Ethernet Switching

Ethernet switching has been an obvious NBT since the concept was introduced to the world by Kalpana.   When combined with the “fat pipe” concept (originally Fast Ethernet”, it really made sense.  Many companies were acquired (mostly by Cisco as they shifted from being all about routers to being about end-to-end networks) at the 10/100 stage. When it was time to go to 100/1000 (Gigabit switching), combined L2/L3 switching was thrown into the mix as well.  Foundry and Extreme were spawned and still survive today.   Many others were acquired and the vestiges live with-in some big companies today.

Ethernet switching success is mostly attributable to the interoperation with the installed base of Ethernet hubs and previous generation switches, enabling very smooth migration.

Historically, each Ethernet generation has been an NBT on the switching side, but not the NIC side.  

10 GigE

If it worked a couple of times before it should work again, right?  Probably wrong. The “need for speed” isn’t there today, slowing down adoption considerably.  The applications that may drive the need for 10GigE  - Metro Ethernet, Ethernet-based storage networks, and high speed server connections - are slow to develop. Also, we may have crossed over a line in terms of copper/fiber physics which will drive most 10 GigE to be fiber-based and therefore more costly.  This creates an awful situation for infrastructure companies – a customer set actually waiting to see the need for a new generation of technology before adopting it!  As a final factor for start-ups, Cisco has decided to be the price leader in 10 GigE!  So you don’t even get to pick up the crumbs they have decided to leave on the table. 

Generally a big company game as they add 10 GigE to existing platforms and benefit from customers buying those platforms in anticipation of going to 10 GigE in the future as opposed to actually doing it. 

Metro Ethernet

I think it’s going to happen.  I really do.  But will the carriers who offer these services find enough of what they need from Cisco?  Nortel?  Alcatel?  I think so. Maybe those vendors need an acquisition to freshen up their offering.  But maybe not.  

From an end user’s point of view and maybe from a service provider’s point of view, I still think this is an NBT.  But I don’t think the world needs any more equipment companies launched into this space.

Content Distribution Networks

They were hot. They were dead. Now they seem to be coming back a bit.  Sometimes technology is like that.  If it hangs around long enough it eventually finds its niche.  Usually just in time for people to propose a new NBT to replace it. In this case, the niche may be related to streaming content distribution.  PVRs are probably the biggest threat to this opportunity as they represent a residential point for caching and are already being embraced by MSOs and satellite TV distributors.   

This market tends to be defined as Akamai and a small number of pretenders to the throne. In the future, it will probably be more about Time Warner and Comcast and Verizon and SBC.

Content Switching

Also known more pedantically as Server Load Balancing (SLB), Content Switching was a NBT for a few months in 1999 as Cisco and Nortel paid exorbitant prices for Arrowpoint and Alteon respectively.  The technology is still very important and may form part of the basis for Grid Computing. But the market suffers from the products being too powerful and low priced.  Their “virtual” nature limits the number of boxes needed per customer.  Pricing has been an issue ever since Alteon priced their capability as a low cost adder to their L2 switches to try to make those switches more attractive.  Foundry followed with their “anything you can do we can do cheaper” pricing strategy for dogging Alteon. $50,000 products became $7,000 products overnight.  The market size has been modest ever since. 

Many new aspirants in this space have misread the issue as one of scale and are now history. In reality, the issue is features and down-scaling solutions so they can be deployed cost effectively in conjunction with Enterprise applications.

Lighting won’t strike twice here.  The vendors who have “gotten it right” as far as adding value to Enterprise applications, like F5 Networks, Redline Networks, and Netscalar are hardened survivors and won’t leave holes open for new entrants. 

2.5G/3G/4G cellular

So much has been written about this it seems silly for me to pontificate much.  The issues are coverage, applications, cost-of-spectrum, and cost of infrastructure upgrade, which is a bigger issue in the US than a lot of other places.  In the US, we have so many other alternatives for communication, the imperative has not existed to aggressively build out packet cellular. 

2.5G/3G is a big company game.  4G/5G is still a potential NBT. But the market won’t start in the US.  See next month’s Musing.

