CN1602480A - Manage memory resources attached to data networks - Google Patents

Abstract

A computer network includes a plurality of memory nodes (103 a-106 a) each having physical memory resources (121 a). A system management server (150) on a network (100) identifies and collects physical storage (121a) on the network (100) into a virtual storage pool (160). When an application (121) running on a storage client accesses network storage, the system management server (150) allocates a segment of the virtual storage pool (160) to the application. Segments of the virtual memory pool (160) are stored on physical memory (121a) resources on the network (100). The system management server monitors the application's use of network memory and transparently and dynamically reallocates the virtual segments to optimal physical memory resources.

Description

Manage memory resources attached to data networks

Cross References to Related Applications

This application claims priority to Israeli Patent Application No. 147073, filed December 10, 2001, under 35 U.S.C. §119.

technical field

The present invention relates to the field of data networks. More specifically, the present invention relates to a method for dynamically managing memory resources attached to a data network and dynamically allocating them to a plurality of workstations also connected to said data network.

Background technique

In a typical network computing environment, the amount of memory available is measured in many terabytes, and the complexity of managing this memory at an enterprise level also complicates its efficient use. Across the enterprise, many different versions of the same computer files cluttered users' hard drives. Efforts to quickly verify memory usage face substantial implementation problems. Implementing general memory allocation policies and performing memory usage analysis from an enterprise perspective is also complex.

In recent years, enterprises have encountered the problem of being unable to effectively implement and manage a centralized storage strategy without centralizing all of their storage resources together. Furthermore, inconsistencies between different file versions arise and make efficient updates difficult.

In the prior art, a central dedicated file server was used as a repository for the network's computer memory. If the number of files is large, the file server can be distributed among several computer systems. However, as the capacity of computer memory increases, the use of dedicated file servers for storage displays a potential bottleneck. The data throughput required to send many files to and from a central dedicated file server is a major contributor to network congestion.

The cost of attaching computer storage to a dedicated file server and the complexity of managing that storage grows rapidly when the need exceeds a certain limit. The need to frequently back up the contents of this memory places a heavier load on the dedicated file server.

As the load on the file server increases, a large portion of its operating system needs to be dedicated to the internal management of the server itself. The complexity of managing file server storage increases as more hardware components are added to increase available storage.

Conventional memory facilities cannot efficiently allocate memory resources because they do not take into account the frequency with which particular data items are accessed. For example, in an email application, the Inbox folder is accessed much more frequently than the Deleted Items folder. Furthermore, in many cases, static allocation of memory resources to servers creates a situation where the available memory available to other servers is underutilized.

Another disadvantage of conventional memory allocation systems is low quality of service (QoS). This means that applications that require a lot of computer resources are starved of resources, while the required memory resources are allocated to applications that do not require them as strongly. Furthermore, ineffective memory management and allocation often leads to memory corruption, which also causes applications using the corrupted memory to also crash. This is also known as system downtime (time during which an application is unresponsive due to a failure). Another disadvantage of conventional memory management systems arises when memory resources should be maintained, upgraded, added or removed. In these situations, several applications (or even all applications) should be suspended, which leads to further extended system downtime.

Therefore, there is a need for a new method of efficiently managing memory resources and distributing files over a data network. In the current state of technology, for data exchange, efficiently distributing data across many disks is a better solution.

It is therefore an object of the present invention to provide a method for dynamically managing and allocating memory resources which overcomes the drawbacks of the prior art.

Another object of the present invention is to provide a method for dynamically managing and allocating memory resources which reduces the amount of unused memory resources.

Yet another object of the present invention is to provide a method for dynamically managing and allocating memory resources which improves the quality of service provided to applications using the memory resources.

Yet another object of the present invention is to provide a method for dynamically managing and allocating memory resources that improves the reliability of memory resources used by applications by reducing system downtime.

Yet another object of the present invention is to provide a method for dynamically managing and allocating memory resources that dynamically balances the load imposed by each application program among the memory resources.

