Presentation of virtual arrays using n-port ID virtualization

The present invention relates generally to the field of storage systems, and particularly to ways of presenting virtual arrays.

Today's enterprise data centers store ever-larger amounts of business critical data that must be immediately and continuously available. Ever larger and more complex storage systems are used for storage of the data. Many different hosts and applications access data on these storage systems. In order to provide security and prevent data corruption, it is often necessary to ensure that the applications and hosts have exclusive access to particular areas of storage in the system.

One mechanism for partitioning storage systems employs the concept of “virtual arrays”. Accordingly, software is provided within a storage array to logically partition the array into separate storage groups. These groups are accessible only to hosts that have been granted access by the storage array. Other hosts cannot access a storage group to which they have not been granted access. Unfortunately, the current methods for partitioning storage arrays into virtual arrays are highly complex and expensive, and operate only at the storage array level. It is desirable to provide a simpler, inexpensive means of presenting virtual arrays to host systems, and to provide a way of centralizing array partitioning from another part of the system—for example, the fabric.

In accordance with the principles of the invention, a storage array includes a plurality of groups of logical units of storage. At least one physical port is coupled to the groups. The groups are coupled to a switch through the at least one physical port. Each group is assigned a unique virtual port ID for each physical port to which it is coupled. Further in accordance with the invention, the virtual port IDs are assignable by the switch. The virtual port IDs are is used by hosts coupled to the switch to exchange data with the groups to which the virtual port IDs are assigned.

The invention is particularly applicable in a Fibre Channel fabric environment. Each group of logical units thereby has its own unique virtual port ID through which it can be addressed by a host. The groups thus appear to the host as separate virtual arrays In accordance with a further aspect of the invention, the switch includes host facing ports. Each host is coupled to the switch through a host facing port. A zoning table in the switch associates each virtual port ID to a host facing port. Each group communicates only with hosts coupled to host facing ports associated with the virtual ID assigned to the group. Each group now appears to a host as a separate physical storage array.

In order to facilitate a fuller understanding of the present invention, reference is now made to the appended drawings. These drawings should not be construed as limiting the present invention, but are intended to be exemplary only.

FIG. 1 is a representation of a storage area network. The storage area network includes a Fibre Channel array and hosts coupled to a Fibre Channel fabric switch.

FIG. 2 is a representation of the storage area network of FIG. 1 , showing Fibre Channel port IDs and LUN names and numbers.

FIG. 3A is a flow diagram representing the operation of the array controller during Fibre Channel system initialization.

FIG. 3B is a flow diagram representing the operation of the switch during Fibre Channel system initialization.

FIG. 4 is a representation of a name table managed by the name server database in the switch.

FIG. 5 is a representation of a LUN table in a host.

FIG. 6 is a representation of a storage area network wherein the LUNs in the storage array are arranged into groups and are separately addressable via virtual port IDs in accordance with the invention.

FIG. 7A is a flow diagram representing the operation of the array controller during Fibre Channel system initialization when virtual port IDs are implemented on the storage array.

FIG. 7B is a flow diagram representing the operation of the switch during Fibre Channel system initialization when virtual port IDs are implemented on the storage array.

FIG. 8 is a representation of the name table in the switch in accordance with the invention.

FIG. 9 is a representation of a LUN table in a host in accordance with the invention.

FIG. 10 is a representation of the system of FIG. 6 wherein the name server database in the switch further includes a zoning table.

FIG. 11 is a representation of the zoning table of FIG. 10 .

FIG. 12A is a representation of a LUN table in one host when the zoning table in the switch is operational.

FIG. 12B is a representation of a LUN table in a second host when the zoning table in the switch is operational.

FIG. 13 is a flow diagram representing the operation of the switch when a zoning table is in use.

