
occur in the latter half of 2009.
Taiwan is very engaged in SAS in the server
space. For the next generation of SAS in storage subsystems
or external arrays I think the opportunities will be there
for some of the ODMs to start scaling up. Two to three
Taiwan ODMs have been building their storage capabilities
over the past few years and we look to them for certain
parts of products.
But still, Taiwan makers mostly target
enclosures and perhaps controllers in the low end of the
market. If you look further up the market, such as to RAID
stacks that have been hardened, the Taiwan makers haven't
had the level of road miles or enterprise hardening yet. |