IP over Optics

Remember when SONET was dead, to be replaced by various forms of IP-over-optics?   In the long run, this still makes sense to me. The earliest deployments may be as part of the backbone infrastructure to support fiber-to-the-home or fiber-to-the-curb. It remains to be seen how carriers will dice and slice the spectrum to deliver services to the home. If done completely with packet multiplexing, IP-over-optics will be needed all over the place.  If done with a mixture of FDMA, digital modulation (QAM), and TDM, then not.  This will take a while to play out.

Once the next generation residential service delivery architectures are figured out, there may be NBTs here.  Timing is critical and tricky.  Also, there may not be one market here but several (carrier vs. MSO, US vs. Pac Rim, etc.) 

Lambda Switching

Wavelength switching still makes a lot of sense in the future as the carrier migrate their networks to the next level.  Some day there will be a need for the types of things that Monterrey networks and BrightLink were building. 

This will be a “Phoenix” NBT some day.  It’s a market timing issue.  Now is not the time. 

PON

If you believe in the ability of Sprint and other carriers with no access to copper and coax subscriber access to drive a market, then PON will be huge. In reality, it WILL be a fairly big deal eventually.  But the economic cost of keeping vendors in play until it really takes off makes it unclear whether this is a good area for incremental investment or not.  This is a market that is easy to enter prematurely. 

Whoever benefits from PON adoption, if it gets adopted at all, is either already in play or is a big company that picks up the pieces later.   

VoIP

Voice over IP WILL happen.   But it won’t take the world by storm.  It will ooze into our consciousness, as it already is. 

Mostly a big company play, except for some peripheral niches. 

iSCSI

I’ve gone back and forth on iSCSI.  At first I thought it was completely silly.  Now I see how it can be used on a small scale to back end NAS heads and to enable new uses of DAS/NAS combinations.  But at the end of the day, it’s a protocol, not a market.  Start-ups building systems products around iSCSI (HBAs, JBODs, etc.) may get some early traction if they represent the only choices on the market early on, but these products are almost by definition commodities right out of the shoot so you better figure out how to succeed on 10-20% gross margins as a start-up, which is almost theoretically impossible. 

iSCSI-to-FC translation is a niche market.  Most SANs and SANlets will be one or the other.  

When the market becomes large enough to be interesting to them, this will become a game for the big boys and for low cost manufacturers. 

There may be some arms dealer opportunities of the 2^nd generation type if the 1^st wave of suppliers dies on the vine waiting for the market to happen.   

SATA-based storage

SATA and SAS-based storage will be huge!!  Almost all storage will be accessed via these technologies in a few years.  But it is still disk drive economics.  A few pennies in manufacturing cost will make or break companies. 

It’s a low cost manufacturing game combined with big company channel leverage.

Storage/Data Management

The reliable and cost effective management of data competes with security and improved supply-chain-management (and other business processes) and the number 1, 2, and 3 things on CIOs minds.  Improvements probably need to come from start-ups who can afford to think out of the box. The problem is, the issue is too critical to trust the solution to start-ups.   This has created a Catch-22 which is slowing down the pace of investment in this space and therefore the delivery of new solution options.

Remains an NBT for start-ups, but they need to ally with bigger companies early on and probably flip into one too. 

Grid Computing

The term grid is used in a lot of different ways.  There are also equivalent terms like utility computing, flexible computing, and virtual computing being bandied about.  In general, marketers are trying to avoid the term “clustering” because “that’s so 90s!”.  But, basically, gird computing is about dynamic clustering – moving resources in and out of service for delivering specific applications as user demands shift throughout the day, week, quarter, month and year.    

Although many are looking at Layer 1-3 technological solutions for Grid computing, the bigger technical roadblocks are at the OS/application level – the biggest of which is how to quiesce and existing application so that the resource can be freed up to start to deliver a different application. 

Once that gets solved, the adoption of more flexible computing models with-in enterprises and ASPs will start to ramp up.  However, the market opportunity tends to get over-forecasted here.  Only certain servers and services will be added to “the grid” in the early years of adoption – distributed databases, applications that can be clustered, and NAS serving such applications.  Many servers will continue to operate “off-grid” until there is a compelling reason to move them on.  Any projection that assumes more than 1,000,000 servers will be grid connected or part of HPC in 2007 is probably too high.

Here are opportunities here.  But so many technologies from so many different angles need to come together, it will be slow to develop.  Start-us need to be well funded to wait out the market and they’ll need Tier 1 server vendor cooperation to get to the end customers.  That adds up to acquisition by server vendors and server vendor wanna-bes, I think.  

