Cisco Cisco UCS C200 M2 High-Density Rack Server White Paper

Página de 44
38   © 2016 Cisco and/or its affiliates. All rights reserved.
Performance Characterization
Table 7 lists the workloads suited to run on the Cisco UCS C220 M4 and StorMagic solution. Each virtual machine running on 
the system can run the I/O patterns listed in the table. The table does not list all the workloads that this solution can run, but 
the workload characterization provides a baseline for further estimation and sizing.
Table 7.  Application I/O Profile: Cisco UCS C220 M4 and StorMagic Solution
Application Profile
RAID Type
Access Mode
Read:WriteRatio
Block Size
Metric
Web front end
5
Random
100:0
8 KB
I/O operations per second 
(IOPS) and response time 
(milliseconds [ms])
Online transaction 
processing (OLTP)
5
Random
80:20
8 KB
IOPS and response time (ms)
Decision support system, 
business intelligence, and 
video on demand (VoD)
5
Sequential
100:0 and 0:100
256/512 KB
Transfer rate (MBps)
For the Cisco UCS C220 M4 and StorMagic solution, each server is configured with eight 1.2-TB 10,000-rpm disks. Each 
disk provides about 140 to 150 IOPS with a response time of less than 20 milliseconds. The large size of the disk results in 
greater variation in the response times, because of the mechanical head movement between the inner and outer cylinders. 
Performance varies based on the size of the volume that is in use. For the test validation purposes here, each virtual machine 
is allocated with 100 GB of disk space on the mirrored volume. 
Workload Scenarios
The following two scenarios were tested and validated on the Cisco UCS C220 M4 and StorMagic solution:
•  Scenario A, with one virtual machine per mirrored volume: This scenario sets the baseline for I/O performance because no 
resource contention exists at the computing and network levels. The workload generated is purely I/O intensive and helps 
identify the maximum achievable throughput and response time for the system.
•  Scenario B, with 10 virtual machines per mirrored volume: This scenario helps estimate the achievable throughput and 
response time under a loaded system. Ten virtual machines are running on two servers, with each host carrying five virtual 
machines. This scenario mimics a typical deployment use case for a Cisco UCS C220 M4 and StorMagic solution. 
In both scenarios, the I/O profiles shown in Table 7 were run, and the results are discussed in the following sections. Tests 
were run for 4-KB block size and the results do not differ much of 8-KB block size; hence, the graphs show only for 8-KB 
block size. Scenario A: One Virtual Machine per Mirrored Volume.
Figure 48 shows that the system is capable of delivering random read operations at up to about 3400 IOPS within a response 
time of 20 milliseconds, which aligns with the expected disk performance.