Patent Application
20130311492
The invention relates to the field of computer systems, and, more particularly, to a computer implemented packet look-up mechanism for very large tables and very high throughput.
An Ethernet port may provide a connection between a computer and a computer network. For a 100 Gbps Ethernet port, the smallest 64 bytes packet duration time is 6.7 nanoseconds, which gives only 3 clock cycles at 500 Mhz for the logic doing the look up.
According to one embodiment of the invention, an improved computer system may include a controller having a computer processor, the controller to reduce insertion times and/or hashing collisions when interfacing with new components introduced to the controller. The system may also include a collision avoidance apparatus that reduces the hashing collisions by using a plurality of tables and a plurality of keys per bucket. The system may further include a hash apparatus in communication with the controller to map the plurality of keys to the plurality of tables where the hash apparatus uses a single hash logic to provide an avalanche effect when one key is changed which results in nearly half of bits changing in the plurality of tables.
The single hash logic may be based upon the Cuckoo algorithm. The single hash logic may comprise a configurable cyclic redundancy check polynomial. The hash apparatus may provide parallel table lookup of the plurality of tables based upon the avalanche effect.
The avalanche effect may be based upon orthogonal hash functions for each of the plurality of tables, and the single hash logic implements each orthogonal hash function. Each bit output of the single hash logic may comprise funneled results of the key bits.
The funneled results may be generated by an XOR function. The plurality of tables may be configurable. The global load of the plurality of tables may be configurable by controlling the number of bits for the single hash logic output.
Another aspect of the invention is a method for improving a computer system. The method may include reducing insertion times and/or hashing collisions when new components are introduced to a controller that includes a computer processor. The method may also include reducing the hashing collisions by using a plurality of tables and a plurality of keys per bucket through a collision avoidance apparatus. The method may further include mapping the plurality of keys to the plurality of tables with a hash apparatus in communication with the controller where the hash apparatus uses a single hash logic to provide an avalanche effect when one key is changed which results in nearly half of bits changing in the plurality of tables.
The method may additionally include providing parallel table lookup of the plurality of tables via the hash apparatus based upon the avalanche effect. The method may also include basing the avalanche effect upon orthogonal hash functions for each of the plurality of tables, and the single hash logic implements each orthogonal hash function.
The method may further include funneling results of the key bits for each bit output of the single hash logic. The method may additionally include making the plurality of tables configurable. The method may also include controlling the number of bits for the single hash logic output so global load of the plurality of tables is configurable.
Another aspect of the invention is computer readable program codes coupled to tangible media to improve a computer system. The computer readable program codes may be configured to cause the program to reduce insertion times and/or hashing collisions when new components are introduced to a controller that includes a computer processor. The computer readable program codes may also reduce the hashing collisions by using a plurality of tables and a plurality of keys per bucket through a collision avoidance apparatus. The computer readable program codes may further map the plurality of keys to the plurality of tables with a hash apparatus in communication with the controller where the hash apparatus uses a single hash logic to provide an avalanche effect when one key is changed which results in nearly half of bits changing in the plurality of tables.
The computer readable program codes may additionally provide parallel table lookup of the plurality of tables via the hash apparatus based upon the avalanche effect. The computer readable program codes may also base the avalanche effect upon orthogonal hash functions for each of the plurality of tables, and the single hash logic implements each orthogonal hash function.
The computer readable program codes may further funnel results of the key bits for each bit output of the single hash logic. The computer readable program codes may additionally make the plurality of tables configurable. The computer readable program codes may also control the number of bits for the single hash logic output so global load of the plurality of tables is configurable.
The invention will now be described more fully hereinafter with reference to the accompanying drawings, in which preferred embodiments of the invention are shown. Like numbers refer to like elements throughout.
With reference now to
In one embodiment, the single hash logic 22 is based upon the Cuckoo algorithm. In another embodiment, the single hash logic 22 comprises a configurable cyclic redundancy check (CRC) polynomial. In another embodiment, the hash apparatus 20 provides parallel table lookup of the plurality of tables based upon the avalanche effect.
In one embodiment, the avalanche effect is based upon orthogonal hash functions for each of the plurality of tables, and the single hash logic 22 implements each orthogonal hash function. In another embodiment, each bit output of the single hash logic 22 comprises funneled results of the key bits.
In one embodiment, the funneled results are generated by an XOR function. In another embodiment, the plurality of tables are configurable. In another embodiment, the global load of the plurality of tables are configurable by controlling the number of bits for the single hash logic 22 output.
Another aspect of the invention is a method for improving a computer system, which is now described with reference to flowchart 24 of
In another method embodiment, which is now described with reference to flowchart 36 of
In another method embodiment, which is now described with reference to flowchart 44 of
In another method embodiment, which is now described with reference to flowchart 52 of
In another method embodiment, which is now described with reference to flowchart 60 of
In another method embodiment, which is now described with reference to flowchart 68 of
Another aspect of the invention is computer readable program codes coupled to tangible media to improve a computer system. The computer readable program codes may be configured to cause the program to reduce insertion times and/or hashing collisions when new components 16 are introduced to a controller 12 that includes a computer processor 14. The computer readable program codes may also reduce hashing collisions by using a plurality of tables and a plurality of keys per bucket through a collision avoidance apparatus 18. The computer readable program codes may further map the plurality of keys to the plurality of tables with a hash apparatus 20 in communication with the controller 12 where the hash apparatus uses a single hash logic 22 to provide an avalanche effect when one key is changed which results in nearly half of bits changing in the plurality of tables.
