Solidna – A core strategic cloud-based storage product and service area for a major Infrastructure as a Service provider
Auf einen Blick
- Projektleiter/in : Prof. Dr. Thomas Michael Bohnert
- Co-Projektleiter/in : Andrew Edmonds, Christof Marti
- Projektstatus : abgeschlossen
- Drittmittelgeber : KTI (KTI-Projekt / Projekt Nr. 15157.2 PFES-ES)
- Projektpartner : CloudSigma AG
- Kontaktperson : Christof Marti
Beschreibung
Solidna developed a core strategic cloud-based storage product and service area for a major Infrastructure as a Service provider (CloudSigma). The three key innovations that will be developed in Solidna are:
1. Upgraded Compute Storage Performance: this will focus on stability and dependability, guaranteeing a minimum performance level for critical systems. Here we will focus on stability and dependability, guaranteeing a minimum performance level for critical systems. Customers directly affect each other’s performance on many IaaS platforms today. This is the most critical problem for a public cloud provider to solve. By having a virtual drive in a public cloud and storing that drive across hundreds of physical drives, that performance limitation can be reduced significantly.
Solidna will develop the means to deliver a cloud storage solution with a high level of stability and dependability and to guarantee a minimum performance level for critical systems. This innovation will be delivered through the following technical innovations of:
- Mechanisms to guarantee a minimum expected performance
- Reliable clients that ensure the data is read/written consistently
- Definition of specific performance critical system metrics and reporting of those metrics
- Optimisation of the system based on system metrics (e.g. variable block sizing based on data stored)
- Data segmentation optimisations including block-size optimisation and distributed striping
- On-demand performance guarantees can grow as requested by the user
2. Advanced Storage Management Functionality will be another focus in Solidna. This will focus on enabling a number of abilities including creation of live snapshots, the backup of virtual drives and to geo-replicate a drive to one or more additional locations. The same rich feature-set as a high end commercial SAN product will result from this project but using standard low-cost, commodity hardware and a new upgraded software storage system. These new features will form the basis of new revenue streams. Key features include:
* The ability to create live snapshots and backups of virtual drives. This ability allows to backup data from drives and keep these as separate copies and is important for the data resilience and security reasons, * The capability to geo-replicate a drive to one or more additional locations.
This innovation will be delivered through the follwing technical innovations:
* System agents to watch and discover failed or potential failing system nodes * Mechanisms and algorithms for deregistration, recreation and associated redistribution and rebalancing of the storage nodes * Active reliability automated testing of the cloud storage service * Logically centralised control centre for the entire system * Storage system with the ability to rebalance the storage nodes * Expansion the ICCLab framework to accommodate the DFS * Functionality of policy-defined geo-replication * Functionality of volume migration
3. Object-based Storage Environment: Massive capacity cloud storage and multi-modal API access to a reliable storage. The scalability of the storage offered to customers is limited to the maximum drive size of 2TB per drive. Although a server can mount multiple drives to form a larger storage volume, the practical maximum size per server could be estimated at around 20-30 TB. Even with multiple drives this becomes difficult to manage. As well as the usual API interface allowing access to virtual drives, the proposed work aims to expose directories of files in the object storage as network mount points to the compute cloud. In effect this gives customers two access points to their storage, based on usage needs, a network drive API and an object storage API interface. This innovation will be delivered through the following technical innovations:
Accessing stored data using POSIX and HTTP from within the VM with implementation of file system drivers and HTTP APIReview of existing storage APIs and recommendation