Server Statistics
From apnsGuide
Contents |
Uses
- Provides various server metrics to browse over.
OS Information
- Canonical Hostname
- hostname of the server, typically of the form server name.apisnetworks.com
- Primary IP Address
- IP address of the server
- Kernel
- Linux kernel version; we run the 2.6 branch at this time with the minimum version being 2.6.19
- Distro
- informal name for "distribution". Although it may be reported as "Red Hat Enterprise", we run the free variant CentOS on all of the servers. Currently we are using CentOS 4 with a few features pulled from RHEL5, most notably glibc and coreutils
- Uptime
- an indication of how long the server has been running without a reboot. Note well that this is solely an indication of how long the kernel has been running and not an indication of the Web, FTP, or e-mail service availability. For Web server availability see the public uptime reports.
- Load average
- load averages are used as a rough approximation for server utilization. Depending upon the number of processors seen by the server, a value of 1, 2, 4, or even 8 may represent 100% "utilization". The term utilization is used loosely as many factors can attribute to artificially high load averages such as high I/O wait times or NFS backlogs (which cause high I/O wait times incidentally). For our servers a load average of 4.00 represents 100% utilization, but we prefer to keep the servers under 1.00 (25% utilization) to ensure peak performance with minimal congestion. Performance penalties are appreciable once a server ventures above 100% utilization.
Hardware Information
- # Processors
- number of physical + logical processors. Physical includes processors and cores whereas logical counts processors split through hyperthreading. The value in this column dictates what 100% utilization is in the load average field. A value of 2 for example dictates that 2.00 is 100% utilization according to the load average value. All of our servers are dual core, dual processor platforms.
- Model Name
- processor model name
- CPU Speed
- processor speed of each CPU core
- Cache Size
- L2 cache size embedded on each physical processor
- Bogomips
- an arbitrary measurement of the "the number of million times per second a processor can do absolutely nothing." The figure provided in the column should not be used for an absolute comparison among processors.
- PCI Devices
- all of the PCI devices inside the server.
Network Information
- Device
- at least two interfaces will appear, eth0 and lo. eth0 refers to the primary network card on the server.
- Received (packets)
- size, in gigabytes, of inbound traffic to the server
- Sent (packets)
- similar to received column, but instead reflects traffic originating from the server
- Err/Drop
- number of errors/dropped packets on a particular interface. This number should be 0.
Memory
- Type
- Either physical or swap memory. Physical memory refers to the physical RAM sticks in the server while swap refers to the slower virtual memory on the hard drives. Systems will exhaust the physical memory before moving to swap space for obvious performance reasons (a few nanoseconds versus a few milliseconds to retrieve data).
- Physical memory usage is partitioned into three categories: kernel + applications, buffers, and cache. The first is kernel + applications, which refers to the memory used by programs running on the server and the kernel itself.
- The next partition, buffers, are used as temporary holding queues for process output.
- cache on the other hand is frequently requested disk I/O. An example of how the server uses memory for caching would be a directory of symbolic links. On the first directory listing the request may take several seconds to complete; however, on subsequent iterations each request would complete in a fraction of a second. This is because the server is pulling directory information from its cache. It is normal for a server to utilize all of its physical RAM by allocating excess memory not used by applications and buffers to the filesystem cache.
- Disk swap utilization occurs once the server exhausts its physical RAM. It shuffles memory to the drive through paging which is significantly slower than physical RAM by the magnitude of 700,000 times for example (4.7 ms versus 7 ns; these are inexact values in favor of the hard drive).
Filesystems
- Mount
- device name
- Type
- filesystem type
- Percent Capacity
- amount of capacity
- Free/Used/Size
- numeric capacity values
