Tuesday, March 24, 2009

Radia Perlman: She Radiates Intelligence and Humor

In honor of Ada Lovelace Day, my blog post today features Dr. Radia Perlamn. Dr. Perlman is a Sun Fellow at Sun Microsystems. She specializes in network and security protocols. In the past she worked at DEC and Novell and she is the author of two textbooks. She has an MS in math and a PhD in computer science from MIT.

Radia, which by the way is pronounced to sound like "radio" but with an "a" at the end, is brilliant, quirky, funny, and a terrific teacher. She radiates intelligence, humor, mother-earth friendliness, and strength. I first met her in 2002 when I took her USENIX course on network security protocols. I had admired her for many years before that, though. Like most networking people of a certain age, I cut my teeth on her terrific book, Interconnections: Bridges, Routers, Switches, and Internetworking Protocols.

Radia invented mechanisms that make modern link-state routing protocols efficient and robust. Her thesis on routing in the presence of malicious failures remains the most important work in routing security. She has also made contributions in such areas as strong password protocols, analysis and redesign of IPsec's IKE protocols, PKI models, efficient certificate revocation, and distributed authorization. She recently was granted a patent for her work on ephemeral decryption which can make files reliably disappear. (Oh, I wish I could use that to make some emails I've sent disappear! :-)

Radia is the inventor of the spanning tree algorithm used by network bridges (switches). The photo above shows her posing as a spanning tree. Many computer scientists have worked on spanning tree algorithms, but Radia's work is noted for having a long life outside the academic lab. Her spanning-tree algorithm is at the heart of most enterprise networks.

Bridges (and the spanning tree protocol) were invented to deal with endnodes that run applications directly above Ethernet, without the benefit of a "Layer 3" protocol (e.g., IP). Bridges were a method of moving Ethernet frames around without a Layer 3 header. According to a recent email discussion I had with Radia, she assumed once everyone implemented Layer 3, bridges would go away. This probably would have happened if the world had adopted ISO's Layer 3, the Connectionless Network Layer Protocol (CLNP), Radia said. But the world adopted IP.

CLNP supported a prefix that an entire corporate network shared. Nodes could move around within the corporate network and keep their Layer 3 address. In contrast, with IP, every link requires its own prefix, which means routers have to be configured, addresses have to change if a node moves, and addresses get wasted. Once the world adopted IP, network administrators tended to avoid the extra configuration required with IP by using bridges (switches). The spanning tree algorithm lets administrators create large Ethernet bridged networks with redundant links that take over upon failure of the links chosen to be part of the spanning tree.

The spanning-tree algorithm has many shortcomings. It doesn't use optimal paths, doesn't allow splitting network traffic across multiple paths, and is lacking many other features of Layer 3 routing protocols. It is also intrinsically fragile because the Ethernet header doesn't contain a Time to Live (TTL) or hop count field, so any loop that forms can be devastating. Frames loops forever and bridges flood broadcasts over and over again. Something as simple as missed spanning-tree protocol messages can result in these deadly loops.

Radia is currently working on a new type of device that has the ease of management of bridging, but the robustness and features of routers. This is being standardized in the Transparent Interconnection of Lots of Links (TRILL) working group of the IETF. The TRILL working group is designing a solution for shortest-path frame forwarding in multihop Ethernet networks with arbitrary topologies, using Radia's work on link-state routing protocols. Radia isn't afraid to say it's time to retire her famous spanning-tree protocol.

One of the reasons I admire Radia is because she has a knack for remaining friendly while speaking controversial or unwelcome truths. In 2005, the Anita Borg Institute for Women and Technology honored Radia with a Women of Vision award. I highly recommend listening to her keynote speech which is available here. She recently told me that some of the things she said in this speech make her cringe, but I think her thoughts on the topic of women in technology are awesome, just like I think she's awesome. Well, now I've probably made her cringe again. Oh well!


  1. By the way, you can view all the Ada Lovelace Day posts here. Some of the posts are fantastic and lead one deeper and deeper into unknown parts of the Web, where there's no telling what you might find. I highly recommend, for example, the Heidi, Geek Girl Detective web comic books I stumbled upon.

    The 1st chapter of Book 1 is "I am the Network Guy," something Heidi has to say to a consultant who wouldn't believe he was supposed to meet with her. I can relate to that. :-)

  2. Priscilla, I've been following your blog and enjoying it very much. It's important to know what women have accomplished in Computer Science Technology and related fields. I think that people are being inspired by your articles to enter the tech profession and follow their dreams of creating innovational 'who knows what' (it's a discovery process).

    I happen to be an older woman who in her 50's (after having been in the 3D gaming and multimedia industry for years) realized that I was good at analyzing problems and seeing how to implement technical solutions. I was writing psuedo code and giving it to programmers.

    So (despite my fears that I wouldn't be able to keep up with the young men who knew far more than me), I went back to school and got my degree in CS and am writing my own code. These young men were very helpful and supportive. I did very well - graduated Summa Cum Laude. But more importantly, I was enjoying learning and creating 'stuff', including being given the opportunity to do an internship for Harry and David Corporation (with Neil Hedgepeth, co-analyzed and implemented the Sarbane-Oxeley code tracking/audit reporting system for their E-Com dept).

    I had the pleasure of being one of Priscilla's students (3 classes - couldn't get enough of her great teaching) while at Southern Orregon University. I was greatly inspired by her extensive knowledge in Networks and other knowledge in CS (she's a sponge of information and 'how to'). She always told me I could 'do it' and to believe in myself (and filled me in on 'Imposter Complex' - which is a topic in itself).

    And here she is, still devoted to inspiring by making this great web site and blog. Did you know she also has her degree in English? So, anyone reading this blog of Priscilla's - both sides of your brain work just fine. Don't let anyone talk you into just using one side (don't let anyone tell you your gender or anything else keeps you from being good at something). I'm an artist and a programmer - go for it, 'cause you can do anything that you WANT to do and be an inspiration for those who are watching what YOU do. Thanks Priscilla for all you're doing.

  3. Thanks for your comment, Linda! Your story is inspirational. I worked at Harry and David also (for 10 weeks). They have a sharp, well-run IT department, with many strong, supportive women and men. Very cool that you got to work on Sarbanes Oxley. That's so important. Thanks for mentioning the impostor complex, sometimes called the impostor syndrome. It affects many women in tech. I may have to write about it! :-)

  4. Great article. Thanks! I'm in the network field. She's a giant in the field. Every network engineer learns about spanning-tree on Day 1. Her work on IS-IS/CLNP has helped service providers tremendously, as that is their protocol of choice. Her work on TRILL is revolutionizing data center networks.