1. Field of the Invention
The present invention relates generally to networks, and more particularly to a way to assign flexible prefixes to Switches in Fibre Channel Fabrics while using the currently defined FC_ID address space. This allows end devices in different Fibre Channel Fabrics to communicate with one another, without requiring modifications to existing end devices, or the need to perform Network Address Translation between Fabrics.
2. Description of the Related Art
With the increasing popularity of Internet commerce and network centric computing, businesses and other organizations are becoming more and more reliant on information. To handle all of this data, storage area networks or SANs have become very popular. A SAN typically includes a number of storage devices, a plurality of Hosts, and a number of Switches arranged in a Switching Fabric that connects the storage devices and the Hosts.
Most SANs rely on the Fibre Channel protocol for communication within the Fabric. For a detailed explanation of the Fibre Channel protocol and Fibre Channel Switching Fabrics and Services, see the Fibre Channel Framing and Signaling Standard, Rev 1.90, International Committee for Information Technology Standards (INCITS), Apr. 9, 2003, and the Fibre Channel Switch Fabric-2, Rev. 5.4, INCITS, Jun. 26, 2001, and the Fibre Channel Generic Services-3, Rev. 7.01, INCITS, Nov. 28, 2000, all incorporated by reference herein for all purposes.
In a Fabric, each Host and storage device is identified by a three byte-wide Fibre Channel address, also called FC_ID. Today the address is statically subdivided in three fields denoted Domain_ID, Area_ID, and Port_ID, each one byte long respectively. Within a Fabric, each Switch is assigned a Domain_ID. The end devices attached to a particular Switch are assigned a FC_ID having the Domain_ID of that Switch. The Switch manages the allocation of the Area_ID and Port_ID fields for each end device to guarantee the uniqueness of the assigned addresses in that Domain_ID. For example, if a Switch is assigned a Domain_ID number five and the Switch subdivides its address space in two areas each having three connected end devices, then a possible Fibre Channel address FC_ID allocation is: 5:1:1, 5:1:2, 5:1:3, 5:2:1, 5:2:2, and 5:2:3 respectively.
Fibre Channel frames are used for communication between Hosts and storage devices within a Fabric. A Fibre Channel frame header carries the source and destination Fibre Channel addresses. When a Host wishes to access a storage device, the FC_IDs of the Host and the storage device are inserted into the source and destination fields of the header respectively. The Switches within the Fabric then route the frame to the target end device using the destination FC_ID. The target end device generates a response frame that includes its own FC_ID in the source field and the FC_ID of the Host in the destination field. The frame is then routed across the Fabric in a similar manner as that described above.
The information infrastructure within a large enterprise typically has a number of independent Fabrics, each dedicated to a different organization or application within the enterprise. For example, a large corporation may have Fabrics for the corporate department, sales organization, engineering group, etc. Each Fabric is separate and distinct. The Hosts of one Fabric cannot access or use a resource in another Fabric. The aforementioned arrangement has a number of disadvantages. The Hosts in a given Fabric can communicate only with the storage devices in that same Fabric. Typically there is no way that a Host in one Fabric can communicate with a storage device in a second Fabric. This arrangement is not only inefficient, it is expensive. Since storage devices cannot be shared among Fabrics, separate storage devices are required for each Fabric.
On occasion, it is desirable for resources to be shared across Fabrics, while keeping the Fabrics separate. For example, it may be convenient for a Host in a first Fabric to be able to access a tape drive in a second Fabric. Fibre Channel Fabrics maintain, in operation, several kinds of information, such as the Name Server database, or the Fabric Configuration Server database, or the Zoning database. In addition, Fibre Channel Fabrics notify end devices of any relevant change in the state of the Fabric itself and of other end devices. There are a number of reasons to interconnect different Fabrics without merging them into a single Fabric. The above mentioned databases typically grow more than linearly with the size of a Fabric. The size of the databases is therefore a limiting factor in the size of a given Fabric. Also the notification mechanism does not scale to big Fabrics. Interconnecting separate Fabrics allows confining the above mentioned databases and notifications inside each Fabric, to keep them manageable, while inter-Fabric protocols may allow communication between selected end devices across multiple Fabrics. A number of solutions to enable resource sharing have been proposed.
