Electrical computers and digital processing systems: multicomput – Computer-to-computer session/connection establishing
Reexamination Certificate
2000-03-15
2004-11-30
Barot, Bharat (Department: 2155)
Electrical computers and digital processing systems: multicomput
Computer-to-computer session/connection establishing
C709S202000, C709S203000, C709S213000, C709S215000
Reexamination Certificate
active
06826613
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to transparent access to network attached storage devices, whether configured as SCSI over IP, NAS or NASD devices. In particular, the present invention provides a method and device for using a switch as a virtual storage device, with the advantage that physical devices can be added to, replaced on or removed from a network without reconfiguring network clients or applications running at levels above the network clients.
2. Description of Related Art
There is a trend towards use of cluster devices on networks to improve performance, fail over, load-balancing, robustness and other characteristics of network devices. In a cluster device, multiple network devices share the workload of what was originally handled by one device, increasing capacity and scalability while minimizing vulnerability to a single point of failure or a single bottleneck. Transparency can be achieved when addresses are available via mechanisms such as round-robin domain name service (“DNS”), where the cluster shares a single fully qualified domain name (“FQDN”) and the name resolution process returns different IP addresses for devices sharing the same FDQN. Protocols such as dynamic host configuration protocol (“DHCP”) have been widely adopted for allocating available addresses. However, allocation of addresses as an approach to transparency requires available addresses and DNS services. It also may require modification network client software and special or modified application software.
Sometimes it is desirable for backward compatibility with older network clients or due to lack of available addresses for cluster of devices to share one virtual IP address. To maintain transparency when cluster devices share the same virtual IP address, the network client must believe that the transport session endpoint is the virtual IP address. The client must address a single logical device without being aware that there are multiple physical devices. It is desirable for the network transport session to be able to change the device or endpoint within a cluster which is communicating with the client without the client being aware of the change. Having a technology that facilitates such transparency, without any need to change existing client software or IP stacks, may significantly increase the rate of introduction of new clustering technologies. Transparency technology also may facilitate the development of wireless systems.
For networks relying on the transport control protocol (“TCP”), there are three logical approaches for supporting cluster devices. One approach is to replay transport connections from one device to the next. The second approach is for the server to instruct the client software to use a specific device within a cluster. Finally, there is the approach to the present invention, to handoff connections transparently among devices within a cluster. A disadvantage of the replay approach is that it generates additional traffic and introduces latency. A disadvantage of the EAP approach is that it requires potentially significant structural changes to the IP stacks in the client and the use of IP options. In essence, the IP stack must be changed so that it understands the existence of a cluster and distinguishes among devices within the cluster. The handoff approach avoids these problems.
Handoffs clearly have benefits when working with clustered systems, server area networks, network attached storage, and other similar to loosely distributed models. Handoffs allow the systems to appear as a virtual IP host through which the transport connections are directly forwarded to the node being utilized; other nodes in the system are not affected. Resource utilization is more efficient and transparent fail over is more easily accomplished. Handoffs may help solve problems with address transparent leases, as in the proposed IP version 6 re-numbering. Handoffs also may aid servers in communication with a network address translation (“NAT”) device, if the NAT is performing a cluster-like role.
A variety of network devices may benefit from virtual IP addressing. Disk drives with built-in file systems, sometimes referred to as network attached storage devices (“NAS”), are one type of the device the would benefit from or function as a cluster. Web servers, database servers, networked computing clusters and load balancing servers also may benefit from virtual IP addressing. In general, any type of network device that would benefit from a cluster being addressed by a single virtual IP address may benefit from transparency technologies. Virtual addressing can be cascaded, so that a virtual IP cluster may appear as a single address within another virtual IP cluster. Network attached storage is prominent among the variety of network devices that may benefit from the present invention.
Network Attached Storage is a storage paradigm in which disks are detached from the server and placed on the network. Ideally, the server is removed from the datapath between client and data. The goal of a NAS system is to increase the overall performance of the system while reducing the total cost of ownership (TCO). New functionality, such as the appearance of infinite disk capacity and plug-and-play configuration, can be incorporated. Improved performance and functionality at a reduced TCO are made possible by the ever increasing disk and switch device capabilities. These capabilities allow offload of processing from a centralized server to smarter devices and possible elimination of the server itself.
Three different strategies can be pursued to develop a NAS solution. The ultimate strategy would be a serverless network attached storage. The strategy names are related to the client-disk datapath; Strategy 1—Server-centric NAS; Strategy 2—Serverless NAS; and Strategy 3—Master/Slave NAS.
A traditional file system is managed as a client/server system. The client accesses the server which has all the required disks integrated with the server. The storage is referred to as Server Integrated Disk (SID) storage. The server-centric strategy has begun migrating from the Server Integrated Disk model to use of internal (SCSI) communication paths across a network. The new model allows disks to be arbitrarily placed on the network. It relies on a form of networked SCSI (SCSI over IP) to attach the disks logically to the server. In this context, SCSI over IP is used in an inclusive sense, with the NetSCSI being a particular research implementation by the University of Southern California Information Sciences Institute of SCSI over IP. This scheme is referred to as Server Attached Disk (SAD). However, the Server Attached Disk model is not expected to yield performance gains. Its gains are expected in ease of use and total cost of ownership.
A “serverless” system is not truly serverless as some central point-of-control needs to exist to reduce system complexity, however the “server” may require insignificant resources. In a serverless system, the system's steady state is direct client and disk communications. Server interaction is an insignificant percentage of the communications. In the serverless approach, the central control point when required, would be a switch. A serverless approach requires significant changes to client systems, which is a major barrier to acceptance. A serverless system has been suggested by research work at Carnegie Mellon University (CMU) on Network Attached Secure Disks (NASD). In this context, NASD is used in an inclusive sense, to include Object Based Storage Devices (OBSD) and the proposed SCSI-4 standard. Carnegie Mellon's work also includes overlay systems to ensure backwards compatibility with existing networked file system protocols such as Sun's Network File System (NFS) and Microsoft's Common Internet File System (CIFS). Research results indicate high potential for system scalability. Use of a file overlay system, however, tends to defeat performance gains from the switch-based architecture.
NAS benefits from maintaini
Lee David C.
O'Connell Anne G.
Wang Peter S. S.
3Com Corporation
Barot Bharat
Beffel, Jr. Ernest J.
Haynes Beffel & Wolfeld LLP
LandOfFree
Virtually addressing storage devices through a switch does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Virtually addressing storage devices through a switch, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Virtually addressing storage devices through a switch will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3359346