Model Features

The server model suite implements the following features.

Table 200: Server Model Features

Feature Description
CPU Modeling Multitasking CPU modeling uses a prioritized preemptive round-robin scheduling algorithm for each CPU Partition
CPU Partitions (Processor Sets) CPUs can be grouped into CPU partitions and a CPU can be shared by several CPU partitions. Jobs that are bound to a CPU partition are executed only on CPUs in that partition.
Memory (RAM) The model tracks memory usage to help predict server memory requirements.
Page Faults Hard and soft page faults are modeled.
Storage Partitions Jobs do I/O operations on storage partitions. The following storage partition types are modeled:
- Direct Attached Storage
- Measured Access
- Remote Storage Access
Disk I/O Drives and drive interfaces are modeled.
Multiple RAID levels RAID levels 0,1, 0+1, and 5 are modeled in the direct attached storage partition.
Replication Each write to a storage partition of any type can be replicated to a remote storage partition. Three modes of replication are modeled:
- Write-Replicate
- Replicate-Write
- Simultaneous
In each mode, the replication can be synchronous or asynchronous.
Write Caching Write operations to a storage partition can be cached so that the job that submitted the write does not incur an additional delay. The write operation will occur in the background.
  • Server models are implemented based on information available from the following sources.

Table 201: Reference Documents

CPU2000 Standard benchmarks published by SPEC (www.spec.org)
Manufacturer Specifications Disk and interface performance are obtained from manufacturer data sheets.
  • Workloads are characterized by jobs. A job can be a single thread, a process, or a group of threads and processes that represent the target workload. Each job has its own resource requirements for CPU, memory, and I/O.
  • Jobs can be executed as applications entirely within the server or as networked applications that scale according to the traffic load
  • The model includes many predefined hardware configurations
  • The model can operate with existing simple CPU server models

Creating and Processing Jobs Using Transaction Whiteboard

If you have a license for Riverbed SteelCentral Transaction Analyzer, you can use the Transaction Whiteboard GUI or Python scripting API to create and process server jobs.