Yet another object of the present invention is to provide a method for dynamically managing and allocating memory resources according to the actual memory needs of each application.

Contents of the invention

The invention relates to a method for dynamically managing memory resources attached to a data network and dynamically allocating the memory resources to user-executed applications connected to the data network through an access point. The physical memory resources allocated to each application and the performance of the physical memory resources are periodically monitored. One or more physical storage resources are represented by corresponding virtual storage spaces aggregated in a virtual storage repository. Periodically monitor the physical memory requirements of each application. Each physical memory resource is partitioned into a plurality of physical memory segments, and the performance attribute of each physical memory segment corresponds to the performance of its physical memory resource. The repository is partitioned into multiple virtual memory segments, and each physical memory segment is mapped to a corresponding virtual memory segment with the same performance attributes. For each application, a virtual memory resource is introduced that includes a combination of virtual memory segments optimized for that application according to the performance attributes and requirements of their corresponding physical memory segments. Physical storage space is reallocated for the application by reallocating each virtual memory segment in the combination to a corresponding physical memory segment.

Preferably, the parameters used to evaluate performance are the level of data/data files stored in the physical memory resource used by the application, the reliability of the physical memory resource, the available storage space on the physical memory resource, the The access time of the data and the delay of data exchange between the computer executing the application program and the access point of the physical memory resource. Repeatedly evaluate the performance of each physical memory resource and monitor each application's physical memory requirements. The reallocation of each virtual memory segment to another corresponding physical memory segment is dynamically changed according to changes in performance and/or requirements.

The evaluation is performed by defining a number of memory nodes, each representing an access point to which a physical memory resource is connected. One or more parameters associated with each memory node are monitored and a dynamic score is assigned to each memory node.

In one aspect, storage priorities are assigned to each memory node. Each segment of virtual memory associated with an application having execution priority is reallocated to a set of memory nodes having a higher storage priority value. Dynamically monitor the performance of each storage node, and change the priority of the storage node according to the monitoring results. The reallocation of each virtual memory segment is changed whenever required.

By storing copies of data files on several different storage nodes and allowing the application to access the copy stored in the storage node with the best performance, the application access time for the required data blocks can be shortened .

Reallocating at least one virtual memory segment to the added physical memory resource by updating the contents of the repository according to the addition/removal of physical memory resources, evaluating the performance of each added physical memory resource, and dynamically changing according to the performance. The physical memory segment obtained by the memory resource and/or another corresponding physical memory segment, adds/removes the physical memory resource to/from the data network in a manner transparent to the currently executing application.

A data read operation may be performed on a virtual memory resource by sending a request from an application program such that the request specifies the location of the requested data within the virtual memory resource. The location of the requested data within the virtual memory resource is mapped to a pool of at least one memory node containing at least a portion of the requested data. One or more memory nodes having the shortest response time to execute the request are selected from the pool. The request is sent to the selected memory node with the lowest data exchange load and the application is allowed to read the requested data from the selected memory node.

A data write operation may be performed on a virtual memory resource by sending a request from an application program such that the request determines the requested data to be written and where the data should be written on the virtual memory resource. Create a pool of possible storage nodes for storing data. At least one memory node is selected from the pool whose physical location on the data network has the shortest response time to execute the request. The request is sent to the selected memory node with the lowest data exchange load and the application is allowed to write data to the selected memory node.

Each application may access each memory node using a computer linked to at least one memory node and providing access to physical memory resources inaccessible to the application as an intermediary between the application and the inaccessible memory resource access.

Preferably, the data throughput performance of each intermediary is evaluated for each application, and based on the results of the evaluation, each application will be provided with a load required for accessibility to inaccessible memory resources within two or Dynamic distribution among more intermediaries.

Reallocate physical storage space for each application by reallocating virtual memory segments corresponding to applications to two or more memory nodes such that the two or more memory nodes are allocated according to their respective scores Dynamically distribute load to achieve load balancing between two or more storage nodes.