In FIG. 1 there is shown a functional block diagram of an exemplary storage area network 10 in which the invention can be implemented. The storage area network 10 employs a Fibre Channel fabric topology. Fibre Channel is a high speed serial transport used in storage systems. It is described in a series of standards that can be found at X3T9.3 Task Group of ANSI: Fibre Channel Physical and Signaling Interface (FC-PH), Rev. 4.2 Oct. 8, 1993. Hosts 12 , shown individually as 12 a , 12 b , 12 c , 12 d and 12 e are coupled to a Fibre Channel “fabric” in the storage system, herein shown as a switch 14 . A storage array 16 including disk drives 18 is also coupled to the switch 14 . The hosts 12 communicate with the disk drives 18 via a cross-point Fibre Channel connection through the switch 14 .

The switch 14 includes switch ports 20 . Host facing switch ports are labeled as 20 h . Array facing switch ports are labeled as 20 a . Host ports 22 on the hosts are coupled via Fibre Channel links 24 to host-facing switch ports 20 h on the switch 14 . Physical array ports 26 on the array 16 are coupled via Fibre Channel links 24 to array-facing switch ports 20 a on the switch 14 . The disks 18 within the array 16 are organized into logical units (“LUNs”) 30 . “LUN”, originally a SCSI (small computer system interface) term, is now commonly used to describe a logical unit of physical storage space. The LUNs are exported by the array ports 26 for access by the hosts 12 via the fibre channel links 24 and switch 14 . As herein shown, each disk appears to be configured as a separate LUN, though it is understood that a LUN can encompass part of a disk, or parts of multiple disks, or multiple complete disks. The arrangement shown is chosen for convenience of description.

In a Fibre Channel system such as that of FIG. 1 , each Fibre Channel device (including but not limited to host ports and array ports) has two identifying characteristics—a name and an address. Fibre Channel names, known as “world wide names”, are unique—every Fibre Channel device in the world has its own unique name. Each Fibre Channel device in a system also has an address, referred to in Fibre Channel parlance as an “ID”, that is dynamic and dependent upon the configuration of the system. The IDs are used for directing information between hosts and arrays in the system. Port addresses are commonly referred to as “port IDs”. LUN addresses are commonly referred to as “LUN numbers”. After initialization, the hosts 12 communicate with the array ports 26 and LUNs 30 by sending messages to the appropriate port ID and LUN number.

In a Fabric topology, the switch 14 assigns IDs to the host ports 22 and array ports 26 during initialization. IDs as described in the Fibre Channel specification are actually 24 bit quantities containing several fields. In FIG. 2 names and IDs are shown symbolically with alphanumeric symbols for simplicity of explanation. The names and IDs of each port and LUN are shown as a pair (name, ID). For instance, the host port 22 on the host 12 a is shown to have a name and ID of (ha, 100 ). The names and IDs for the ports 22 on the hosts 12 b - e are (hb, 101 ), (hc, 102 ), (hd, 103 ) and (he, 104 ). The host-facing switch ports 20 h are shown to have names and IDs (hfa, 200 ), (hfb, 201 ), (hfc, 202 ), (hfd, 203 ), and (hfe, 204 ). Array-facing switch ports 20 h are shown to have names and IDs (afa, 300 ) and (afb, 301 ). The array ports 26 are shown to have names and IDs (apa, 0 ), and (apb, 1 ). Each LUN 30 has a name and LUN number. For example, the LUN of name a 0 is LUN number L 00 . LUN numbers L 00 , L 01 , L 02 , L 10 , L 11 , L 12 , L 20 , L 21 , L 22 , L 30 , L 31 , L 32 , L 40 , L 41 , L 42 , L 50 , L 51 , AND L 52 are shown as accessible via array port ID 0 . LUN numbers L 60 , L 61 , L 62 , L 70 , L 71 , L 72 , L 80 , L 81 , L 82 , L 90 , L 91 , and L 92 are shown as accessible via array port ID 1 .

