Distributed Virtual Disk

Distributed Virtual Disk

Introduction:

The ideal storage system is globally accessible, always available, provides unlimited performance and capacity for a large number of clients, and requires no management. It is conventionally believed that nothing is ideal in the universe. But still, if one tries to incorporate majority of these features in a single system, then a cost-effective scalable system becomes available. As mentioned earlier the ideal storage system is always available and is incrementally expandable. Existing storage systems are far from this ideal. The recent introduction of low-cost, scalable, high-performance networks allows us to re-examine the way we build storage systems and to investigate storage architectures that bring us closer to the ideal storage system. Petal, attempts to approximate this ideal in practice through a novel combination of features.

Currently, managing large storage systems is an expensive and complicated process. Various estimates suggest that for $1 of storage $5-$10 is spent to manage it. Adding a new storage device frequently requires a dozen or more distinct steps, many of which require reasoning about the system as a whole and are, therefore, difficult to automate. Often a single component failure can halt the entire system, and requires considerable time and effort to resume operation. For instance, in a network using a ring topology or a star topology, the hub acts like a server through which all the transactions occur. Thus, hub is a single element whose failure may result in a system halt. Moreover, the capacity and performance of individual components in the system must be periodically monitored and equipoised to reduce fragmentation and elimination of hot spots. This usually requires a lot of manual work in moving, partitioning, or replicating files and directories. Replicating of data blatantly increases the space complexity of the system. The ultimate solution to overcome all these problems is, Petal, a cost-effective scalable network that is an easy-to-manage distributed storage system.

Petal consists of a collection of network connected servers that cooperatively manage a pool of physical disks. To a Petal client, this collection appears as a highly available block-level storage system that provides large abstract containers called virtual disks. A virtual disk is globally accessible to all Petal clients on the network. A client can create a virtual disk on demand to tap the entire capacity and performance of the underlying physical resources. Furthermore, additional resources, such as servers and disks, can be automatically incorporated into Petal.

Petal, additionally exhibits the following novel combination of characteristics, which we believe will succor in reducing the complexity of managing large storage systems:

1. It can tolerate and recover from any single component failure such as disk, server, or network.
2. It can be geographically distributed to tolerate site failures such as power outages and natural disasters.
3. It transparently reconfigures to expand in performance and capacity as new servers and disks to be added.
4. It uniformly balances load and capacity throughout the servers in the system.
5. It provides fast, efficient support for backup and recovery in environments with multiple types of clients, such as file servers and databases.

Petal’s virtual disks allow us to cleanly separate a client’s view of storage from the physical resources that are used to implement it. This allows us to share the physical resources more flexibly among many clients, and to offer important services such as “snapshots” and incremental expandability in an efficient manner. The disk-like interface offered by Petal provides a lower level service than a distributed file system; however, we believe that a distributed file system can be efficiently implemented on top of Petal, and that the resulting system as a whole will be as cost-effective as a comparable distributed file system implementation.


for more info visit.
http://www.enjineer.com