By continuously or periodically monitoring the requirement level of the actual physical storage space, allocating the actual physical storage space to the application program according to the requirement level during the time period when the application program actually needs the physical storage space, and dynamically changing the allocation level according to the change of the requirement level , physical memory resources can be reallocated per application.

The invention also relates to a system for dynamically managing memory resources attached to a data network and dynamically allocating the memory resources to applications executed by users connected to the data network through an access point and operating according to the method described above .

Description of drawings

These and other features and advantages of the present invention will be better understood from the following illustrative, non-limiting description of preferred embodiments of the invention with reference to the accompanying drawings, which include:

FIG. 1 schematically shows the architecture of a system for dynamically managing memory resources and dynamically allocating memory resources to application servers/workstations connected to a data network according to a preferred embodiment of the present invention;

Fig. 2 schematically shows the structure of physical storage resources and virtual storage resources and the mapping between them according to a preferred embodiment of the present invention; and

3A and 3B schematically illustrate read and write operations performed within a system for dynamically managing memory resources and dynamically allocating memory resources to application servers/workstations connected to a data network according to a preferred embodiment of the present invention .

Detailed ways

The present invention comprises following parts:

a memory domain supervisor, located on the system management server, for managing memory allocation policies and allocating memory to memory clients;

a storage node agent, located on each computer that has available storage space on its hard disk; and

A storage client, located on each computer that needs to use storage space.

The tasks of each of these components are described in more detail below.

Fig. 1 schematically shows the architecture of a system for dynamically managing memory resources and dynamically allocating memory resources to application servers/workstations connected to a data network according to a preferred embodiment of the present invention. The data network 100 includes a local area network (LAM) 101, and the local area network 101 includes: a network manager 102; a plurality of workstations 103 to 106, respectively having local storage 103a to 106a; and a plurality of network area storage (Network-Area-Storage (NAS)) Servers 110 and 111, each containing a large amount of storage space, are used for LAN purposes. NAS servers 110 and 111 are in continuous communication (via communication path 170 ) with application servers 121 to 123 connected to LAN 100 and running applications used by workstations 102 to 105 thereon. Communication path 170 is used to temporarily store data files required to run applications on workstations on LAN 101 . Application servers 121 to 123 may contain their own (local storage) hard disk 121a, or they may use storage services provided by an external Storage Area Network (Storage Area Network (SAN)) 140 by using their several storage disks 141 to 143 . Each access point of an independent memory resource (a physical memory component such as a hard disk) to the network is called a memory node.

According to the prior art, the application servers 121 to 123 respectively store the data of their application programs on their own corresponding hard disk 121 a (if large enough), or on corresponding disks 141 to 143 allocated by the SAN 140 . In order to overcome the disadvantages of unused storage space, system downtime and insufficient quality of service, a management server 150 is added to the network manager 101 . The management server 150 identifies all physical memory resources (i.e., all hard disks) connected to the network 100 and aggregates them into a virtual memory pool 160. The distribution of data is implemented in segments among physical memory resources, making this distribution transparent to each application. Furthermore, the management server 150 monitors (by running the storage domain supervisor installed therein) all of the various applications currently used by the workstations 103 to 106 on the network. Thus, server 150 can detect how much disk space each application implementation occupies on the application server running the application. Using this knowledge and criteria, server 150 reallocates virtual memory resources to each application based on each application's actual needs and usage levels. In order to generate dynamic instructions to the network manager 102, the server 150 processes the gathered knowledge to adjust and reallocate the available storage space among running applications, while providing each application with the virtual storage desired by that application to function properly amount of space. The server 150 is arranged so that it runs parallel to the network communication path 171 between the LAN 101 and the application servers 121 to 123 . This configuration ensures that server 150 is not a bottleneck for data flowing through communication path 171, thus eliminating data congestion.

