“We’ve increased the number of CPUs that a virtual guest can support, from the previous limit of 64 up to a new limit of 160,” Tim Burke, Vice-President of Linux Engineering at Red Hat, told InternetNews. “Scalability is important because we’re seeing more high-end compute-intensive workloads that are being run on virtualized guests.”
On the topic of beefy virtual guests, Red Hat has also grown the amount of memory that can be allocated to virtual guests. In RHEL 6.3, memory per virtual guest has moved from 512 GB in previous releases up to 2 TB. Virtual disks have also been scaled up in the SCSI subsystem from 25 disks per guest up to thousands per guest.
From an availability perspective, Red Hat is introducing Dynamic CPU Hotplug, technology that allows a virtual guest to add more virtual CPUs while it is still running.
“Previously, you had to momentarily shut down the virtual guest in order to gor it, so this is really a run-time enhancement,” Burke said. “Similarly, at run time you can now dynamically scale the size of the disk volumes for a virtual guest.”
Some of the scaling improvements that Red Hat users may see come from new Intel Xeon E5 CPUs, although Burke stressed that most of the scaling improvement comes from Red Hat coding changes.
“It’s fundamentally the KVM architecture. KVM is a thin layer on top of the bare metal hardware, so we don’t have to do everything twice,” Burke said.