One proposed solution involves the virtualization of the end devices so that there are “local instances” of each end device in each Fabric. See for example US Patent Publication 2003/0012204. With this approach, a gateway is needed to couple two (or more) Fabrics. The gateway is required to perform FC_ID translations (i.e., Network Address translations or NATs) for the source and destination end devices. Several problems are associated with NAT translations. If the gateway that performs the translations fails, an alternative or fail-over path across a second gateway needs to be created. The second gateway must have the same state information of the failed gateway. This requires a great deal of management by the Fabric administrator. In addition, with certain FC frames, both the source and/or destination FC_IDs may be carried in the frame payload. A mechanism that identifies and translates these FC_IDs must therefore be provided. This solution also does not work if encryption or a proprietary protocol is used between the source and destination end devices because there is no way for the translating gateway to process the proprietary payloads or decrypt the frames to identify the source and destination FC_IDs possibly carried inside the payload. While in certain cases this approach may be enough, in the general case, the management overhead of this proposal is highly burdensome and is very not practical.
Another proposed solution is the extension of the address space beyond the current three bytes to include the addition of a source Fabric_ID field and a destination Fabric_ID. During communication, the source and destination Fabric_IDs are specified in the FC frame header along with the source and destination FC_IDs. While this proposal eliminates the NAT translation, it is also impractical because it is not backward compatible with existing networking infrastructure. Existing Switches and end devices do not recognize Fabric_IDs. To implement this solution, an entirely new networking infrastructure that supports the use of Fabric_IDs would be needed. In particular, not only the Switches, but also all the end devices need to be replaced. This solution is therefore extremely impractical.
A way to assign flexible prefixes to Switches in Fibre Channel Fabrics while using the currently defined FC_ID address space is therefore desired, in order to allow end devices in different Fibre Channel Fabrics to communicate with one another, without requiring modifications to existing end devices, or the need to perform Network Address Translation between Fabrics.
The present invention relates to a way to assign flexible prefixes to Switches in Fibre Channel Fabrics while using the currently defined FC_ID address space. This allows end devices in different Fibre Channel Fabrics to communicate with one another, without requiring modifications to existing end devices, or the need to perform Network Address Translation between Fabrics. With the present invention, the existing address space for each Switch includes a dynamically configurable number of host bits sufficient to address all the end devices coupled to the Switch and the Switch itself. The remaining bits, called the Switch prefix, are used to identify the Switch in the switching Fabric. In an alternative embodiment, the Switch prefix bits may be further divided into a configurable first sub-set of bits used to identify a specific Fabric (Fabric prefix) and a second configurable sub-set of bits used to identify the Switch in the Fabric (Switch_ID). Thus the flexible addressing scheme of the present invention enables end devices in different Fabrics to communicate with one another without expanding the Fibre Channel address space or the need to perform Network Address Translations.
The invention may best be understood by reference to the following description taken in conjunction with the accompanying drawings in which:
In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced without some or all of these specific details. In other instances, well known process steps have not been described in detail in order to avoid unnecessarily obscuring the present invention.
Referring to
Under the current Fibre Channel practice and standards, the Domain_ID, Area_ID and Port_ID sub-division of the FC_ID assigned to each end device is statically set. In other words, the 24 bits of the address space are divided so that the first byte is used to identify the Domain_ID of the Switch and the second and third bytes are used to identify a particular end device connected to the Switch. Since the address space allocated for Area_IDs and Port_IDs are each eight bits wide, a Switch could potentially address up to (216=65,536) end devices. Since the vast majority of Switches connect 16 or 32 end devices in a typical Fabric, and the biggest Switches have today up to 256 ports at most, the current static address space allocation scheme is terribly inefficient.