The reallocation process is based on the fact that many applications, despite being disk intensive, actually only use some of the resources. The remaining resources not used by the application are only needed by the application when it is aware of its existence, not running on it. For example, an application can take up 15GB of memory, while only 10GB is actually used by the disk for installation and data files. To run correctly, the application requires the remaining 5GB to be available on its allocated disk, but they are almost never (or never) used. The reallocation process takes over such unused portions of disk resources and allocates them to the applications that actually need them to run. In this way, the size of the virtual storage volume of the network can be larger than the actual physical storage space. This increases the flexibility of the network to the limit of the formatting capabilities of its operating system's physical storage space. Each application is allocated actual physical storage space on demand (dynamically) and only for the period of time that the application actually needs. The demand level is monitored continuously or periodically, and if a decrease in the demand level is detected, the amount of allocated physical storage space is correspondingly reduced for the application and can be allocated to other applications currently increasing their demand level. The same can be done for allocating virtual memory resources to each application.

Another optional feature that the system can implement is its liquidity—it is an indication of how much additional memory resources the system should allocate for immediate use by applications. Fluidity provides better memory allocation performance and ensures that applications do not run out of memory resources due to unexpected increases in memory requirements. The storage volume usage indicator alerts the system manager before applications exhaust available memory resources.

Yet another optional feature of the system is its accessibility—even though some of these storage devices can only be accessed by a limited number of computers in the network, it still enables the application server to have access to all network storage devices ( storage nodes) for storage. This is accomplished by using computers that can access the inaccessible disks as intermediaries and giving them access to applications that request the inaccessible data. For each application, the data throughput performance of each intermediary (i.e., the amount of data successfully processed by the intermediary in a given period of time) is evaluated in detail, and based on the evaluation results, the performance of each intermediary is dynamically evaluated among different intermediaries. Distribute the load needed to achieve accessibility.

To ensure that applications whose resources are exempted can still run without failure, server 150 creates virtual storage volumes 161, 162 and 163 (in virtual storage pool 160) for application servers 121, 122 and 123, respectively. These virtual storage volumes are reflected as virtual disks 121b, 122b and 123b. This means that even though an application does not own all the physical disk resources it needs to run, it still receives an indication from the network manager 102 that it can use all of these resources, when in fact its unused resources are allocated to other applications . Therefore, the application server only knows the size of its virtual disk and not its physical disk. Because the resource requirements of each application change frequently, the size of the virtual disk as seen by the application server also changes. Each virtual storage volume is partitioned into predetermined storage segments ("chunks") that can be dynamically mapped back to physical memory resources (e.g., disks) by allocating the predetermined storage segments between corresponding physical memory resources. 121a, 141 to 143).

A storage node agent is provided for each storage node, and the storage node agent is a software component that performs reallocation of data exchange between allocated physical and virtual memory resources. According to a preferred embodiment of the present invention, the resources of each memory node linked to the end user's workstation are also added to the virtual memory pool 160 . Mapping is performed by defining a plurality of memory nodes 130a to 130i respectively connected to corresponding physical memory resources. Each memory node is evaluated with performance parameters obtained from predetermined criteria (e.g., physical memory available on the node, latency of data reaching the node over a data network, response to disks connected to the memory node) access time, etc.) as a feature.

To optimize the reallocation process, the server 150 dynamically evaluates each memory node, and in a manner transparent to the application, for each application, distributes (by allocating) among the memory nodes found to be most suitable for the application corresponding to The application's physical memory segment. Each request made by an application to access its data files is sent to the corresponding storage node that currently contains those data files. The evaluation process is repeated, and data files are moved from one node to another based on the evaluation results.

Management console 164 controls the operation of server 150 , and management console 150 communicates with server 150 via LAN/WAN 165 and provides dynamic instructions to network manager 102 .

Server 150 includes pointers to locations within virtual memory pool 160 corresponding to each file within the system, so applications making file requests need not know their actual physical location. Virtual storage pool 160 maintains a set of tables that map virtual storage space to a set of storage physical volumes located on different disks (storage nodes) throughout the network.

Using the virtual storage pool 160, any client application can access every file on every storage disk associated with the network. During forwarding of a data request, the client application identifies itself so it can extract the security level of its access from the correct table in the virtual memory pool 160 .

