A recent edition of the UK The Economist had an interesting article called “The Server Market; Shifting Sands”. What the article highlighted was the fragile and fluid nature of the global server market. What stands out in the article is the decline in server revenue for the two leading server manufacturers HP and IBM when comparing Q1 2013 to Q1 2012 ($3.0 B down from $3.5 B). What is surprising is the declines are at a time when the demand for data centre based content is growing. While not directly analogous to the server market growth, content is processed and accessed by the server, so the demand for processing resources and storage go hand in hand. Understanding the growth in content helps frame the challenge the data centre engineers face as they plan ahead. At the end of 2012, an article in Computer Weekly referenced a study by EMC estimating 2.8 ZB of data was created in 2012 (2.8 trillion gigabytes) and will double every year until 2020. The Economist article continued to comment on the trend of the mega data centre builders, such as Facebook and Google. These companies are building mega data centres to host a large percentage of this ever expanding content. What The Economist highlights is that some organizations are avoiding the mainstream manufacturers and designing their own servers and buying directly from contract manufacturers in Taiwan to reduce cost. The reported revenue decline of the top two server manufactures was approximately 15%, and out of the next three largest server manufacturers (based on revenue), only Dell managed to increase their revenue. The underlying question is whether this is the start of a shift in server buying patterns or whether it remains limited to a small subset of the large data centre builders.
One potential factor which could be impacting server revenue is virtualization. Software virtualization vendors continue to claim increasing server performance and optimization benefits. What’s not clear is the real impact on server sales. The server manufacturers are most certainly studying the trends and they will be more than aware of the impact of market changes in the adjacent PC manufacturing industry. Dell and HP worked hard to maintain market leadership in the Netbook and Desktop PC markets, only to see an overnight shift from PC to tablets. The top four manufacturers of tablets are Apple, Samsung, Amazon and ASUS. HP and Dell are nowhere to be seen. While a similar disruptive technology such as tablets is not on the horizon for the server market, the trend towards low cost custom servers must be giving cause for concern. This may explain the recent publicized (but failed) discussion between IBM and Lenovo over selling IBM’s server business.
Given the fluid nature of the server market, the increasing impact of virtualization and changing trends in storage technologies, the network infrastructure needs to be a scalable and flexible adaption layer, which can be uncoupled from the physical server and storage technologies. This includes the ability to adapt to changes in physical server and storage technologies, but more importantly adapt and integrate with emerging trends in virtualization technologies. Network virtualization has traditionally been loosely coupled to server virtualization layer. This is now changing and software will play a critical role in the next generation network infrastructure. The Juniper Networks JunosV contrail is a standards based SDN controller solutions, which can abstract application service requirements and automate end to end orchestration of the network and service-chaining of L4-L7 services. This is the industry’s first truly open standards-based IP solution. The potential for this type of controller is wide and we can expect an evolution of standards and protocols to enhance new service functionality. The controller is a software based platform, which will be able to adapt quickly to new standards, but it is also important that the network layer can adapt as well. This is where Juniper Networks believes the choice of network infrastructure is going to be important. Switching technology which is programmable at the hardware and software level is going to be required. This is where we believe Juniper has further leadership. A good example is the new EX9200, which was recently launched as the industries first programmable SDN switch; this allows the hardware and software to be adapted to support new functionality, avoiding costly hardware refresh cycles. In a fluid market it is always good to be prepared for change. For more information on Juniper Networks’ SDN strategy read Bob Muglia’s blog post Decoding SDN.