White Papers

39 BP1062 | Best Practices for Decision Support Systems with Microsoft SQL Server 2012 using Dell EqualLogic PS Series
Storage Arrays
storage processing controllers (including on-board cache), and the network interfaces. The increased
storage throughput that became available as the number of arrays increased allowed faster data fetching
by the database. For a constant user load, the queries were also completed faster resulting in decreased
completion times when the array was scaled. With two arrays, the completion time was 20% less than with
a single array.
Proportional scaling in the array throughput was not observed because of the memory limitation at the
server. Figure 18 shows the available memory (MBytes) collected using Perfmon during the test duration in
both single array and scaled arrays scenarios with constant load from a single server. The average available
MBytes reduced by 16% when the second array was added and this limited the linear throughput
scalability. Here adding more RAM would help to achieve linear scalability.
Available memory comparison Figure 18
Since DW user queries consumed substantial CPU processing, to check the CPU utilization at the hosts
during the array scaling tests, the average and maximum CPU utilization percentages were measured
using Perfmon. As shown Figure 19, the average and the maximum CPU utilizations observed in this test
were well below 80%. This was a result of the higher processor power of the Dell PowerEdge R820 servers
(four sockets with eight cores in each). An increase of 33% in the average CPU utilization was observed
when the array was scaled from one to two for a constant workload from a single host.
0
500
1000
1500
2000
2500
3000
3500
4000
Available Memory (MByte)
Test Duration
Available Memory
1xarray
2xarrays