The enterprise distribution's second update of its sixth edition offers numerous optimisations for virtualisation, resource management and filesystems. New and improved drivers and a major X Server update enhance the distribution's hardware support.
Thirteen months after the introduction of Red Hat Enterprise Linux (RHEL) 6, Red Hat presented the second update to its Linux distribution for corporate customers on 6 December. As the RHEL 6 series is still in the first part of its life cycle, version 6.2 not only contains the fixes that have been approved since the introduction of RHEL 6.1, but also an extensive list of improvements.
Like the process scheduler of Linux 3.2, the RHEL kernel's scheduler now offers a "control group (cgroup) CPU ceiling enforcement" feature. This feature prevents applications within a cgroup from using more CPU resources than are allocated to this group, even if this means that available CPU resources remain unused. This is relevant, for example, in cloud environments with "pay-per-use" accessibility, because it prevents users from accessing more resources than they have purchased. Ceiling enforcement is also of interest for service level guarantees, to ensure that a user's purchased CPU resources are accessible in a guest system instantly and at any time.
Red Hat says that it has also improved the scalability of the cgroups code. According to the release notes, this will allow several hundred groups to be used without any performance implications. The Red Hat developers note that they have also improved the performance of the cgroup controllers for I/O tasks and for working memory.
The Linux 2.6.32-based RHEL kernel offers pstore, a feature that was integrated in Linux 2.6.39 and which allows crash analysis data to be stored in non-volatile memory in the event of a system crash; some recent systems include such "platform-persistent storage", which is a small memory area for storing this very information.
As usual, Red Hat has updated a whole range of drivers to support new hardware as well as hardware that is to be released in the coming months. A particularly large number of changes affect the storage and network hardware drivers; Red Hat highlights that they have improved the distribution's support for 10-Gigabit network hardware. The audio drivers, and those for USB 3.0 and PCI Express (PCIe), have also been updated.
Red Hat usually keeps the distribution's core components at the level of the version that is used when a RHEL family is first introduced. However, this time the Linux distributor has made an exception for the graphics stack, switching to version 7.11 of Mesa 3D and to version 1.10 of the X Server. The developers in that scope have updated the graphics drivers for AMD, Intel and NVIDIA hardware to improve the support for recent components. The release notes vaguely mention the provision of graphics support for Intel's next generation of chipsets, probably referring to the Ivy Bridge platform which is expected next year. An updated Synaptics touchpad driver adds multi-touch support.
Red Hat has raised the perf performance monitoring tool and its kernel-side components to the level of Linux 3.1, which, among other improvements, adds cgroup support.
Supported in the scope of the Scalable File System add-on of RHEL 6, the XFS filesystem has been improved to better handle metadata-intensive workloads – for instance when accessing a directory with a large number of small files, which, the release notes say, could previously cause degraded performance.
A new addition is the Parallel NFS (pNFS) architecture, which has been included as an unsupported Technology Preview. Specified as part of NFS 4.1, this technology is designed to significantly increase data throughput by processing metadata and data in parallel.
Ext4 filesystems can now be created much more quickly because the kernel uses "Lazy Inode Table Initialisation" and will only create some of the filesystem structures during formatting, while the remaining structures are built later on. Asynchronous data writes allow write operations via the CIFS filesystem to be executed up to twice as fast as before; the kernel uses CIFS to mount Samba and Windows shares. This improvement has recently also been integrated into the Linux main development branch and will be included in kernel version 3.2.
Previously rated as a Technology Preview, the operation of Clustered Samba (CTDB) together with the GFS2 cluster filesystem is now fully supported. The High Availability (HA) add-on enables RHEL, from version 6.2, to support the HA configuration of applications that run under a VMware virtualisation solution on RHEL 6.2 guests.
The newly added Stream Control Transmission Protocol (SCTP) multihomingsupport allows nodes to be reached under several IP addresses. The RHEL 6 kernel now offers sendmmsg, a syscall that was added in Linux 3.0. This syscall allows whole message batches to be submitted with one function call, which reduces syscall overhead and can improve network throughput. New to RHEL, but supported in the official kernel since version 2.6.38, the Transmit Packet Steering (XPS) feature is designed to increase sending throughput.