The Fibre Channel switch 14 includes a name server database 40 . The name server database 40 is used by the switch 14 to assign IDs to host ports 22 and array ports 26 during initialization. The name server database 40 includes a name server table 42 that is used by the switch to resolve IDs to names. The general process by which port IDs are assigned in accordance with the Fibre Channel T1 standard is shown in FIGS. 3A and 3B . FIG. 3A shows the process by which the switch 14 assigns Port IDs. FIG. 3B shows the process by which the fibre channel array controller 44 in the array 16 ( FIG. 2 ) communicates with the switch 14 . First, each array port (i.e. ports 0 and 1 ) is logged in to the switch 14 by the array controller 44 ( FIG. 3A step 50 ). The port logins are received by the name server database 40 ( FIG. 3B step 52 .) When the switch 14 receives a port login (“FLOGI”) command, it responds by sending an acknowledgement message to the array controller 44 ( FIG. 3B step 54 ). This acknowledgement message contains a Port ID for the array port that was logged in. The acknowledgement message is received by the array controller 44 ( FIG. 3A step 56 ). Then, a registration process is performed by the switch 14 ( FIG. 3A step 58 , FIG. 3B step 60 ). During the registration process, the name server table 42 is built and distributed to nodes registered to receive it.

An example of the name server table 42 is shown in FIG. 4 . The full contents of the name server table 42 are described in the Fibre Channel Name Server MIB, described in the IETF RFC 4044 , “Fibre Channel Management MB”, herein incorporated by reference. FIG. 3 shows only enough of the table 42 to contribute to understanding of the invention. The table 42 includes multiple entries 62 , each including a port name field 64 and a port address ID field 66 . During the registration process of FIGS. 3A and 3B , the entries 62 are populated with the switch port 26 names and Port IDs assigned by the switch 14 . For the example array 16 shown in FIG. 1 , an entry 44 includes the port name apa and Port ID 0 , while another entry 62 includes the port name apb and Port ID 1 . The switch 14 then sends this table 42 to all members of the SAN 10 registered to receive state change notifications. This includes the hosts 12 . The hosts 12 now have the Port IDs of the array ports 26 so Fibre Channel communications between the hosts 12 and array 16 can ensue.

Now that the hosts have IDs to access the ports, they can learn what. LUNs are available. LUN names and numbers are managed at the array level. Each host 12 sends a query to each array port 26 ID in turn, requesting a list of available LUN numbers. Once the LUN numbers for a given array port ID are known, the host is able to query each LUN 30 by using a combination of the port ID and LUN number to access the LUNs 30 . The host 12 then queries each LUN 30 for its corresponding LUN name. Once the host has gathered all this information, it builds a directory LUN table 70 that relates LUN names, port IDs, and LUN numbers. A representation of such a LUN table 70 is shown in FIG. 5 . The table includes an entry 72 for each LUN it has discovered. Each entry includes a LUN name field 74 , a port ID field 76 and a LUN number field 78 , which in combination identify the LUN 30 . The table 70 for a host 12 of FIG. 2 includes the LUN names, port IDs, and LUN numbers for the LUNs 30 on the array 16 for each port ID 0 and 1 . For example, one entry 72 shows a LUN with name a 0 and LUN number L 00 associated with Port ID 0 . Another entry 72 shows a LUN with name k 0 and address L 31 associated with Port ID 0 . Yet another entry 72 shows a LUN with name e 1 and LUN number L 71 associated with Port ID 1 .

During operation, hosts 12 refer to LUNs 30 by their LUN numbers. In order to access a LUN 30 , a host 12 port 22 sends a message whose Fibre Channel address includes the array port ID and LUN number. The switch 14 parses the port ID portion of the address in order to forward the message to the identified array port 26 . The array 16 then uses the LUN number portion of the address to access the proper LUN 30 within the array 16 . So, for example, if host 12 a needs to access LUN #L 71 , the host 12 a port 22 sends a message to an address including the port ID 1 and the LUN number L 71 . The switch 14 sees the port ID 1 and sends the message to the array port 26 with ID 1 . The array sees that the message is directed to LUN # L 71 and thus proceeds to perform the appropriate operation on LUN #L 71 .

