May 5, 2003

A small review - Networld+Interop 2003

This is a small review of the Networld+Interop show held in Las Vegas April 29 through May 1, 2003.

I've been attending Interop events ever since it was the TCP/IP Interoperability Conference in 1987.  I've helped design and construct most of the networks used in those shows over the years.  Over the last few years I've moved to the iLabs, where we continue the tradition of assembling and operating impossibly complicated multi-vendor networks using the latest (and sometimes yet unreleased) technologies.

Here's a photo of one of our many equipment racks in the iLabs.  If you look closely you'll see just about every vendor of heavy duty gear.  (Photo: 186k bytes)

And yes, those are OC192 and 10Gig interfaces.  In case you are wondering, the key locks are there to keep people from accidentally frying their eyeballs.  (Photo: 163k bytes)

interfaces.jpg (162926 bytes)

My main work at the show was part of the IP Storage Initiative.  Our nascent iSCSI FAQ may be found at  We assembled a fairly broad spectrum of iSCSI, fiberchannel, and fiberchannel-over-IP products and implementations (some of which are open source or copylefted) - nearly all of which interoperated.  (As usual some folks had to spend a couple of days working on their code on the show floor to accomplish this.)  In addition to our pile of heavy-duty gear we had the world's first USB RAID 5!  We exported three USB flash memory drives on a Linux box as a JBOD of iSCSI drives.  We then went over a 54megabit wireless link to a Windows 2K box that mapped these drives into a RAID 5.  And yes, it worked, even when I used my Maxwell product to introduce some fairly significant levels of traffic impairments.  (Photo: 30k bytes)

usb-raid.jpg (30155 bytes)

IP Multicast - This was the first show in a decade in which we did not have full IP multicast connectivity to the outside.  It's sad to see such a useful technology fall by the wayside.

IPv6 was also missing in action.  It may be dead.  The IETF has spent far too long arguing like Lilliputians over what seem to be trivial details.  It is unclear if there remains anything pushing us towards IPv6.  The once present need to reduce the pressure on the IP address space has been diminished through the use of NATs (Network Address Translators) and CIDR.  It is my prediction that the internet will devolve into a collection of independent national, academic, and corporate networks, each with its own IPv4 address space, and interconnected through a relatively few formal points of exchange that require address translation.  In addition, IPv6 brings no solution to the internet's real problem - routing.

Wireless was everywhere.  There were many wireless vendors on the floor.  And at the iLabs we had an 802.1x interoperability test.  We had a few dozen access points in a 2m by 4m area - probably cumulatively emitting enough radiation to bake a chicken.  The folks at AirMagnet found several hundred access points on the show floor.  Overall, my sense is that the wireless landscape - security, bandwidth, coverage, reliability, reparability - will all be rapidly improving over the next year or two.  Hopefully the regulatory bodies like the FCC and NTIA don't become impediments.

iSCSI - I was a skeptic when I first encountered iSCSI a couple of months ago.  However, I've used Maxwell to stress iSCSI and I've seen people using it over transoceanic links.  iSCSI is much more robust than I had originally anticipated.  However, it will stress networks and is vulnerable to predatory implementations that don't follow the rules (such as TCP congestion avoidence).  I'll be posting some materials on this and on how to tune IP networks to support iSCSI in the developing iSCSI FAQ.

Voice over IP (VOIP) - IP based telephony has become rather routine.  We used IP phones without really caring or noticing that they were IP based phones.  They mostly worked (except when I needed to generate touch-tone tones to trigger my answering machine.)  I still have not seen any products that incorporate ENUM in any meaningful way.

Security - There were lots of folks touting security this and security that.  The tide is moving in the direction of building walls rather than enabling interconnectivity.  As I mentioned in my previous blog entry, "The death of the "Global Uniform Name Space", it seems to me that people are willing to place their internet fiefdom behind strong walls, even to the degree of making it difficult or impossible for outsiders to utter even the names of objects and resources protected by those walls.

What wasn't at the show - Few of the amazing number of open source and copylefted tools were visible in the vendor booths - but they were omnipresent on most of the computers used to manage and operate the network and on the laptops of the techies.  The just released survey of network security tools shows the vibrancy of this sector of the internet.

Judging from the show attendance it's going to be a while before the networking industry recovers - potential customers were rare.  However, it is also pretty clear that innovation continues in certain limited areas such as wireless.  One wonders whether Networld+Interop will return next year.  I hope so, although I'd recommend that it remember its roots and return to San Jose, California.

Posted by karl at May 5, 2003 12:57 AM