Red Hat Enterprise Linux provides many useful options for performance tuning. Large systems, scaling to hundreds of processors, can be tuned to deliver superb performance. But tuning these systems requires considerable expertise and a well-defined workload. When large systems were expensive and few in number, it was acceptable to give them special treatment. Now that these systems are mainstream, more effective tools are needed.
To further complicate things, more powerful systems are being used now for service consolidation. Workloads that may have been running on four to eight older servers are now placed into a single server. And as discussed earlier in Section 18.104.22.168, “Parallel Computing”
, many mid-range systems nowadays contain more cores than yesterday's high-performance machines.
Many modern applications are designed for parallel processing, using multiple threads or processes to improve performance. However, few applications can make effective use of more than eight threads. Thus, multiple applications typically need to be installed on a 32-CPU system to maximize capacity.
Consider the situation: small, inexpensive mainstream systems are now at parity with the performance of yesterday's expensive, high-performance machines. Cheaper high-performance machines gave system architects the ability to consolidate more services to fewer machines.
However, some resources (such as I/O and network communications) are shared, and do not grow as fast as CPU count. As such, a system housing multiple applications can experience degraded overall performance when one application hogs too much of a single resource.
To address this, Red Hat Enterprise Linux 6 now supports control groups (cgroups). Cgroups allow administrators to allocate resources to specific tasks as needed. This means, for example, being able to allocate 80% of four CPUs, 60GB of memory, and 40% of disk I/O to a database application. A web application running on the same system could be given two CPUs, 2GB of memory, and 50% of available network bandwidth.
As a result, both database and web applications deliver good performance, as the system prevents both from excessively consuming system resources. In addition, many aspects of cgroups are self-tuning, allowing the system to respond accordingly to changes in workload.
A cgroup has two major components:
Tasks assigned to the cgroup run within the cgroup. Any child tasks they spawn also run within the cgroup. This allows an administrator to manage an entire application as a single unit. An administrator can also configure allocations for the following resources:
Within CPUsets, cgroups allow administrators to configure the number of CPUs, affinity for specific CPUs or nodes , and the amount of CPU time used by a set of tasks. Using cgroups to configure CPUsets is vital for ensuring good overall performance, preventing an application from consuming excessive resources at the cost of other tasks while simultaneously ensuring that the application is not starved for CPU time.
I/O bandwidth and network bandwidth are managed by other resource controllers. Again, the resource controllers allow you to determine how much bandwidth the tasks in a cgroup can consume, and ensure that the tasks in a cgroup neither consume excessive resources nor are starved of resources.
Cgroups allow the administrator to define and allocate, at a high level, the system resources that various applications need (and will) consume. The system then automatically manages and balances the various applications, delivering good predictable performance and optimizing the performance of the overall system.