Referring to
With the current model of address allocation, to each of the Switches SW1, SW2 and SW3 is assigned one (or more) Domain_ID(s). A routing protocol, called FSPF, propagates Domain_IDs “reachability” information among the Switches of a Fabric, and allows each Switch to maintain a routing table. Assuming that Switches SW1, SW2, and SW3 of
With the present invention, the same Switch illustrated in
The present invention contemplates a flexible allocation of the Fibre Channel address space. Specifically, the 24 bits of the FC_ID are divided into host bits and Switch prefix bits in a flexible manner. The number of host bits is configured by the fabric administrator to satisfy the current number of end devices coupled to the Switch. The remaining bits, called the Switch prefix, are used to uniquely identify a Switch. In one embodiment particularly suited for inter-fabric communication, the bits of the prefix can be further configured into two sub-sets. The first sub-set (Fabric prefix) can be used to identify the Fabric to which a Switch belongs. The second sub-set (Switch_ID) can be used to identify a particular Switch within that Fabric. In other words, different Fabrics will have different Fabric prefix values. All the Switches within a single Fabric will have the same Fabric prefix value, but different Switch_ID values.
The number of host bits and Switch prefix bits for a particular Switch is configured by (i): ascertaining the total number of host_IDs needed for the Switch. The number of host_IDs is calculated by adding the number of end devices connected to the Switch plus one host_ID reserved for the Switch itself. As is described in more detail below, a host_ID for the Switch is needed for Switch-to-Switch communication; (ii) reserving the necessary number of host bits required to address each of the total number of host_IDs; and (iii) calculating the number of Switch prefix bits by subtracting the number of reserved host bits from the address width (24 bits). As previously noted, the Switch prefix bits can also be further divided into the Fabric prefix and Switch_ID sub-sets to differentiate Fabrics and the Switches within a Fabric.
Referring to
Referring to
Referring to
According to various alternative embodiments, the Fabric prefix is set to a specified size (e.g. eight bits wide) or it can be configured to any size, provided a sufficient number of bits are set aside for the Switch_ID field. This means that the number of bits in the Switch_ID field is flexible and is determined by the number of host bits needed by the Switch and the size of the Fabric prefix. In other embodiments of the invention, however, the number of bits in Switch_ID field can be fixed and the number of bits in the Fabric prefix can be flexible. Alternatively, both the Fabric Prefix and the Switch_ID fields can be variable or set to a fixed length (e.g., five bits wide each). As an example, a Fabric with a lot of Switches may be identified by a three bits long Fabric prefix, to leave more address space to identify Switches. A smaller Fabric may be identified by a longer Fabric prefix because it needs less address space to identify the Switches.
In current Fibre Channel Fabrics, the protocol used to compute the routing information needed to route frames between the Switches in the Fabric (routing protocol), is based on Domain_IDs and is called FSPF. When the FSPF protocol is implemented, a routing table is constructed that defines the Domain_ID and Port along the best path to each Switch in the Fabric. When an intra-switch communication having the domain controller address format (FF-FC-Domain_ID) is received, the Domain_ID is extracted and the communication is forwarded along the Port defined by the best path toward that Switch in the routing table.
With the present invention, the Domain_IDs have been replaced by a variable length Switch prefix. Accordingly, the FSPF protocol as defined herein, can not be used. Rather, the well known OSPFv3 or the IS-IS routing protocols may be adapted to Fibre Channel and used to construct a Fibre Channel routing table, or FSPF may be enhanced to convey prefixes information. Switch-to-Switch communications is then handled using the FC_ID that identifies a specific Switch without any need of special address format such as the domain controller address format mentioned above. For more information on the OSPFv3 or IS-IS routing protocols, see respectively the IETF document RFC 2740 and RFC 1142, incorporated herein by reference.
