I finally got a chance to catch up on some trade rags and came across this interesting blurb from Network Magazine about XORP, the Linux of Routing. While an early project from UC Berkeley, I encourage you to take a look and keep it on your radar. This is yet another example of the potential commoditization of high-end products. Here is an excerpt:
“Since the routing code and the OS are free, the biggest expense will be the hardware. Commodity PCs make notoriously poor routing platforms, so they’ll need a sufficiently gast bus structure to boost their total processing and throughput. The recently standardized PCI-X 2.0 fits that billing, providing bus sppeds reaching 700,000 64-byte Ethernet packets per second. That’s good news not only on the performance front, but for the price tag as well. “A machine with 1GByte of RAM could easily be assembled today for less than $1,500,” says Orion Hodson, a XORP developer. By comparison, a Cisco 7304-Cisco Systems’ highest-grade enterprise router with software forwarding-runs $22,000.”
It is still early days for XORP and the platform still needs to address performance and security issues, but the point is that any software product with a large enough installed base can be vulnerable to open source competition.
Speaking of open source software, MySQL just announced a new version of its database which has built in load-balancing and automated failover so it can be deployed in large transactional environments. This is a big deal and grealy expands the market opportunity for MySQL and will better position it against Oracle and IBM. One other open opportunity for attack in the database market is the reporting and analytics end. One of my portfolio companies which I have written about before, Metapa, is leveraging open souce technology, mainly Linux and PostgreSQL, to deliver terabyte-scale data warehousing on a cluster of commodity hardware. The secret sauce is its proprietary Linux database clustering software which is “purpose-built” for Business Intelligence. In early benchmark tests, the product has shown up to 10-50x performance improvements over existing data warehouses run on traditional enterprise systems. So if I were an incumbent, I would be concerned about these developments.
The advent of open source networking devices doesn’t surprise me. If you ever cracked open networking gear that costed tens of thousands of dollars, you would see a lot of off-the-shelf circa-1999 parts you could pick up from Best Buy for a few hundred dollars – 5400RPM ATA100 harddrives, PCI add-on cards, CompactFlash cards, PC100 SDRAM, Pentium III chips, etc. Of course, you have to purchase “Cisco approved” or “Nortel approved” parts (which naturally cost 10-20 times more) or risk voiding the warranty.
The only unique parts are the firmware and some of the circuit blades.
It’s not strictly true that these are the same parts. You can build out the same board with higher quality transistors and other components that are electrically the same but have a much lower failure rate. This can double or triple the price compared to stuff that is made of components, each of which barely passes QC and collectively that translates into a higher failure rate and a shorter warranty period.
For some uses, the open source solution will end up killing off both Cisco and Nortel just as white box PCs collectively are larger than any individual brand name PC. But that’s not going to be the end of the brand names. There are plenty of situations where you can’t afford the down time so you have to buy high quality hardware. The price difference becomes much less important at that point.
Where XORP will succeed is in the dlink level category. A great open source product that can provide great performance. Perfect for the branch office, or the SOHO.
I can’t see open source routing ever replcing Cisco. For supporting business critical applications and servers I would only trust a proven componet such as Cisco. Performance and stability wise there will be no comparrison. Cisco’s uptime speaks for itself.
I agree about the Cisco comment for now. Thomas in an earlier comment says it best when he believes that XORP will best succeed in the SOHO category with DLink.