Note that, in accordance with the prior art arrangement of FIG. 2 , a host has actual access to all LUNs on each array port to which it has access. For example, the host 12 a has access to port ID 0 , and therefore has access to LUNs L 00 -L 52 .

It is often desirable to separate a storage array into several distinctly accessible sub-arrays, or “virtual arrays”. Each host or application has access to a virtual array, but does not have access to the other virtual arrays within the storage array. For example, it may be desirable to arrange the LUN numbers L 00 -L 12 as a first virtual array accessible only to the host 12 a , and LUN numbers L 20 -L 32 as a second virtual array accessible only to the host 12 b . Such an arrangement can provide security against data corruption and can provide ease of management for host applications. But, in the prior art example of FIG. 2 , all the LUNs L 00 -L 52 are exposed via the same port ID 0 , and thus cannot be hidden at the fabric level from either host 12 a or host 12 b . Virtual arrays have therefore previously been provided only through implementation of complex software on the storage array.

In accordance with the principles of the invention, the storage array and fabric are employed to present virtual arrays to the hosts. The LUNs in a storage array 16 are arranged into several storage groups. The term “storage group” can have different meanings in different contexts, so for clarity, a “storage group” as used herein is simply a group of LUNs. Virtual Port IDs are established over each physical port on the array. Each storage group has assigned to it at least one virtual port ID, used by the hosts to access the storage groups. Each storage group is thus separately accessible via at least one unique virtual port ID. A host 12 can access only the LUNs 30 in a storage group with a virtual port ID to which the switch 14 allows it access. As will be seen, the provision of unique virtual IDs for each storage group allows zoning to be applied by the switch 14 such that each host 12 has access to only designated storage groups. The storage groups thus appear as individual virtual arrays to the hosts 12 . Therefore, the storage groups will herein further be referred to as “presented virtual arrays”.

In FIG. 6 , the storage system 10 has been modified so that presented virtual arrays can be accessed by the hosts. The modified storage array 16 a is shown to include presented virtual arrays 210 a , 210 b , 210 c , 210 d , and 210 e . The LUN 30 storage groups forming the presented virtual arrays can be arranged for example by a modified fibre channel controller 44 a (shown) or by a separate controller, or by pre-configured software programming or dynamic user programming of the array 16 , or any combination thereof. Each presented virtual array 210 a - 210 e is associated with at least one “virtual port ID” v 0 -v 5 . Generally, each presented virtual array 210 a - 210 e is assigned one or more virtual port IDs depending upon how many physical ports the virtual array is accessible through. As shown by example, the presented virtual array 210 a is associated with the physical array Port ID 0 and is assigned one virtual port ID v 0 . The presented virtual array 210 b is also associated with the physical array Port ID 0 and is assigned one virtual port ID v 1 . The presented virtual array 210 c is associated with both the physical array ports Port ID 0 , 1 , and is thus assigned two virtual port IDs v 2 and v 5 . The presented virtual arrays 210 d and 210 e are both associated with the physical array port ID 1 and are assigned virtual port IDs v 3 and v 4 respectively.

In accordance with one implementation of the virtual Port IDs of the invention, the virtual port IDs are assigned by the modified switch 14 a . The ANSI T11 standard “ ”, which currently defines virtual ports used by hosts, is extended to support storage arrays. The process by which virtual Port IDs are provided by the switch 14 a is shown in FIGS. 7A and 7B . FIG. 7A shows the process by which the fibre channel array controller 44 a in the array 16 communicates with the switch 14 a . FIG. 7B shows the process by which the switch 14 a assigns Port IDs. First, each array port (i.e. ports 0 and 1 ) is logged in to the switch 14 a by the array controller 44 a ( FIG. 7A step 220 ). The port logins are received by the switch 14 a ( FIG. 7B step 222 .) When the switch 14 a receives a port login (“FLOGI”) command, it responds by sending an acknowledgement message to the array controller 44 a ( FIG. 7B step 224 ). This acknowledgement message contains a Port ID for the array port that logged in. The acknowledgement message is received by the array 16 controller 44 a ( FIG. 7A step 226 ). Then, a registration process is performed by the switch 14 a ( FIG. 7A step 228 , FIG. 7B step 230 ). During the registration process, the name server table 42 a is built as previously described with regard to name server table 42 .

