Exiting the Lost Decade and Trying to Frame Some Random Thoughts
I have spent the last two weeks talking to a lot of people in the tech business in 15 companies across most of the networking, service provider, optical, CDN, storage and compute segments of the market. Most are former co-workers or industry colleagues. It is an interesting process as it had enabled me to look at the technology landscape or cloud world as everyone seems to want to call it from multiple viewpoints. I also spent time talking to a number of VCs about their investing funnel and these conversations have been the opposite in terms of timeframe compared to how I spent the last three years. During the last three years I was focused on the near term, typically 3-6 months. My first observation from the last few weeks is it feels like 2000-2001 all over again.
I think we could go back and look at PPT presentations from a number of failed startups from that era, add ten years and declare the business plan and market development to be spot on. I am calling the first ten years of 2000s the lost decade for startups. Yes, I know there were successes but the point I am making is the network evolution, internet, mobile devices, media did evolve as many had thought, it just too a little longer for a number of reasons. I find this interesting because it implies the supposition was correct, but the adoption rate was miscalculated.
I spent part of Friday morning reading data center fabric white papers from a variety of vendors. It took awhile because I feel asleep a few times :-). I did this because I have been developing some strong opinions on how the network, the data center, service providers of all types and the users will interact in the future. With that said here are some thoughts:
I have written before that I believe storage, compute and bandwidth prices are on a buy more for less curves. The easiest way to illustrate the curve is from Matthew Komorowki’s site (see the chart at the bottom). The affect those curves will have is interesting and a big driver now that we have exited the lost decade.
– The compute point is becoming increasing important as to where it exists in the network. Persistent network connections like Gmail, Facebook, applications, non-streaming content, games, etc are changing the rules as to how content is stored and handled. Where and how the security and policy authentication occurs is affecting content distribution.
– The future is more bandwidth with shorter distances between the compute elements. Despite all the papers on data center fabrics I read, the winners will be the companies that deliver constant improvements in bandwidth and enable a closing of the distance gap between compute elements.
– Data centers go all photonic. In the future, something like 85-95% of all compute and storage will be in the DC. Data sizes continue to increase. Someone is going to have to build PB level switches. The only way to reduce the power and cooling requirements is all photonic.
– Service provider optical upgrades roll on with OTN playing a bigger role. Getting the electronics out of the network will start in the data center and work into the service provider network. Aggregate long haul data rates, in the C band reach 25Tbit/s per fiber, which drive a new set of DC switching requirements, hence the all photonic DC.
– Goodbye STBs…Cisco already seeing this move. Hello femtocells or some sort of hybrid home network interface device with storage and security capable of fiber, copper, HFC and wireless access? Maybe it uses open source with a specific vendor frontend as the user interface. Kind of like Android or Windows Mobile on your smartphone and each device supplier builds their own interface. Just a thought.
Cloud, cloud, cloud…I am falling asleep with all the cloud talk…
Every few years the tech industry manufactures a new buzzword and then works itself up into lather. If it sounds really big, then Wall Street notices and a bunch of people create baskets of themed stocks. Cloud is the new winner. We have cloud stocks, cloud baskets, cheap ways to play the cloud, blah, blah, blah.
All this cloud talk is really about two elements. Content distribution and the compute point. I am using content as a catch all term to include digital media and apps. Where is the content stored and where is the compute point in the network. That is what we are really talking about when we are talking cloud and when I was looking at network designs in late 1980s and early 1990s we were doing the same so I must been doing cloud networking 20+ years ago. We had software as a service in the 1980s it was called SNA. I competed with service providers who were outsourcing networks in the 1990s so that must have been infrastructure as a service. All we are really doing is putting content and compute back in the data center after spending 25 years distributing it. I am just wondering if we distribute it again, but in the former of small DCs to augment the massive warehouse DCs. That is the thesis I am thinking about at this point.
** It is all about the network stupid, because it is all about compute. **