All posts by James Kuo

IOPS calculations

Read and Write IOPs

Total the read and write IOPs and calculate the number of recommended spindles. For example, 10K RPM spindles can be assumed to support 130 IOPs, and 15K spindles can be assumed to support 180 IOPs. The total Backend IOPs workload is a function of the front-end (host) workload and the RAID overhead using these formulas:
R5: Raid-Adjusted IOPs = (Disk IOPS * Read Ratio) + ((Disk IOPs * Write Ratio) * 4)
R10: Raid-Adjusted IOPs = (Disk IOPS * Read Ratio) + ((Disk IOPs * Write Ratio) * 2)
For example, an Exchange server with a peak of approximately 389 Read IOPs and 419 Write IOPs in an R10 configuration would need to support approximately 808 back-end disk IOPs, which translates into a minimum of eight 10K spindles or six 15K spindles (this calculation does not take into consideration storage requirements).
15k RPM Drive=180 IOPs
10k RPM Drive=120 IOPs
7200 = 50 IOPs
5400 = 40 IOPs

NTP tips

Checking your time against a reference NTP server

w32tm /stripchart / /samples:5 /dataonly

Checking time across multiple systems using ‘psexec’ (from

psexec \\computer1,computer2,computer2 w32tm /stripchart / /samples:5 /dataonly

Setting the time services to use NTP and server NTP requests (for all AD servers).

Step 1:

 W32tm /config /syncfromflags:manual / /reliable:yes /update
W32tm /resync

Step 2:


If a guest VM, it is best NOT to sync with the Hyper-V host server …

reg add HKLM\SYSTEM\CurrentControlSet\Services\W32Time\TimeProviders\VMICTimeProvider /v Enabled /t reg_dword /d 0

Step 3:

Net stop w32time
Net start w32time

Checking time accross all DCs:
w32tm /monitor

Checking time configuration on a DC:

w32tm /query /configuration