For IT departments, 2016 looks to be a trying year: Most are planning hybrid cloud deployments and starting the...
scale-out process to absorb much of the data center, all while the need to respond to growing workloads creates pressure for additional virtualized server clusters. As if that were not enough, this must all be accomplished as server, storage and network technologies make significant shifts in direction and performance.
The challenge in handling a transition year like this is to balance shorter and longer term needs so that equipment doesn't have too short a life and doesn't add additional cost for features needed in future years.
The key is to buy only what is needed short term. Traditional buying strategies factor expected growth, but as organizations shift more workloads to the cloud, this approach can result in too many drives and servers for the job on hand. This is a bad policy in a world of rapidly declining prices, but a worse policy still when technology looks to drive much greater performance in next generation products.
The rationale for buying ahead is homogeneity, but virtualization and the cloud has abstracted the working platform from the hardware. Commercial off-the-shelf (COTS) gear is at the point that servers from most major vendors behave the same, operationally and under management, through the baseboard management controller. There are typically some fine details in management that differ, but these are minor.
This means that tomorrow's COTS gear will deliver VMs just like those of today. There may be more VMs, or they may run faster, but questions of needing homogeneity are no longer an issue.
This brings up a whole new set of questions; the most important is, "can I use white box servers for virtualization?" These are servers from the likes of Quanta, Lenovo or SuperMicro. They are typically much cheaper than servers from traditional vendors and have no vendor lock-ins on key components, such as drives. You can use any solid-state drive, for instance, avoiding the large price uplifts of the past.
Due to the huge volume sold to cloud-scale organizations, white box servers are top-quality products. They are as compatible as anything from Dell or HPE. Provided you buy with care -- ideally from a well-known distributor -- the drives will be much cheaper than typical enterprise storage, with bulk Serial Advanced Technology Attachment drives around $30 per terabyte.
Generally, support from these new vendors is a bit weaker than from traditional vendors. This would matter more if the architecture was proprietary. However, self-discovery built into today's plug-and-play elements don't include complicated setup processes.
These white box servers are often sophisticated. For instance, we're talking about designs tuned for multiple graphics processors with specialized airflow. These are every bit as competitive as units from HPE, Dell or EMC. In fact, a customer base that includes mega cloud service providers acts as a technology forcing function; often the new white boxes are slightly ahead of the traditional competition in delivering features and performance.
The much lower acquisition costs will free up funds for other improvements, such as in networking or storage.
We aren't talking pennies here. IDC noted that white box vendors shipped 7% of server units but only accounted for 2% of revenue in Europe, the Middle East and Africa in Q3 2015, which is another way of saying that those white box servers cost only one-third the cost of the traditional vendors' units.
The same process of "white boxing" can be expected for storage in 2016 and 2017 while networks undergo a migration to software-defined networks that utilizes merchant silicon bare-metal switches. All of these changes should achieve major cost reductions for the end-user data center. In 2016, we will witness a year of changing vendor loyalties as businesses try out white box servers and, eventually, begin to embrace them.