Storage architecture providing scalable performance through...

Details Storage architecture providing scalable performance through... Storage architecture providing scalable performance through...

1999-03-04

2002-02-19

Kim, Matthew (Department: 2186)

Electrical computers and digital processing systems: memory

Storage accessing and control

Specific memory composition

C710S120000, C711S154000

Reexamination Certificate

active

06349357

ABSTRACT:

BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention broadly relates to storage systems having at least one controller that manages data transfers between a host computer system and one or more storage devices. In particular, the present invention relates to a storage system having data transfer bandwidth independently scalable of I/O operation execution rate.
2. Description of the Related Art
Auxiliary storage devices, such as magnetic or optical disk arrays, are usually preferred for high-volume data storage. Many modern computer applications, such as high resolution video or graphic displays involving on-demand video servers, may heavily depend on the capacity of the host computer to perform in a data-intensive environment. In other words, necessity for external storage of data in relatively slower auxiliary data storage devices demands that the host computer system accomplish requisite data transfers at a rate that does not severely restrict the utility of the application that necessitated high-volume data transfers. Due to the speed differential between a host processor and an external storage device, a storage controller is almost invariably employed to manage data transfers to/from the host from/to the storage device.
The purpose of a storage controller is to manage the storage for the host processor, leaving the higher speed host processor to perform other tasks during the time the storage controller accomplishes the requested data transfer to/from the external storage. The host generally performs simple data operations such as data reads and data writes. It is the duty of the storage controller to manage storage redundancy, hardware failure recovery, and volume organization for the data in the auxiliary storage. RAID (Redundant Array of Independent Disks) algorithms are often used to manage data storage among a number of disk drives.
FIG. 1
shows a computer system
10
having a conventional storage controller
14
linking the host computer
12
with the external auxiliary storage device
16
. The storage device
16
may include more than one disk drives and may also employ different RAID levels to store the data received from the host
12
. The connecting links
13
and
15
may employ fibre channels, SCSI (Small Computer System Interface) interface, FC-AL (Fibre Channel Arbitrated Loop) interface, HIPPI (High Performance Parallel Interface) interface, USB (Universal Serial Bus) interface, ATM (Asynchronous Transfer Mode) interface, FireWire (High Performance Serial Bus) interface, an SSA (Serial Storage Architecture) interface or any other suitable interface standard for I/O data transfers.
As shown in
FIG. 1
, the conventional storage controller
14
receives every command, status and data packet during the host-requested data transfer. In other words, every binary information packet passes through the controller
14
. An exemplary flow of packets during a data read operation through the conventional storage controller
14
is illustrated in FIG.
2
. The data transfer protocol in
FIG. 2
is typically a two-party point-to-point communication protocol, e.g. fibre channel protocol. The links
13
and
15
have been assumed to be fibre channels. However, the discussion of the general flow of packets depicted in
FIG. 1
holds true for other interfaces as well.
Referring now to
FIGS. 1 and 2
together, a read command identifying the storage controller
14
as its recipient (XID=A) is issued from the host
12
to the storage controller
14
over the link
13
. The storage controller
14
performs necessary command translation and transmits another command packet to the storage device
16
over the link
15
. The command packet from the controller
14
identifies the storage device
15
as its intended recipient (XID=B) and functions to instruct the storage device
16
to initiate the necessary data transfer, i.e. to transmit the host-requested data (as identified by the command from the controller
14
). The storage drive or storage device
16
accesses the requested data and transmits data and status packets to the controller
14
over the interface link
15
. The status packet may indicate to the controller
14
whether the read operation was successful, i.e. whether the data read was valid. The controller
14
then inserts its own ID (XID=A) into the received data and status packets and forwards those packets to the host
12
. This completes the read operation initiated by the host
12
. In the event that the status signals from the storage
16
indicate a faulty data read operation, the host
12
may reinitiate or abort the previous read transaction. In general, the packet transmission depicted in
FIG. 2
is typical of a two-party point-to-point interface protocol, e.g. a fibre channel data transfer protocol.
Two parameters play a major role in measuring the performance of a storage system: (1) Input/Output (I/O) operations per second (iops), and (2) Data transfer rate. Generally, rate of execution of iops by a storage controller is governed by the type, speed and number of CPU's within the controller. However, the data transfer rate depends on the storage controller internal bandwidth that is dedicated for data transfer. Current storage systems have restricted scalability because of the storage controllers having a relatively inflexible ratio of CPU to bandwidth capability. In other words, as shown in
FIGS. 1 and 2
, the data transfer between the host and the storage is made dependent on the control functions (i.e., command and status packets) executed by the storage controller. This interdependence or interlocking of iops with the data transfer results in less efficient scalability of performance parameters. For example, in the conventional storage controller architectures, an increase in the data transfer bandwidth may unnecessarily, and sometimes quite expensively, require a similar increase in the number of CPU's residing within the controller.
Therefore, it is desirable to have a storage controller where control functionality (as measured by the iops parameter) is scalable independently of the data transfer bandwidth (which determines the data transfer rate), and vice versa. It may be further desirable to achieve independence in scalability without necessitating a change in the existing interface protocol managing the host-controller-storage interface.
SUMMARY OF THE INVENTION
The problems outlined above are in large part solved by a storage system as disclosed herein. The storage system includes a storage controller whose architecture is organized into functional units. The control function and the data transfer functions are separated so as to allow independent scalability of one or both. In other words, the storage system may be viewed as including a combination of a control module and a data switch. The command and status information may go to/from the control module, but the data may be moved directly between the host and the storage via the data switch. Thus, the control module and, hence, control functions or iops are effectively separated from the physical data transfer path. This allows data paths to be sized to the required bandwidth independently of the rate of execution of control packets by the control module or of controller bandwidth. Similarly, the number of control modules may be chosen independently of the data transfer function to meet the iops requirement.
Broadly speaking, a computer system according to the present invention includes a storage controller coupled to the host computer and the storage device. The storage controller includes a switch that links the host computer with the storage device via a control path and a data path. In one embodiment, the control and the data paths may be at least partially physically separate from each other. The control module of the storage controller is coupled to the switch through the control path. Any data transfer command from the host computer is transmitted over the control path, and, hence, passes through the control module. However, data tra

Affiliated with

Also associated with

Rating

Storage architecture providing scalable performance through... does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Storage architecture providing scalable performance through..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Storage architecture providing scalable performance through... will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2962674