One feature achieved by updating the routing protocol as described above is to scale the other Fibre Channel control protocols to an environment composed of interconnected Fabrics. An intra-Fabric routing protocol suitable for use in an environment of interconnected Fabrics needs to be hierarchical, in the sense that it has to provide a way to differentiate between intra-Fabric and inter-Fabric routes in the routing messages. Both OSPFv3 and IS-IS are hierarchical, and FSPF may be enhanced to support this capability. The hierarchical routing protocol is then the foundation to make hierarchical also the remaining Fibre Channel protocols.
A Switch participating in a hierarchical routing protocol is in fact able to distinguish the prefixes (and so the Switches) belonging to its own Fabric from the prefixes belonging to different Fabrics, because the first ones are announced by the routing protocol as intra-Fabric routes, while the latter ones are announced as inter-Fabric routes. The Switch may then continue to use the existing databases and notification protocols for the intra-Fabric communication, while implementing new inter-Fabric protocols for the selected allowed inter-Fabric communications.
Referring to
In respect to the routing tables computed currently by FSPF, in which each entry contains a fixed length Domain_ID and one (or more) exit ports, each entry of the routing tables populated according to the present invention contains an indication of whether the route is internal or external, a variable length Switch Prefix, and one (or more) exit ports. In addition, while the routing tables computed today by FSPF may be searched using an exact match lookup technique, the routing tables populated according to the present invention are searched with a longest prefix match algorithm.
In operation, the routing table represented in
Referring to
By way of example, assume that a first fabric “Fabric 1” is identified by the Fabric Prefix ‘0000 0001’ and a second fabric “Fabric 2” is identified by the Fabric Prefix ‘0000 0010’. Two prefixes therefore need to be assigned to Gateway 30, one belonging to Fabric 1 and the other belonging to Fabric 2. Further assume that: (i) Gateway 30 has Switch Prefix ‘0000 0001 0000 0000 10’ in Fabric 1 and Switch Prefix ‘0000 0010 0000 0000 10’ in Fabric 2; (ii) Switches SW1, SW2, and SW3 have prefixes ‘0000 0001 0000 0000 0000 1’, ‘0000 0001 0000 0000 0001’, and ‘0000 0001 0000 0000 01’ respectively and that Switches SW4, SW5, and SW6 have prefixes ‘0000 0010 0000 0000 0000 1’, ‘0000 0010 0000 0000 0001’, and ‘0000 0010 0000 0000 01’ respectively; and (iii) finally, assume that Gateway 30 has to facilitate inter-Fabric communications among some end devices connected to Switch SW1 in Fabric 1, and Switch SW4 in Fabric 2.
An alternative embodiment of this invention is implemented by the application of the hierarchical routing protocols to the current FC_ID assignment model. This may allow re-using existing hardware (i.e., existing switching logics) with an updated control plane to provide better scalability properties to the current suite of Fibre Channel control protocols. The invention remains as described, but with 8 bits long fixed length prefixes (which reduce to Domain_IDs). This particular embodiment, having fixed length prefixes, allows the use of an exact match lookup technique to search the routing or forwarding tables, instead than requiring the longest prefix match algorithm.
Although illustrative embodiments and applications of this invention are shown and described herein, many variations and modifications are possible which remain within the concept, scope, and spirit of the invention, and these variations would become clear to those of ordinary skill in the art. For example, the present invention may be used also in Fabrics that implements extended addressing, i.e. in Fabrics that extend the address space by including in the FC header also a source Fabric_ID and a destination Fabric_ID in addition to the source and destination FC_IDs. Accordingly, the present embodiments are to be considered as illustrative and not restrictive, and the invention is not to be limited to the details given herein, but may be modified within the scope and equivalents of the appended claims.
Number | Name | Date | Kind |
---|---|---|---|
20020034178 | Schmidt et al. | Mar 2002 | A1 |
20020191615 | Paul et al. | Dec 2002 | A1 |
20060034302 | Peterson | Feb 2006 | A1 |
Number | Date | Country | |
---|---|---|---|
60488116 | Jul 2003 | US |