A whole range of advancements affect the virtualisation area; this is bound to benefit Red Hat Enterprise Virtualization (RHEV) 3.0, which is currently still in development and whose kernel is related to that of RHEL 6.2.
From version 6.2, RHEL will offer "virtual CPU timeslice sharing" for KVM guest systems that have access to multiple processors. This technology allows guest systems to use their allocated resources more efficiently and hence work faster; it is thought to be advantageous particularly for large systems with numerous processor cores, where performance previously degraded because of pre-emption issues related to the locks that ensure exclusive access to data structures. The required processor-side support is called "Pause Filter" by AMD, and "Pause Loop Exiting" (PLE) by Intel.
The Red Hat developers say that they have improved the processing speed of UDP data that is handled via the virtio-net driver; RHEL 6.2 is supposed to be more efficient when processing network packets that are smaller than 4 Kbytes. Network communication I/O path optimisations are designed to enhance performance and will, according to the release notes, allow the operation of guest systems that are faster than the host. Red Hat also notes that it has increased the speed of the macvtap and vhost network technologies; however, these technologies continue to be Technology Previews and are, therefore, excluded from the support contract. The same applies to the distribution's "live snapshot" support that allows a guest system's filesystem status to be saved during operation – for example to create a backup.
With NUMA systems, the libvirt can now not only determine which processors, but also which memory area, will be used by the guest; ideally this will result in using the working memory that is most accessible to the applicable NUMA node. Qemu, which is used for KVM, now also offers a USB 2.0 emulation. When operated as a paravirtualised Xen guest, RHEL 6.2 now supports memory ballooning, a feature that allows the memory to be adjusted at run-time.
Also newly supported is the operation of applications in a Linux Container (LXC) in order to separate applications from the rest of the system and to manage resource usage via control groups (cgroups) and namespaces. This function can be used via the libvirt API and the virt-manager GUI; however, it has also been given Technology Preview status for the time being.
- As in the latest kernel versions, the Logical Volume Manager (LVM) can now use the MD subsystem's software RAID functions, which are usually accessed via the mdadm tool. For example, a RAID 5 can be created using the
lvcreate --type raid5 -i 3 -L 1G -n my_lv my_vginstruction. However, this function is also rated as an unsupported Technology Preview. Red Hat says that LVM devices can now be activated and deactivated quicker than before, and that the distribution now supports the ability to notify the underlying storage media of newly deallocated storage areas ("discard") for more device mapper targets.
- New versions of the mdadm and mdmon tools, and the appropriate kernel support, allow RHEL to offer Array Auto-Rebuild, RAID level migrations and SAS-SATA "drive roaming".
- The iSER (iSCSI extension for RDMA) initiator and target functions are now fully supported, allowing RHEL to be used as an iSCSI initiator and storage server. Also available as a Technology Preview is the new target mode for Fibre Channel over Ethernet (FCoE).
- Storage media that is connected via Fibre Channel or Serial Attach SCSI (SAS) can now be addressed via its World Wide Name (WWN) or its World Wide Identifier (WWID) during installation.
- Introduced in RHEL 6.1, the distribution's FreeIPA-based Identity Management (IPA/Identity, Policy and Audit) service is no longer a Technology Preview and is now fully supported by Red Hat. It interoperates with Microsoft Active Directory and is designed to reduce the Linux infrastructure's administrative overhead.
- As of RHEL 6.2 Beta, the Linux distribution is under evaluation forCommon Criteria for Evaluation Assurance Level (EAL) 4+. Certification according to FIPS-140 has been applied for and is currently being evaluated.
- RHEL is now capable of Trusted Boot (tboot). Together with Intel's Trusted Execution Technology (TXT), this technology, which is not to be confused with Secure Boot, can ensure that no malicious code – such as a kernel that includes a backdoor – is loaded while the system is being booted.
- The Network Manager now supports wireless roaming and can scan wireless networks in the background.
Details of these and other new RHEL 6.2 features can be found in the information documents provided by Red Hat. A quick overview is available in "Red Hat Enterprise Linux 6.2 – What's New". More comprehensive information from a technological perspective can be found in the release notes; those who want to dig even deeper can consult Red Hat's technical notes.