Fig. 2 schematically shows the structures of physical memory resources and virtual memory resources and the mapping between them according to a preferred embodiment of the present invention. Each virtual storage volume (e.g., 161) associated with each application is partitioned into equal storage "blocks," which in turn are subdivided into segments such that each segment is associated with the optimal storage node (via continuous evaluation). Each segment of a block is mapped by its corresponding optimal memory node to a "chunk" located on the corresponding portion of the disk associated with that node. It can be seen from this figure that each block can be mapped to multiple disks (distributed among multiple disks) each having different performance and located at different locations on the data network.

The layered architecture suggested by the present invention makes the storage network expandable while maintaining its performance substantially. The network is divided into areas (eg separate LANs) that are connected to each other. Selected computers within each predetermined region maintain their local routing tables that map virtual storage space to a set of physical memory resources located within that region. Whenever an unmapped storage volume needs to be accessed, the computer searches for the location of the requested storage volume within the virtual storage pool 160, and then accesses its data. The local routing table is updated each time the data in the storage area changes. Of all regions, only the virtual memory pool 160 maintains a complete picture of changes in metadata (ie, data related to the attributes, structure, and location of stored data files). In this way, especially for large storage networks, the number of times that virtual storage pool 160 should be accessed in order to access files on any storage node on the network can be minimized, and the number of times required to update the local routing table can be minimized. The amount of metadata traffic is reduced to a minimum.

Physical memory resources can be implemented using Redundant Array of Independent Disks (RAID—redundantly storing the same data on multiple hard disks (ie, different locations)). Saving multiple copies of files is a more cost-effective method because there is no operational delay in redistributing them, and backups of these files are available immediately.

3A and 3B schematically illustrate read and write operations performed within a system for dynamically managing memory resources and dynamically allocating memory resources to application servers/workstations connected to a data network according to a preferred embodiment of the present invention .

In a write operation, the user application (running on the storage client) issues a request to read some data, and attaches three parameters to the request—the virtual volume to be read, the offset of the requested data within the volume, and the data length. The request is forwarded through the file system and accessed to a low-level device component of the storage client, usually a disk. Then, the low-level device calls the block allocator. The block allocator uses the volume mapping table to translate the virtual location (the allocation virtual drive within the virtual memory pool 160) of the requested data (volume-specified and the requested offset parameters) into the physical location within the network where the data is actually stored (the storage node ).

Often, there are situations where request data is written to more than one location within the network. To decide which storage node is best to retrieve data from, the storage client periodically sends requests for the read file to each storage node on the network, and then measures the response time. Thereafter, it creates a table of the best memory node with the shortest read access time (highest priority) for the memory client's location. The load balancer uses this table to calculate the best storage node from which to retrieve the requested data. Data can be retrieved from the memory node with the highest priority. Alternatively, if the memory node with the highest priority is congested due to concurrent requests from other applications, data is retrieved from another memory node with similar or next-best priority. Since the performance of each memory node is evaluated continuously (or periodically) for each application, data retrieval can be dynamically distributed among all the different memory nodes that contain part of the data required by each application ( load balancing between storage nodes). For each application, the combination of memory nodes used for each read operation can be changed according to the change of the evaluation result.

After determining the retrieval location, the RAID controller in charge of I/O operation in the system sends requests through various network communication cards. It then accesses the correct memory node and then retrieves the requested data.

Using a similar approach, write operations are performed. The request to write data received from the user application also has 3 parameters, but this time, instead of the data length (which occurs in a read operation), it is the actual data to be written. The initial steps are the same until the block allocator extracts from the volume image table the precise location within which the data should be written. The block allocator then uses the node speed results and the usage information table to examine all available memory nodes across the network and form a pool of possible storage space for writing data. The block allocator allocates the memory required to create at least two copies of each requested data block for the user to create a new data file.

To select storage nodes from the pool, the load balancer evaluates each remote storage node in order to allocate storage in the most efficient manner, according to the priority determined by the following parameters:

Amount of storage remaining on the storage node.