Then, if virtual port IDs are needed by the array 16 a ports 26 (step 232 ), the array controller 44 a sends an “FDISC” command to the switch 14 a (step 234 ). The switch 14 a receives the FDISC command (step 236 ) and responds by sending a virtual port ID to the array controller 44 a (step 238 ). The array controller 44 a receives the virtual port ID from the switch 14 a (step 240 ). The switch 14 a and array controller 44 a then perform the registration process to add the virtual Port ID to the name server table 42 a , as will be described (steps 242 , 244 ). The FDISC command and response is repeated for each virtual ID required for each physical port (steps 232 - 244 ).

Now the switch 14 a can build the name server table 42 a in a manner similar to that previously described with respect to name server table 42 , except the name server table 42 a associates multiple virtual port IDs with the physical port names. An example of such a name server table 42 a is shown in FIG. 8 . The physical array port 26 with name apa and Port ID 0 is also associated with the virtual port IDs v 0 , v 1 , v 2 , and v 5 . Likewise, the physical array port 26 with name apb and Port ID 1 is associated with the virtual port IDs v 3 , v 4 , and v 5 . At this point, the switch 14 a can update the hosts 12 with the contents of the name server table 42 a as previously described. The hosts 12 will thereby receive all the virtual port IDs v 0 -v 5 .

Now that the hosts 12 have the virtual port IDs v 0 -v 5 , they can build their directory LUN tables in a manner similar to that previously described with regard to FIG. 5 , except that now each virtual port ID will be associated with LUN names and numbers. Accordingly, each host 12 sends a query to each virtual array port ID v 0 -v 5 in turn, requesting a list of available LUN numbers. Once the LUN numbers for a given virtual array port ID are known, the host is able to query each LUN 30 by using a combination of the virtual port ID and LUN number to access the LUNs. The host 12 then queries each LUN 30 for its corresponding LUN name. Once the host has gathered all this information, it builds a directory LUN table 50 a that relates LUN names, virtual port IDs, and LUN numbers. A representation of such a LUN table 70 a is shown in FIG. 9 . The table includes an entry 72 a for each LUN it has discovered. Each entry includes a LUN name field 74 a , a virtual port ID field 76 a and a LUN number field 78 a , which in combination identify the LUN. For example, the LUNs 30 associated with the presented virtual array 210 a —i.e. virtual port ID v 0 —are (a 0 , L 00 ), (b 0 , L 01 ), (c 0 , L 02 ), (d 0 , L 10 ), (e 0 , L 11 ), and (f 0 , L 12 ). Note also that the presented virtual array 210 c (LUN#s L 40 -L 52 ) is dual-ported; that is, it is accessible via virtual port IDs v 2 and v 5 . LUNs in dual-ported presented virtual arrays may have two separate LUN numbers as well, one for use when accessed on the first virtual port ID, and the second for use when accessed on the second virtual port ID. Dual-ported presented virtual arrays can be useful for high availability purposes.

In accordance with one advantage of the invention, storage array “zoning” can be implemented at the fabric switch in order to physically separate the presented virtual arrays for access only by certain hosts. Fibre Channel switches are able to implement zoning, whereby access between host ports and array ports is specified. But zoning can only be implemented at the port level; that is, it cannot be implemented at the LUN level. In the prior art arrangement of FIG. 2 , zoning cannot be used to separate the storage groups of LUNs 210 a , 210 b , 210 c as shown FIG. 3 , because all the LUNs have the same Port ID 0 .

