| Multicore Architecture | | | Hyper-Treading | | | Power & cooling advantages | | | Multicore Issue |
Multi core processing systems will bring new, challenging issues for the IT community and the infrastructure that it runs on. IDC has identified many key areas that will be impacted by the rollout and adoption of this technology. Although some observers feel that multi core processing spells an end to Moore's law, whereby processor performance is based on improved circuit density and speed, IDC believes that both initiatives provide a potentially potent one-two punch to commodity-based systems over established proprietary systems often found in the midrange and high-end systems markets.
The market dynamics affected by multi core processing include a broad array of areas, as follows:
The value proposition that ISVs bring to software pricing is already under stress as companies are challenged to reconsider pricing models to accommodate utility-based architectures and modular (or service orientated) frameworks. Now, the hardware industry has raised that ante by allowing the consumer much more compute power within a comparable platform. Consequently, given this analysis, multi core processing systems could be disruptive to the software industry by allowing previously less sophisticated platforms to compete for established markets at a low point on a revenue/price curve.
One of the current challenges for blade server deployments has been the power and cooling requirements of individual blades. Consequently, systems builders have to walk a fine line to identify the break even point for the chassis to accommodate a cost-effective blade configuration. Given that early results for multi core processors have shown them to be good at reducing power and cooling, IDC believes that they will help to accelerate the deployment of a fully loaded blade server chassis and therefore strengthen their total cost of ownership (TCO) case. Conversely, IDC also believes that multi core processing systems will extend the life cycle of large-scale systems for the very same reason (by lowering the overall power and cooling needs as more dense systems are built).
Multi core systems offer up the opportunity to process workloads with specific performance characteristics (e.g., particular transaction applications that benefit from parallel processing). However, one should be cautious that the multi core systems do not dominate a workload environment without being able to keep overall systems performances in balance. Here, computer, I/O, and memory needs have to be regulated to allow the overall system experience to be positive and not introduce systems latency. Therefore, systems architects building IT infrastructures should be on the lookout for a wide range of benchmarking results before deploying widespread multi core systems.
Applications, operating systems, and databases are developing their ability to take advantage of multithreading and parallel processing systems, and IDC believes that this is a key area for which disruptive and sustaining views will be expressed. IDC believes that multi core processing systems will be sustaining to existing processor architectures such as x86, POWER, and SPARC. By giving these platforms additional thread and parallel processing capabilities, their place in the infrastructure environment is sustained. However, platforms being built by companies such as Azul Systems and Clearspeed show that they could be disruptive to specific areas such as the Java Virtual Machine and computational processing, respectively.
The storage industry may provide a good reference point for multi core processing systems by comparing the storage capabilities of the disk subsystem. Disk capacity expanded rapidly and allowed new industries to be created (storage area networks and network attached storage). IDC believes that the server and systems market could face a similar expansion by proper use of the added compute cycles that multi core processing systems bring. Equally important is today's development of server capabilities being inserted into storage subsystems (e.g., IBM's POWER chipsets) to raise their intelligence and functions. IDC believes that, in the future, if multi core processing systems are placed into storage subsystems, new capabilities (e.g., security and encryption) or functions could be more easily performed by placing huge computing power very close to huge I/O subsystems.
It is clear that if multi core processing systems are accepted by the IT industry at large, there will be a shift in workload processing. IDC believes that workload migrations happen slower than systems vendors hope for, but multi core processing systems could be a major catalyst for such shifts. Reasons include the ability of blades to do more work, industry standard server platforms to move further in the enterprise, the simplification of grid computing networks (once a resource is found to do repetitive processing, that resource can do a lot more than today's server configuration), shifts in the application server tier (expansion in capability, but contraction in complexity), and, finally, changes in the size and capability of clustering farms.
Multi core processing systems will drive the need for systems vendors to expand their systems virtualization capabilities. In doing so, virtualization at a hardware level (partitioning) will take on a new meaning, with the potential need to partition the processing cores too. This function could add value to the virtualization stack and therefore create revenue opportunities.
Multi core processing systems will force new discussions on the merits of a scale-up or scale-out infrastructure, given that both directions will have more compute capabilities. The position that supporters of both infrastructures will need to take is showing the value that these new systems can bring to the IT infrastructure. CIO and IT management will not be ready to accept these systems unless the vendor can build a strong value proposition for either build out direction
| Multicore Architecture | | | Hyper-Treading | | | Power & cooling advantages | | | Multicore Issue |