Next Gen Cluster Servers

We haven’t really had the successful introduction of a new server vendor into the market for many years.  With the convergence of the concepts of blade servers, function-specific servers, and compute/network integration via Infiniband, PCIE, or RDMA/TCP (choose your favorite), there have been a spate of business plans in the high end server space recently.   Egenera’s success to date is creating hope for others. 

I don’t know.  There can’t be THAT many applications that require the scale of processing of a single data set that brokerage companies, weather forecasting agencies, and gene splicing drug companies need. 

Probably the Cray type market space is up for grabs, but how big is that? 

Application-Aware Firewalls

Traditional L4-type firewalls are an anachronism.  They provide no security.  But people keep buying them and using them because their neighbors do.  Security is more about CYA than technical reality, so that provides a huge amount of inertia for anything that gets viewed as “best practice” at any given point in time.

In reality, firewalls need to get a lot better at classification in order to provide any traffic management benefit.  P2P traffic can hide under the cover of Port 80 or it can dynamically handshake for different TCP port numbers per connection, using port numbers also used by legitimate applications.  VoIP and RTP exhibit similar characteristics, where to know a packet is associated with VoIP requires looking at handshaking at the time of connection set-up.  Those issues are still at the networking layers.  In addition, there are increasing needs to makes security or traffic management decisions based on how users (or fake users) specifically interact with applications, or based on payloads or content, or based on behavior over time.

Therefore security devices need to both get a lot smarter AND faster, as they are needed more and more with-in LANs and data centers, not just at the (traditionally)  lower speed LAN/WAN boundary. 

Probably the best NBT opportunity of all I am analyzing here.  New technology advancement is needed to do more complex things faster.  Many companies (like P-Cube, Sanctum, Datapower, Packeteer, to name a very few) are in play today going after various niches, but I think there are still opportunities. 

Route Traffic Management

Due to the efforts of a couple of well-known industry analysts, this became a perceived NBT a couple of years ago.  Give me a break.  The concept was not new.  The key needs had been perceived and fulfilled by Radware several years ago.  And for them this was more of a feature than a product.  Certainly not a company.  I never thought anybody would want the level of router control (basically a way to tune the logic of dynamic routing protocols) proposed by these vendors and I was right.  This was a classic example of “it can be done therefore I shall do it” combined with “hey, someone is doing this we better fund someone else to do it just in case”.  Something like $100M of VC burnt money later; it’s easy to see that something like $20M value was created. 

Not even worth summarizing.

Wireless Switches

Creating an 802.11 wireless network used to be about installing an 802.11 access point, some client cards, and that was it.  That still works fine for SOHO and Small-to-Medium business environments, where the vast majority of the Access Point market (in terms of units) is.  But in a larger enterprise, where you need lot’s of Access Points for coverage, getting them to work together and managing the can be problematical. So the concept of wireless switch solutions came into play, where a wired device interdicts traffic coming from a de-featured Access Point and between the two (and possibly a supplementary management application), a large scale enterprise solution results.

Silicon Valley VCs are enamored with large enterprise right now. So a rash of so-called wireless switch start-ups have been launched.  The problem is – the market isn’t that huge, Cisco will take 60-80% of what there is, and the barrier to entry to create yet another competitor is not that high.  In the meantime, Accton and Taiwanese manufacturers are dominating the bigger residential and SMB markets. 

WLAN is going to continue to come on strong. But systems prices and margins are already low and will get lower. For the first time in a long time, service provider’s business cases look better than equipment vendor’s.

This will be a game for Cisco and low cost manufacturers (and arms dealers that supply them).  The wireless switch vendors will get bought up soon, as the larger OEMs decide they need that capability, starting with the companies that begin with A and moving (perhaps) through to C, 

Home Networking

I don’t think home networking is about building tiny LANs and environments replicating what we have in our offices on a small scale.  It starts there obviously. But I think home networking is really about sharing and distributing the use of home entertainment and home service devices. Since mobility is important in the home, I expect wireless technology in one form or another to be heavily used.  

This is another market where the number of units will be huge but the ASP per unit tiny.  Also, the channel is a bigger challenge than the technology. Therefore this is really a game for the incumbent consumer electronics players and the technology arms dealers that supply them.  

Web Services

Beats the hell out of me.