But in accordance with this further aspect of the invention, since each presented virtual array 210 a - e is associated with its own unique virtual Port ID v 0 -v 5 , the switch 14 a can differentiate between each presented virtual array 210 a - e based upon its virtual Port ID. The switch 14 a can be programmed to allow or disallow access to each virtual port address from each host facing array port address through the use of its zoning process. Host access to the presented virtual arrays 210 a - e can thus be physically separated, enhancing security, data integrity, and ease of storage management for host applications.

Referring now to FIG. 10 , the switch name server database 40 a is shown to include a zoning table 43 a as well as a name table 42 a . The full contents of a zoning table as used by a prior art fibre channel switch is described at “ ”. The zoning table 43 a is modified in accordance with the invention to include virtual array port IDs. Only the portions of the zoning table 43 a relevant to an understanding of the invention are shown here. Accordingly, the zoning table 43 a has entries 252 including a virtual port ID field 254 and a host facing switch port ID field 256 . For each virtual port ID recognized by the switch 14 , the name server database 40 a can associate one or more host facing switch port IDs with the virtual port ID. For example, in the table shown, virtual port ID v 0 is associated with host facing switch port ID 200 . Virtual port ID v 1 is associated with host facing switch port ID 201 . Virtual port ID v 2 is associated with host facing switch port ID 202 . Virtual port ID v 4 is associated with host facing switch port ID 204 . Virtual port ID v 5 is associated with host facing switch port ID 202 . (Host facing array port 202 is dual ported on virtual port IDs v 2 and v 5 .)

Now, when the switch 14 a updates the hosts 12 with the contents of the name server table 42 a , it uses the zoning table 43 a to filter the presentation of the name server table 42 a information to the hosts 12 . Referring to FIG. 12 there are shown several examples of the LUN tables 70 a in the hosts 12 a and 12 b . When the switch 14 a updates the hosts 12 , the switch 14 a refers to the zoning table 43 a —for example the table of FIG. 11 . The process by which the switch 14 a uses the zoning table 43 a is shown in FIG. 13 . The switch 14 a checks the zoning table 43 a for the first host facing port ID 200 (step 260 ). The switch 14 a sees that the host facing array port ID 200 is authorized to access only virtual array port v 0 . So, only the name table information associated with virtual port ID 0 is forwarded to the host 12 a that is coupled to the host facing switch port ID 200 (step 262 ). The host 12 a thus receives the address information for the virtual port v 0 . The LUN table 70 a for the host 12 a thus looks like that shown in FIG. 12A . The host 12 a is limited to access to the LUNs L 00 -L 12 on virtual Port ID v 0 . The switch 14 a then checks the zoning table 43 a for access information for the host facing switch port ID 201 (steps 266 , 268 , 260 ). The switch 14 a sees that the host facing switch port ID 201 is authorized to access only virtual array port v 1 . So, only the name table information associated with virtual array port v 1 is forwarded to the host 12 b that is coupled to the host facing switch port ID 201 (step 262 ). The host 12 b thus receives the address information for the virtual port v 1 . The LUN table 70 a for the host 12 b thus looks like that shown in FIG. 12B . The host 12 b is limited to access to the LUNs L 20 -L 32 on virtual Port ID v 1 . This process continues for each host facing switch port ID (steps 266 - 262 ).

Now each host has access to only the LUNs 30 on the virtual array ports allowed by the zoning table 43 a in the switch 14 , rather than to all LUNs 30 on a physical array port 26 . The invention thus allows a very simple and efficient means of presenting virtual arrays to the hosts, without requiring complex array level software.

The present invention is not to be limited in scope by the specific embodiments described herein. Various modifications of the present invention, in addition to those described herein, will be apparent to those of ordinary skill in the art from the foregoing description and accompanying drawings. For example, the disclosed controllers can be implemented in hardware, software, or both. All such modifications are intended to fall within the scope of the invention. Further, although aspects of the present invention have been described herein in the context of a particular implementation in a particular environment for a particular purpose, those of ordinary skill in the art will recognize that its usefulness is not limited thereto and that the present invention can be beneficially implemented in any number of environments for any number of purposes.