Other requests sent by other applications to this storage node to access data.

Data congestion on the path to the node.

Data is written to the memory node with the highest priority, or by continuously (or periodically) evaluating the performance of each memory node for each application. For each application, data write operations are dynamically distributed among different (or even all) memory nodes according to the evaluation results (load balancing among memory nodes). For each application, the combination of memory nodes used for each write operation can be changed according to the change of the evaluation result.

After selecting the storage nodes to be used, the RAID controller sends write requests to the correct NAS and SAN devices and sends data to them through various network communication cards. The data is then received and stored on the correct storage nodes within the correct NAS and SAN devices.

Since its users' requests for data stored on the network are constantly changing, the memory distribution of this data is dynamically modified according to changing memory requests. Finally, the number of instances of this data is optimized and its physical location between different memory nodes on the network is changed according to the user's requirements for it. In this way, the system keeps adjusting itself until an optimal configuration is reached.

According to a preferred embodiment of the present invention, multiple copies of each file are stored on at least two different nodes on the network as a backup in case of system failure. For each requested file, the file usage pattern stored in the configuration table associated with that file is evaluated. By evaluating the replica of the file and storing the replica of the file on a separate storage node on the network, data throughput on the network is increased by eliminating access contention based on the evaluation results.

Instead of utilizing a central server, file distribution is performed by simultaneously generating multiple copies of the file on different nodes of the network. Thus, the distribution is decentralized and the bottleneck state is eliminated.

Perform image processing dynamically without interrupting applications. Thus, new storage disks can be added to the data network by registering them in the virtual storage pool.

The storage location of each replica of each file and updated metadata about each block (small size memory segment on the hard disk) of memory comprising these files is dynamically maintained in tables of the virtual memory pool 160 .

Redundancy levels for different files are also set dynamically, where files with important data are replicated in more locations across the network, which provides better protection against storage failures.

Of course, the above examples and descriptions are provided only to illustrate the present invention and are not intended to limit the present invention in any respect. Those skilled in the art should understand that the present invention can be implemented in a large number of ways using more than one of the technologies described above, and they all belong to the scope of the present invention.

Claims (30)

