Resource Points - CPU Time
CPU time is a measure of the amount of time a specific process spends using the CPUs on a server.
The datacentre group makes use of something called “process accounting” which is a feature of the Linux operating system. This logs each process run on the server and the amount of CPU time it consumed. We then parse and interpret these logs to ensure our servers are running optimally.
On occasion, we find users that are consuming too much CPU time in relation to other users on the server. This is an extremely important measurement in shared hosting because a user that is consuming a large amount of CPU time can negatively impact the hosting experience for both themselves and other users. We normally use a percentile ranking to find when users are consuming too much CPU time. We have developed the calculations for the ranking extensively in-house and it takes into account many variables when comparing and benchmarking users.
When we notify customers of a problem with their CPU time usage, we use a measurement called “Resource Points”. Essentially, this is the number of minutes that a user’s processes are actively utilizing the CPUs on a server. CPU minutes are real units, however, the amount of resource usage they represent can vary greatly depending on server configuration. Because the datacentre group uses standardized hardware and processors, these numbers are comparable across our server fleet. However, if you attempt to compare the number we give you with CPU minutes obtained on differing hardware configurations, such as those from other hosting companies, the number itself will not tell the whole story.
The reason for this is that different CPUs are capable of processing more or less information in a given amount of time. Additionally, the number of CPUs in a system, as well as whether or not the processes you are running can take advantage of multiple CPUs, can change exactly what the number means. For example, if you are on a server that has 8 cores of processing power, your CPU minute usage will differ on a server that has only 4 cores of processing power. As another example, consider two single CPU servers, but in one server, the CPU is from the late 1990s, and in the other server, the CPU is current. A user consuming 10 CPU minutes on the newer processor may easily use several times more than that on the older processor.
It is also possible to use different units of measurement to express CPU time. For example, it can be expressed as CPU seconds, or even as a unit of measurement defined in the Linux kernel as a “tick” or “jiffy”. A tick or jiffy cannot be converted directly to CPU seconds or minutes without specific information from the server they were generated on. Additionally, CPU seconds, ticks and jiffies are all smaller units of measurement compared to CPU minutes, so any measurements conveyed in those units will appear much larger than those conveyed in CPU minutes. All of these units suffer the same trouble when attempting to compare the numbers across different hardware configurations.
Please keep this information in mind when attempting to compare the number we give you to other numbers you may find in your research. Again, because the datacentre group uses standardized hardware and CPUs, we can safely compare the numbers across our fleet to reach a clear conclusion. Resource Points, along with the formulas and statistics that we’ve developed in-house, provide us with very clear benchmarks when it comes to assessing a user’s impact on our servers and whether or not they are fairly sharing the resources our servers have to offer.
The datacentre group makes use of something called “process accounting” which is a feature of the Linux operating system. This logs each process run on the server and the amount of CPU time it consumed. We then parse and interpret these logs to ensure our servers are running optimally.
On occasion, we find users that are consuming too much CPU time in relation to other users on the server. This is an extremely important measurement in shared hosting because a user that is consuming a large amount of CPU time can negatively impact the hosting experience for both themselves and other users. We normally use a percentile ranking to find when users are consuming too much CPU time. We have developed the calculations for the ranking extensively in-house and it takes into account many variables when comparing and benchmarking users.
When we notify customers of a problem with their CPU time usage, we use a measurement called “Resource Points”. Essentially, this is the number of minutes that a user’s processes are actively utilizing the CPUs on a server. CPU minutes are real units, however, the amount of resource usage they represent can vary greatly depending on server configuration. Because the datacentre group uses standardized hardware and processors, these numbers are comparable across our server fleet. However, if you attempt to compare the number we give you with CPU minutes obtained on differing hardware configurations, such as those from other hosting companies, the number itself will not tell the whole story.
The reason for this is that different CPUs are capable of processing more or less information in a given amount of time. Additionally, the number of CPUs in a system, as well as whether or not the processes you are running can take advantage of multiple CPUs, can change exactly what the number means. For example, if you are on a server that has 8 cores of processing power, your CPU minute usage will differ on a server that has only 4 cores of processing power. As another example, consider two single CPU servers, but in one server, the CPU is from the late 1990s, and in the other server, the CPU is current. A user consuming 10 CPU minutes on the newer processor may easily use several times more than that on the older processor.
It is also possible to use different units of measurement to express CPU time. For example, it can be expressed as CPU seconds, or even as a unit of measurement defined in the Linux kernel as a “tick” or “jiffy”. A tick or jiffy cannot be converted directly to CPU seconds or minutes without specific information from the server they were generated on. Additionally, CPU seconds, ticks and jiffies are all smaller units of measurement compared to CPU minutes, so any measurements conveyed in those units will appear much larger than those conveyed in CPU minutes. All of these units suffer the same trouble when attempting to compare the numbers across different hardware configurations.
Please keep this information in mind when attempting to compare the number we give you to other numbers you may find in your research. Again, because the datacentre group uses standardized hardware and CPUs, we can safely compare the numbers across our fleet to reach a clear conclusion. Resource Points, along with the formulas and statistics that we’ve developed in-house, provide us with very clear benchmarks when it comes to assessing a user’s impact on our servers and whether or not they are fairly sharing the resources our servers have to offer.