In association with heise online

Substructure

RHEL 6 contains drivers for accelerating its operation as a guest system under KVM, VMware and Xen. The RHEL 6 kernel is based on version 2.6.32 of Linux; as usual, Red Hat has incorporated drivers and features that became available in later kernel versions or have never been part of the official kernel. Therefore, Red Hat sometimes refers to its kernel as a hybrid of Linux 2.6.32 and later versions.

Heaps of new features become apparent when comparing the RHEL 6 kernel with the version 2.6.18 kernel of RHEL 5, although more than a few of them are already old hat in many other distributions. For instance, the Completely Fair Scheduler (CFS) highlighted by Red Hat has been part of the Linux kernel since version 2.6.23. The "tickless" kernel, which stops the timer interrupt from going off a hundred or a thousand times per second when a system is idle, is already well-tested. This trick reduces both the power consumption and the basic load of RHEL 6 systems that operate as virtualised guests, which frees up the host CPU for productive tasks.

Overall, the updated kernel promises improved scalability. This means that RHEL 6 performs better on today's ubiquitous systems with one or more multi-core CPUs, and that RHEL is now also suitable for larger systems. In theory, RHEL 6 supports 64-bit x86 machines with up to 4,096 processor cores and 64 TBytes of memory, although only systems with 128 cores and 2 TBytes of memory have so far been tested. RHEL 5 was limited to 64 cores and 1 TByte of RAM. The KVM virtualisation limits have also been raised: guest systems can now be allocated up to 64 processor cores and up to 256 GByte of Memory per guest.

Premier league

Red Hat has highlighted functions, collectively known as "RAS" features, which are designed to improve a system's reliability, availability and serviceability. For instance, RHEL 6 allows processors and working memory to be added during operation on systems whose hardware supports such additions. Defective memory areas can be marked as flawed and decommissioned via Hwpoison. The new RHEL examines the CPU's Machine Check Exceptions (MCE) more closely and will, in certain cases, only terminate the affected processes instead of shutting down the whole system with a kernel panic. PCI Express Advanced Error Reporting (AER) and ACPI Platform Error Interface (APEI) are now also supported.


Zoom In terms of virtualisation, Red Hat has now completed the migration to KVM. Virt-Manager is the included configuration and management tool.
Ext4 has superseded its predecessor, Ext3, as the distribution's default file system; however, according to the Red Hat specifications, the maximum file system size of 16 TBytes, remains the same for Ext4 as it was for Ext3 in RHEL 5 (the RHEL specs also say Ext3 supported a maximum file size of 2 TBytes compared to 16 Tbytes for Ext4 in RHEL 6). For larger data storage systems with a capacity of up to 100 TBytes, Red Hat offers the "Scalable File System Add-On" via XFS at a charge. When setting up partitions and volumes, the installer makes sure that proper alignment is achieved to realise maximum performance on hard disks with 4-KByte sectors. The file system's discard feature informs storage devices of any storage areas that have been de-allocated through file editing or deleting – which not only improves the life span and performance of SSDs but is important for network storage systems which offer thin provisioning. The distribution's UEFI support, which is relevant for boot devices with capacities in excess of 2 TBytes is also new. However, RHEL 6 didn't boot via UEFI on a desktop PC whose Intel board is classified as an "UEFI Generation 2 Evaluation Platform" by uefi.org.

Control groups (Cgroups) allow administrators to put limits on the CPU time, memory requirements and network bandwidth of a process or process group in order to manage the sharing of resources. For example, an application or virtual machine can be allocated at least 90% of the network bandwidth to prevent a parallel test VM from disrupting production use.

Kernel technologies such as multi-queue networking, LRO (Large Receive Offload), and GRO (Generic Receive Offload) promise to improve the utilisation of resources in multi-processor systems and increase network throughput. New is support for FcoE (Fiber Channel over Ethernet) protocol in software and the option to implement iSCSI partitions as root or boot devices. These and other storage technologies can conveniently be configured in the graphical installation program, which is also accessible remotely via VNC. The text-based installer now only offers basic functionality.

The web interface of the Conga cluster management tool has been given a major overhaul and offers new functions; cluster communication is now controlled via the Corosync cluster engine. The distribution includes Samba version 3.5, which supports IPv6, works as a Windows 7 domain member and can assume a position of trust in interactions with Windows 2008 R2. Data exchanges with Samba clients can be encrypted, and improved scalability and reliability are available via Clustered Samba (CTDB). MySQL version 5.1.47, Postgres version 8.4.4 and Tomcat version 6.0.24 are also included.

Next: Desktop

Print Version | Permalink: http://h-online.com/-1135949
  • Twitter
  • Facebook
  • submit to slashdot
  • StumbleUpon
  • submit to reddit
 


  • July's Community Calendar





The H Open

The H Security

The H Developer

The H Internet Toolkit