The selected benchmarks are intended to treat (i) Memory IO; (ii) CPU; (iii) Disk IO; (iv) Application; (v) Network.
Several cloud workloads using Hadoop framework are IO-bound, which means that Disk IO performance becomes a bottleneck for achieving higher performance.
Each control group is associated with a limit on the amount of resources that can be consumed on that server, e.g., the percentage of cpu time, memory, disk IO, and network IO.
If a server becomes overloaded by any metric, e.g., cpu load, memory usage, disk IO or network IO, this can be remedied by migrating some of the load to other servers with more available capacity.
The website offers results for a set of regularly used benchmarks for users to explore, from CPU (single-core and multicores), Disk IO (e.g. bonnie++, IOzone) and Memory IO (e.g. STREAM), to Application (e.g. various zip compressions) as well as an Aggregate for each of these categories.
In Section 5, we present performance-per-watt characterization and analysis for disk IO.
The [Wang, Y.-L. Y.-L., Russell, Raeder, J. The Io mass loading disk: model C.T.ulations.
Memcached is memory-bound workload, so disk IO performance-per-watt characterization may not be applicable for Memcached.
The latency is related to the processor latency, Internet access latency, disk IO latency, and latency associated with moving data within a given cluster.
There are a number and variety of (micro-) benchmark results reported for Cloud CPU, Disk IO, Memory, network, and so on, usually involving Amazon Web Services (AWS) [7 13].
