In the age of commodity servers, announcements of new devices hitting the channel have become somewhat tedious. New, more powerful chips are developed, servers gain improved processing and more memory, but the real game-changing features and functionality usually come from the software side.
Hardware is expected to power up and push bits around, drawing attention only when something goes wrong.
But IBM may have changed that pattern with the new eX5 line. In an age where enterprises are looking for the greatest flexibility as they pursue ever-changing virtual and cloud strategies, Big Blue has hit upon the idea of a server with interchangeable components. That is, processors, memory and storage can be swapped out of individual machines, providing an upgrade path that does not require wholesale replacement of entire servers each time you need a boost.
The company calls this "fabric-based computing," with the idea that servers need no longer follow the PC model in which various components are fixed in hardware. So memory, for example, can range up to 24 additional DIMMs per blade unit or 32 for racks, using a peripheral device called the Max5. That should greatly improve consolidation ratios for organizations that are hitting the wall not because they've run out of processing capacity, but memory.
The eX5 also contains an I/O innovation from the System P line called FlexNode. It's essentially an I/O partitioning system that allows you to pool multiple servers to help better navigate the growing confusion surrounding software licensing requirements in virtual environments. As Forrester analyst James Staten puts it:
" Why should you have to pay for a two-socket license for an application that can't even exercise more than two cores on a single socket Xeon 5500 server?"
Ironically, the Ex5 is in fact the result of a new processor design. The new device offers a 600 percent increase in memory over the existing Ex4 line and nearly doubles the performance per watt. So even as new processing or memory capabilities are added to the server itself, it will be easier to stay within operating cost envelopes.
In the computer industry, nothing remains in a vacuum for very long, however. So it's probably a safe bet that HP, Oracle and the rest of the crowd are working on similar designs.
That's good news, though, because for far too long, the server market has been on autopilot -- the answer to every challenge being: "Just add more servers."
Now that there's some flexibility on the hardware level again, we'll see new life in the server farm again.