Red Hat Advances Server Resource Control with RHEL 6.2
Red Hat (NYSE:RHT) is out with the second major update to its flagship Red Hat Enterprise Linux (RHEL) distribution release. The RHEL 6.2 release provides improved resource management, virtualization and storage capabilities.
RHEL 6 first debuted in November 2010 and was updated in May 2011 with the RHEL 6.1 release. One of the major changes in the RHEL 6.x series over the previous RHEL 5.x branch is support for cgroups (control groups) to control system resources.
"Cgroups debuted in RHEL 6.0, so the concept isn't new but what is new in RHEL 6.2 is the ability to set capacity limits," Tim Burke, vice president of Linux Engineering at Red Hat, told InternetNews.com.
Burke explained that previously the way cgroups worked was with ratios for capacity, for example setting a 40 percent of CPU policy. He noted that the trouble with this approach is that if, for example, there were three virtual guests running on a system, each controlled with a cgroup setting for one-third capacity, if two systems were idle, the third could consume all of the systems resources. With the new ability to specify capacities for I/O rates, CPU cycles and the like, there is now a bounded ceiling.
Going a step further, the system is also integrated with the libvirt virtualization APIs, which enable tools like Red Hat Enterprise Virtualization (RHEV) to control and specify system resources. Overall, cgroups are important for administrators to help ensure system integrity and availability. One specific use case sited by Burke is for backup applications that can be very I/O intensive.
"By using cgroups, you can set I/O policy caps and constraints on bare metal just as easily as you can for virtual machines," Burke said.
The new cgroups enhancements are all part of the upstream Linux kernel, from which other vendors will also benefit. Burke stressed that Red Hat has always had an upstream-first policy with code that is then backported to Red Hat's solutions.
"None of the stuff in RHEL 6.2 is Red Hat secret sauce goodies that we're sitting on," Burke said. "We really do hold to the upstream first mantra."
Another area of improvement in RHEL 6.2 is support for the iSCSI extension for RDMA. RDMA (Remote Direct Memory Access) is a specification that helps deliver low-latency characteristics to Ethernet. Among the contributors to the original Linux RDMA effort is networking vendor Cisco.
Burke noted that with Linux RDMA many of the efforts were driven by the Open Fabrics Initiative. In his view, that effort did not have an upstream first approach, and as such it was divergent with the mainline upstream Linux kernel development.
"Red Hat has led of the lot of the work, and we have the main maintainer for the upstream RDMA components," Burke said. "We no longer rebase to Open Fabrics, instead we're working upstream as part of the community to get RDMA enablers in place."
One area that isn't being addressed on the storage side of RHEL 6.2 is new, specific support for the Gluster filesystem. Red Hat acquired Gluster for $136 million in October. Burke noted that there is nothing Gluster-specific in RHEL 6.2 since the acquisition was too recent. That said, Red Hat's community Linux distribution Fedora has been working on improving Gluster. The recent Fedora 16 release included the HekaFS filesystem, which is an enhanced version of Gluster.
"Our intentions of course are to productize Gluster, not just at the filesystem level but also as an integrated stack," Burke said. "We'll partner with hardware vendors for more appliance type reference stacks that we'll be coming out with and so you'll see that coming in our roadmap, but it's not quite in place yet."
Read more on "Server OS Spotlight" »