There’s a lot of debate these days over scale-out versus scale-up IT architectures that stems from a debate that began with symmetrical multiprocessing systems back in the 1980s.
As it was then, the debate itself is being pushed by vendors that have platforms optimized for one computing scenario versus another. In reality, IT organizations have needs for platforms running application workloads with different characteristics. Most of those workloads are best served by a scale-out architecture. But there are application workloads such as databases or a large number of virtual machines that may lend themselves better to scale-up architectures. The challenge has been getting systems designed for applications that need to scale up to fit inside data centers dominated by 2u racks.
To address those types of workloads, Hewlett-Packard this week announced new four-socket members of the HP Proliant Gen8 family of servers that fit in a 2u form factor. In addition, John Gromala, director of product marketing for HP industry standard servers and software, notes that HP Proliant Gen8 servers make use of a memory design that has five times more density than rival offerings to provide up to 1.5TB of memory using 48 DIMM slots. In addition, Gromala says the new HP systems are also the first servers to support the PCIe 3.0 standard, which further enhances I/O performance.
The critical attribute of the HP ProLiant BL660c Gen8 server, says Gromala, is that it offers four-socket density in half the size of previous generations, while reducing the total cost of ownership by up to 30 percent. Priced starting at $5,912, Gromala says the new 4-socket servers fill out the middle of the new HP server lineup that is not only faster than rival platforms, but also includes functionality that automates most of the routine tasks associated with managing a server.
While server vendors are facing more than their fair share of challenges in terms of number of units they need to sell, the economy itself is only one factor at play. The number of application workloads that can be run per server is steadily increasing, which means each server is, in effect, taking on workloads that previously would have required multiple servers. That’s good news for IT organizations struggling with space, power consumption and application performance issues related to virtual machine density.
Whether IT organizations ultimately decide to pursue a scale-up versus scale-out architecture depends on the types of workloads they need to manage. What is apparent, however, is that most existing servers can no longer effectively meet those challenges.