1. A system for managing memory resources on a network, the system comprising: a plurality of memory nodes located on the network, each node being respectively associated with a physical memory resource; a management server, located on the network, for concentrating physical memory resources associated with the memory node into a pool of virtual memory resources; and A storage client for accessing virtual storage resources in a pool centralized by the management server. 2. The system of claim 1, wherein the virtual memory pool includes a plurality of virtual segments, and wherein the virtual segments are adapted to be stored on physical memory resources. 3. The system of claim 2, wherein the virtual segment is set on a virtual storage volume, and wherein, to a storage client, the virtual storage volume appears as a physical storage resource. 4. The system of claim 1, wherein the total number of virtual memory within the pool exceeds the total number of physical memory resources on the network. 5. The system of claim 1, wherein the management server is adapted to monitor access of virtual memory resources by memory clients and dynamically allocate virtual memory resources to physical memory resources based on the access. 6. The system of claim 5, wherein physical memory resources are characterized by performance parameters, and wherein the management server dynamically allocates virtual memory resources to physical memory resources. 7. The system of claim 5, wherein dynamic allocation is transparent to memory clients. 8. The system of claim 5, wherein the management server is adapted to dynamically allocate virtual memory resources to physical memory resources according to the level of use of virtual memory by memory clients. 9. The system of claim 5, wherein the memory client is adapted to execute a plurality of applications, and wherein the management server is adapted to monitor access to virtual memory resources by each of the plurality of applications , and the virtual memory is dynamically allocated to each of the plurality of application programs according to accesses made by the application programs. 10. The system of claim 1, wherein the memory client accesses data held by a plurality of virtual memory resources, and wherein the memory client is further adapted to test the plurality of virtual memory resources holding the data, and A set of optimal virtual memory resources from which to access the data is identified. 11. The system of claim 10, wherein the memory client further comprises: A load balancer adapted to select a virtual memory resource among the groups from which the data is accessed. 12. The system of claim 1 , wherein the memory nodes on the network are inaccessible to the memory client, but are accessible to the intermediary computer system, and wherein the management server is adapted to use the intermediary computer system to make the memory client The machine accesses the physical memory associated with the memory node. 13. The system of claim 1, wherein the network comprises a plurality of regions, each region comprising a plurality of memory nodes, the system further comprising: A computer system has a local routing table that maps a pool of virtual memory resources to physical memory resources associated with a plurality of memory nodes in one of the regions. 14. A computer program product, the computer program product comprising: A computer-readable medium having computer program logic embedded therein for maintaining memory resources on a network comprising a plurality of memory nodes each associated with a physical memory resource, the network further comprising means for accessing A memory client of a memory resource, the computer program logic comprising: The management server logic is used to concentrate the physical storage resources related to the storage node into a virtual storage resource pool, and to provide the storage client with the virtual storage resources in the pool. 15. The computer program product of claim 14, wherein the virtual memory pool includes a plurality of virtual segments, and wherein the virtual segments are adapted to be stored on physical memory resources. 16. The computer program product of claim 15, wherein virtual segmentation is set on a virtual storage volume, and wherein the virtual storage volume appears to a storage client as a physical storage resource. 17. The computer program product of claim 14, wherein the management server logic is further adapted to monitor memory client accesses to memory resources and dynamically allocate virtual memory resources to physical memory resources based on the accesses. 18. The computer program product of claim 17, wherein physical memory resources are characterized by performance parameters, and wherein the management server logic dynamically assigns virtual memory resources to allocated to physical memory resources. 19. The computer program product of claim 17, wherein the memory client is adapted to execute a plurality of applications, and wherein the management server logic is adapted to access memory resources by each of the plurality of applications Monitoring is performed and virtual memory resources are dynamically allocated to each of the plurality of applications based on accesses made by the applications. 20. The computer program product of claim 14, wherein the memory client accesses data held by a plurality of virtual memory resources, the computer program product further comprising: Test logic for testing a plurality of virtual memory resources holding the data and identifying an optimal set of virtual memory resources from which the memory client should access the data. 21. The computer program product of claim 20, further comprising: Load balancer logic adapted to select the virtual memory resource among the group from which the memory client accesses the data. 22. A method for managing memory resources on a network, the method comprising: identifying a plurality of memory nodes on the network, each node being individually associated with a physical memory resource; concentrating physical memory resources associated with the memory node into a pool of virtual memory resources; and Virtual memory resources in the pool are provided to memory clients based on memory clients accessing memory resources on the network. 23. The method of claim 22, wherein the virtual memory pool includes a plurality of virtual segments, and wherein the virtual segments are distributed among physical memory resources. 24. The method of claim 23, wherein virtual segmentation is set on a virtual memory volume, and wherein, to a memory client, the virtual memory volume appears as a physical memory resource. 25. The method of claim 22, wherein the providing step comprises: monitoring accesses to virtual storage by storage clients; and Based on the access, virtual memory resources are dynamically allocated to physical memory resources. 26. The method of claim 25, wherein the physical memory resources are characterized by performance parameters, and wherein the dynamic allocation step comprises: Based on the performance parameters and the nature of the accesses made by the memory clients, virtual memory resources are allocated to physical memory resources. 27. The method of claim 25, wherein the dynamic allocation step comprises: The virtual memory resources are allocated to the physical memory resources according to the level at which the virtual memory is used by the memory clients. 28. The method of claim 22, wherein the memory client accesses data held by a plurality of virtual memory resources, and the method further comprises: testing the multiple virtual memory resources holding the data; and From this test, an optimal set of virtual memory resources from which the memory client can access the data is identified. 29. The method of claim 28, further comprising: A virtual memory resource is selected among the groups from which the memory client will access the data. 30. The method of claim 22, further comprising: identifying a new memory node on the network that is associated with a new physical memory resource; and A portion of the virtual memory resource is allocated to the new physical memory resource.

Images