Electrical computers and digital data processing systems: input/ – Input/output data processing – Input/output access regulation
Reexamination Certificate
2001-06-06
2004-06-29
Gaffin, Jeffrey (Department: 2182)
Electrical computers and digital data processing systems: input/
Input/output data processing
Input/output access regulation
C710S036000
Reexamination Certificate
active
06757753
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to data storage in a computerized storage system, such as a storage area network (SAN). More particularly, the present invention relates to a new and improved technique of handling data access requests within the storage system in such a manner that the devices that issue the requests (e.g. “servers”) have improved flexibility in selecting the path through which to send the requests without adversely affecting the ability of the devices that receive the requests (e.g. “storage arrays”) to respond to the requests.
BACKGROUND OF THE INVENTION
Current prior art high-capacity computerized data storage systems, such as the one shown in
FIG. 1
, typically involve a storage area network (SAN)
100
within which one or more conventional storage arrays
102
store data on behalf of one or more servers
104
. The servers
104
typically service data storage requirements of several client devices
106
, as shown in FIG.
1
. The servers
104
are typically connected through switches or SAN's, such as Fibre Channel (FC) SAN fabrics
108
, to the storage arrays
102
. The servers
104
access a plurality of logical volumes
110
present on the storage arrays
102
on behalf of the client devices
106
.
Each storage array
102
typically includes a bank
112
of individual storage devices (not shown, e.g. hard drives, compact disk (CD) drives, tape drives, etc.), typically arranged in a RAID (Redundant Array of Independent Drives) configuration. The RAID storage devices supply data storage space for the logical volumes
110
. The logical volumes
110
are commonly striped across multiple storage devices in the banks
112
of storage devices, and may be striped across multiple storage arrays
102
. The servers
104
that access a given logical volume
110
must have a striping definition for the logical volume
110
if the logical volume
110
is striped across multiple storage arrays
102
and must have a connection, or path, to the storage array
102
that contains the logical volume
110
, or a portion thereof. A manager device
113
typically sets up the logical volumes
110
and monitors for problems, such as a storage device that has failed or is about to fail. Through a discovery process, the servers
104
typically discover the logical volumes
110
on the storage arrays
102
and the array controllers
114
through which the servers
104
can access the logical volumes
110
. The servers
104
are thus configured to use the discovered logical volumes
110
.
Each storage array
102
also commonly includes more than one array controller
114
, through which the storage devices and logical volumes
110
are accessed. Each array controller
114
typically connects to one of the switched fabrics
108
. Thus, a data transfer path between one of the servers
104
and one of the array controllers
114
in one of the storage arrays
102
is established from a host bus adapter (HBA)
116
in the server
104
, through the switched fabric
108
(to which the host bus adapter
116
is attached), and to the array controller
114
(to which the switched fabric
108
is attached).
Some of the servers
104
may access the same logical volume
110
through more than one data transfer path through the switched fabrics
108
on behalf of the same or different client devices
106
. Therefore, more than one array controller
114
of a storage array
102
may receive a data access request to the same logical volume
110
, or portion thereof, present on the storage array
102
.
When one of the array controllers
114
of a given storage array
102
processes a data access request to a given logical volume
110
, that array controller
114
is said to have access control or “ownership” of the logical volume
110
. When one array controller
114
has ownership of the logical volume
110
, no other array controller
114
in the storage array
102
can access the logical volume
110
without transferring ownership to itself, due to cache coherency issues.
When an array controller
114
receives a data access request to a logical volume
110
that the array controller
114
does not currently own, it transfers ownership of the logical volume
110
to itself in an automatic volume transfer (AVT) process and then processes the data access request. Upon transfer of ownership, the array controller
114
giving up ownership typically must “flush” cached data to the storage devices, so the array controller
114
that is receiving ownership will have the correct, up-to-date data in the storage devices. The time required to perform the cache flush, however, degrades the overall performance of the storage array
102
. Additionally, the data access request issued by the server
104
may “timeout” causing the server
104
to erroneously determine that the array controller
114
is not operating if the cache flush takes too much time. Furthermore, when the same logical volume
110
is repeatedly accessed through different array controllers
114
, then the array controllers
114
repetitively transfer ownership of the logical volume
110
back and forth between themselves. The repetitive ownership transferring is called “thrashing.” Thrashing can severely degrade the performance of data accesses to the affected logical volume
110
, since significant time is taken up performing the AVT processes instead of accessing the affected logical volume
110
.
When the servers
104
access the same logical volume
110
through the same common array controller
114
, then thrashing and unnecessary AVT processes are avoided, because the common array controller
114
can process all of the data access requests. However, it is sometimes advantageous for one or more of the servers
104
to switch between its host bus adapters
116
for accessing the same logical volume
110
, even when no array controller
114
has failed. In this manner, the servers
104
can optimize the use of their host bus adapters
116
, but at the expense of thrashing between the array controllers
114
, since the host bus adapters
116
are connected to the storage array
102
through different data transfer paths.
Additionally, the servers
104
must be able to discover the logical volumes
110
to be able to configure themselves and to route the data access requests to the appropriate array controllers
114
. As the size of the SAN
100
and the number of the storage arrays
102
increases, however, the complexity of the tasks of the servers
104
increases significantly.
Furthermore, whenever a new storage array
102
or storage device is added to (or deleted from) the SAN
100
or whenever the distribution of a logical volume
110
across the banks
112
of storage devices or across the storage arrays
102
is changed, the servers
104
must be reconfigured with respect to volume ownership to handle the changes. The reconfiguration takes time away from handling data access requests by the servers and becomes considerably more complex as the number of storage arrays
102
increases.
It is with respect to these and other background considerations that the present invention has evolved.
SUMMARY OF THE INVENTION
The present invention enables servers in a storage system to issue data access requests through different data transfer paths without causing ownership of the logical volumes to thrash back and forth between controllers of a storage array. Thus, the flexibility of the performance of the servers is increased without sacrificing the performance of the controllers.
In accordance with these features, routing capabilities are included in I/O (input/output) modules that control access to the storage devices. The routing capabilities are also preferably included in other types of devices (e.g. intelligent storage hubs, network attached storage appliances, snapshot/backup appliances, remote data appliances, etc.) that control access to the storage arrays.
The I/O modules (or other routing-capability devices) receive the data access requests from the servers and, independently of the data transfer path through whi
DeKoning Rodney A.
Weber Bret S.
Duft Setter Ollila & Bornsen LLC
Gaffin Jeffrey
LSI Logic Corporation
Martinez David
LandOfFree
Uniform routing of storage access requests through redundant... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Uniform routing of storage access requests through redundant..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Uniform routing of storage access requests through redundant... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3297456