OSL goes Linux - RSIO and Linux - OSL Gesellschaft für offene Systemlösungen mbH

OSL goes Linux: Presenting RSIO on LinuxTag 2012

May 25 2012 - Berlin - Berlin Fair

Under the headline "Linux Kernel" OSL presented the RSIO protocol (High Speed Data Center Block IO over Ethernet).
After presenting the first implementation of RSIO on Solaris 2010 in Frankfurt (SNW Europe), a lot has happened in the meantime. Besides moving to Linux, RSIO has proven its outstanding characteristics in combination with storage virtualisation of OSL Storage Cluster in Real-Life SAP Applications. Alongside SAP KVM-Cluster are on top of our list: Performance, clarity, flexibility on basis of a single common Network-Infrastructure.

OSL congratulates the Host LinuxTag e.V. on a successful event. From our point of view there was a substantial interest in topics like virtual machines (KVM), clustering and storage technologies. For example the very well attended lecture of Thomas Gelf (Netways GmbH) on "DIY cloud out of the Linux Toolbox (german)"
Pleasing for us has been the recognition of our with partners and customers cleverly devised thoughts: We continue on our track and leverage our know-how when it comes to clustering and IO virtualisation, in order to come up with a customized and comprehensive package. Bert Miemietz tried to put our approach in perspective: Where and how, can Block IO and Clustering be combined in a truly simple Ethernet based infrastructure?

Download slides RSIO - Block IO (german)

A selection of Questions:

Why a new protocol? There are comparables - are they no good?

Honestly: We came to the conclusion, a new protocol fullfilling our high expectations is inevitable. It should be data centre ready, flexible and capable. For us that means to come up with a design supporting arbitrary socket types (currently TCP/IP supported). We have seen performance comparable to ATA over Ethernet (or higher), plus we gain routing and are highly scalable. Administration is incredibly easy. RSIO is able to utilise several channels for resilience and throughput (at the same time). RSIO offers a global namespace and clustering for storage-server and storage-client. RSIO may also submit Storage Cluster commands, enabling Clients to administer Storage Ressources as well as supports us to design increasingly powerful High-Availability Cluster (KVM-Cluster).
Where can I get more information? How do I get the software?

For information about RSIO look here: RSIO Technology.
We are thrilled about any user trying to take our RSIO software for a test drive. Please note that the Storage server side requires a development version of OSL Storage Cluster 4.0. Please contact us, to support your efforts to set up a successful test environment.
Are there any published details regarding the Protocol? Why didn't OSL publish its ideas and approaches as soon as possible (for a wider review by the linux community)?

RSIO was started as a commercial Software by OSL focusing on portability. Its first implementation was on Solaris, later on it was ported to Linux. Details regarding the protocol will be published soon, to ease porting to even more Operationg Systems + and/or to gain ideas for more efficient approaches. These obligations would, in the moment, be an obstacle to our challenging development work. Independent of all that, RSIO client will be given away under a proprietary, but no charge License. The Linux kernel module will be available under GPLv2.
What happens if the userspace cores?

It must simply be restarted, IO will not be suspended. Daemons in userspace may also be started in twin mode, observing each other and in cases of failure will restart the missing part. This is also possible on Solaris. Moreover restarts via SMF (Solaris) or inittab are supported.
And what is so special about your KVM-Cluster?

RSIO doesn't only provide Block-IO over Ethernet, it also comes (combined with OSL Storage Cluster) with moore goodies desired/required in clustered environments, like Global Namespace, Access Management etc.. Also to mention, the highly productive, cluster-capable Storage Virtualisation with lots of real-life solutions and an simple, logical and advanced design. The Protocol is also designed to transport more than read/write IOs. On one link (practically two for resilience) all may be mangaged: Block IO, Management, Live-Migration, High-Availability-Clustering. Together with OSL Storage Cluster it will also be possible to manage a whole farm of highly available KVMs from one central point. Moreover even more complex tasks like data-mirroring, backup and disaster recovery may all be managed from this central point, having everything under administrative control.