The computer readable program codes may additionally provide parallel table lookup of the plurality of tables via the hash apparatus 22 based upon the avalanche effect. The computer readable program codes may also base the avalanche effect upon orthogonal hash functions for each of the plurality of tables, and the single hash logic 22 implements each orthogonal hash function.
The computer readable program codes may further funnel results of the key bits for each bit output of the single hash logic 22. The computer readable program codes may additionally make the plurality of tables configurable. The computer readable program codes may also control the number of bits for the single hash logic 22 output so global load of the plurality of tables is configurable.
In view of the foregoing, the system 10 provides an improved computer system. For example, using conventional hashing with chaining such as the “Cuckoo Algorithm” can do a search in a few bonded number of table lookups when used in a 100 Gbps Ethernet application, but its drawback is a long insertion time when a lot of hashing collisions are present in a very large table. In contrast, system 10 enables insertion in the same time range as a search, which is important when new table insertion is done on the fly by packet learning in a plug and play network.
System 10 uses efficient hashing and/or algorithm selection for collision avoidance and reduced insertion time. System 10 provides packet look-up for very large tables and high throughput.
In one embodiment, system 10 provides hardware implementation of a single hashing method capable of using any type of CRC polynomial division, for any number of logic tables, thereby reducing cache collisions. System 10 hardware implementation of a packet lookup function supports a very large number of large keys (hundreds of millions) and very high throughput (100 Gbps Ethernet). In another embodiment, system 10 is based on a single logic 22 in which system 10 can configure any type of CRC polynomials. In another embodiment, system 10 reduces collisions using orthogonality and high avalanche effect for a plurality of hashers thereby enabling the lookup of several tables in parallel.
In one embodiment, efficient orthogonal hashers are used for each different table. In another embodiment, to be efficient, a hasher should provide a high avalanche effect, e.g. around half of the bits of the hash result change when only one bit of the key is changed.
This can be achieved by division or multiplication of the key, but with the very limited clock cycles for 100 Gbps aggregate throughput, such hardware implementation of multiplication is difficult. XOR folding can be implemented in hardware but they do not provide a good avalanche result. CRC can work because it provides a good avalanche effect and can be implemented in hardware with XOR funneling, but it does require a lot of logic XOR circuits, which leads to large logic circuitry for different tables.
To reduce the collisions, system 10 uses 4 different tables and 8 keys per bucket, in one embodiment. In another embodiment, a single programmable hasher 22 is used where each bit of the hasher result is the funnel result of all the bits of the key, which allows the implementation of any type of CRC polynomial. This requires a large number of XOR circuits, but the sharing of the single logic 22 for the 4 different tables makes it affordable.
In one embodiment, the flexible hardware implementation to optimize the solution with the lookup table includes having the number of hash tables configurable from 2 to 5 (basic configuration could be 4 actives+1 standby). In another embodiment, the 5 hashing functions are XOR matrix fully configurable. Even if the matrix allows different types of hash configuration, CRC hashing function is the preferred choice.
In one embodiment, the global table load will be configurable by providing configuration of the number of bits for the hash result (hardware can support up to 19 b hash value 4M entries per table). In another embodiment, “On the Fly” keys comparison help to optimize back to back DMA requests and responses.
With additional reference to
With additional reference to
In one embodiment, probably the most flexible implementation of system 10, all the key bits can participate to any hash bit. This implementation makes the hardware almost independent of refinement to the algorithms and the maximum number of tables because the “intelligence” is in the hash logic 22.
In one embodiment, sizing for a 64 b key, 19 b hash, and 5 hash functions provides:
A rough area estimate for 45 nm implementation include:
In one embodiment, system 10 provides a method for a table lookup. The method may include hardware means for executing a single hash function for mapping one of hundreds of millions of keys to a slot, e.g. bucket, in the table whereby the hashing function produces a high avalanche effect when only one bit of a key is changed. The method may also include means for minimizing hashing collisions by implementing four different tables and eight keys per bucket.
In one embodiment, system 10 provides hardware implementation of a packet lookup function supporting a very large number of large keys, e.g. hundreds of millions, and very high throughput, e.g. 100 Gbps Ethernet, based on a single logic 22 that can be configured to any type of CRC polynomials while reducing collisions with orthogonality and high avalanche effect for a plurality of hashers to enable the lookup of several tables in parallel.
In one embodiment, system 10 provides a method and system for executing a single hashing logic 22 to map a very large set of keys to a slot, e.g. bucket, with a high avalanche effect. In another embodiment, system 10 reduces hashing collision by sharing the single hashing logic 22 for four different tables and eight keys per slot.
It should be noted that in some alternative implementations, the functions noted in a flowchart block may occur out of the order noted in the figures. For instance, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved because the flow diagrams depicted herein are just examples. There may be many variations to these diagrams or the steps (or operations) described therein without departing from the spirit of the invention. For example, the steps may be performed concurrently and/or in a different order, or steps may be added, deleted, and/or modified. All of these variations are considered a part of the claimed invention.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms “a”, “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises” and/or “comprising,” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
The corresponding structures, materials, acts, and equivalents of all means or step plus function elements in the claims below are intended to include any structure, material, or act for performing the function in combination with other claimed elements as specifically claimed. The description of the present invention has been presented for purposes of illustration and description, but is not intended to be exhaustive or limited to the invention in the form disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the invention. The embodiment was chosen and described in order to best explain the principles of the invention and the practical application, and to enable others of ordinary skill in the art to understand the invention for various embodiments with various modifications as are suited to the particular use contemplated.
While the preferred embodiment to the invention has been described, it will be understood that those skilled in the art, both now and in the future, may make various improvements and enhancements which fall within the scope of the claims which follow. These claims should be construed to maintain the proper protection for the invention first described.
