CentricStor is a virtual tape library / archive system which is used in production system worldwide, because of its outstanding features and performance.

Archival of data is one of the core responsibilities in a time of daily growing amounts of data processing. The volume of data in mainframe or other server-farms has to be stored, retrieved and secured. Backups and redundant data storing are highly desired or even required today. Traditional long term or backup storage in data-processing-centres is achieved via automated storage on magnetic tapes in (automated) tape libraries. The downsides of these traditional systems are:

  • enormous cost per unit for an automated tape library
  • therefore: only few physical tape libraries per computing-centre
  • ergo: only sequential usable; bottlenecks
  • tape storage speed can not keep up with datawrite speed of processing-centres

Solution - CentricStor - a virtual archive system CentricStor starts where tape libraries stop:

CentricStor

  • uses the same interface as existing tape libraries
  • acts as go-between for hosts and physical tape libraries
  • disentangles data-processing and data-storage
  • manages the storage of data in a flexible, fast and scalable way, through its special design and architecture.
  • CentricStor presents the same interface to the using host systems as a traditional tape library would present. That means, there is an automated tape library control-interface and a set of drives that are usable by the applications. No changes are necessary in the applications of the hosts.
  • The drives that are presented to the applications for reading and writing of volumes are purely virtual drives and are managed by the CentricStor system. The control information which virtual tape is to be inserted in the logical drives and what volume the application wants to access is as in the traditional case exchanged via the control interface. The differences in the CentricStor approach is, that these control commands are interpreted by CentricStor and not by an automated tape library. This encompasses all actions that are of interest to the host side.

Internal, CentricStor uses a cache of volumes on raid systems which are connected in a fast fibre channel fabric. This cache of volumes can be managed independently by the CentricStor system. This means in essence, that CentricStor decides when to store volumes on real tape media in its backend tape library.

CentricStor ITXperts GmbH Munich

The advantages of this approach are quite clear: Hosts can write in parallel to different virtual drives without being constrained in simultaneous connections by the number of physical available or existing tape drives. Due to the nature of the raid drives, volumes are still as safely stored as on tape. Since raid drives allow higher transfer rates than tape drives and since multiple parallel connections are possible, the total data throughput is definitely higher in the CentricStor approach. Another aspect where CentricStor beats traditional automated tape libraries is in the necessary access times. If the accessed volume is still in the raid-cache drives or if a new volume is stored, there is no penalty for retrieving and accessing the right tapes. Thus reaction times are immensely fast.

High Availability is another aspect of CentricStor. Downtime of the backend tape drives or whole libraries does not register on the application level. Hardware faults can be hidden by the CentricStor level. Accesses are handled with the available volumes in the cache or can be fulfilled via remote accesses. The additional abstraction level provides this advantageous us decoupling for free.

Furthermore the modular architecture and the design of this additional level between data-providers and data-storage allows additional features, which are implemented and realized in the CentricStor system without the help of application modifications. That means features like redundant mirroring, dual-save or tape reorganisation happen in the background and totally invisible.

Design and implementation of CentricStor CentricStor manages to solve all these aspects with the services its architecture provides. Because of its distributed nature and inherent complexity the design of the architecture calls for clear, efficient and ingenious as well as anticipatory planing. This one aspect where our team at ITXperts excels in. As can be seen in the interactions between the diverse components of the CentricStor system, the functionality and the easy extensibility of the system. Equally to the planing stage, the implementation has to be done in the same painstaking way, while being open enough for extensions and while being efficient and fast. As the components developed by ITXperts show, this is another aspect with which we feel right at home.

Contributions of ITXperts to CentricStor

  • Design of the CentricStor architecture - modular, distributed, high performant
  • low-level component development - core driver, device drivers, kernel patches
  • distributed, high performance filesystem
  • middleware supporting higher level applications on low-level components
  • development of a high availability providing middleware
  • development of application-level components for CentricStor
  • continuing further development of CentricStor core components
  • ongoing support for problems