The IT atmosphere is clouding up and, as it does, administrators are going to have to address certain issues, among them, which is the most suitable processor for a particular use of cloud computing. Mr. Benchmark provides a few suggestions in this direction in his blog "blog What processor will fuel your first private cloud: Intel Nehalem or AMD Istanbul?"

As he suggests in his conclusion, this is a complex question, one that comes with a complex answer. He reports that benchmarking results show the AMD Istanbul better suited for calculation-intensive workloads but also show better memory performance of the Intel Nehalem. As a result, he writes, different layers within one's private cloud will need to be profiled in order to determine what is the best choice. The best operating system to use in this determination, he notes, is Solaris or OpenSolaris, which comes equipped with the right set of tools (i.e. Dynamic Tracing) to make the determination.

The blog reports that Sun Solution Centers and Sun Professional Services are starting now to build the first private cloud architectures based on Sun Open Source products, the most common building block being the versatile Sun Blade 6000 because of the capacity of this chassis to host many different type of CPU's (x86 & SPARC) and operating systems (Windows, Linux, OpenSolaris, Solaris or even Vmware vSphere).

Simultaneously, Intel and AMD have released two exceptional chips: the Intel Xeon 5500 (code name Nehalem) and the six-core AMD Opteron (code name Istanbul).

Mr. Benchmark writes that computing capabilities (in integer and floating point) and memory performance are the two key dimensions to explore in determining which is the optimal processor choice for cloud computing applications. He notes that improved caching mechanism (memcached for example) and the commoditization of solid state disks (SSDs) are moving database performance profiles toward memory- or cpu-intensive workloads. Additionally, the exceptional power of 10-Gbit based Hybrid storage appliances, like the Sun Storage 7410 Unified Storage System, allows administrators to be less concerned with I/O and network-bound situations.

The blog finds it useful that both AMD and Intel have produced chips that can use the same sockets as earlier chips. One need only download the latest platform BIOS and check on the OS level to ensure the presence of the latest OS releases, which include upgraded libraries and drivers, these being critical for utmost performance.

For his benchmarking exercise, Mr. Benchmark used iGenCPU, a calculation benchmark written in Java that calculates Benoit Mandelbrot's fractals using a custom Imaginary Numbers library. The main benefit of this workload, he points out, is that it naturally creates a 50% floating point and 50% integer calculation, similar to the floating operations produced by commercial software, which increase every year. This type of performance profile is getting closer and closer to what modern web servers (like Apache) and application servers (like GlassFish) will produce, he observes.

For this test, findings showed very similar peak throughput (984 fractals/s on Intel, 1008 fractals/s on AMD), while the AMD chip produced superior throughput at any level of concurrency. At 8 threads, which is a very common scalability limit for commercial virtualization products, it produces 28% more throughput than Nehalem because of the superiority of the Opteron calculation co-processors. He asserts that calculations require larger rather than faster L1/L2 cache, and the Opteron micro-architecture is naturally a better fit for this workload.

In assessing memory performance, the blogger used the iGenRAM benchmark, which is highly memory intensive using 1Gigabyte of memory per thread. Memory allocation time as well as memory search performance produced a combined throughput number showing that the faster DDR3 memory and higher frequency of the Intel chip make it a better fit for memory-intensive workloads. In peak, the Nehalem based system produce 23% more throughput than its competitor, he writes. For a small number of threads (1 to 4), however, both systems produce very similar numbers.

A second level predictor on this repetitive workload would most likely help the Nehalem-based system improve its scalability curve tangent past four threads, Mr. Benchmark speculates. He adds that his tests used DDR3 1066Mhz. The now available DDR3 1333Mhz would likely increase the Intel chip advantage on this workload, he suggests.

More Information

Cloud Computing Infrastructure and Architecture

Sun Cloud Strategic Planning/Consulting Services

OpenSolaris

Intel Xeon 5500

Sun CMT Servers Offer 1.6 GHz UltraSPARC T2 and T2 Plus Processors [...read more...]

		fullsource
Print Email Current Profile Archive Blog Generate newsletter