This application relates to telecommunications systems, and in particular to a dynamically configurable architecture for mixed data processing and an associated technique for dynamically allocating telecommunications resources based on demand.
The hardware platform on which communications infrastructure, such as base stations and switch gateways, have traditionally been built includes a mix of microprocessors and digital signal processors. Microprocessors typically serve as system host and data router, performing control-oriented tasks, and interface to multiple DSPs. The DSPs are required to perform signal processing tasks such as speech compression, echo cancellation, fax processing, modem data pump functionality, and error correction. Communications infrastructure equipment must process a number of channels including a variety of data types to support system usage. Each channel or data stream may have unique processing requirements depending on its data type. For example, V.42 bis modem processing is control-processing intensive and places greater burden on the microprocessor than speech compression, which is completely signal processing oriented. Therefore, the ratio of signal processing resources to microprocessor resources is different depending on the channel type.
To process such diverse data streams, one technique has been to dedicate hardware and software resources for each channel or data stream in a communications system. This technique works best where the type of processing required is known a priori and does not change substantially. A typical system using this technique would have dedicated resources for speech data and dedicated resources for other data types.
The shortcoming of the dedicated hardware and software technique is that traffic patterns are not static, therefore resources and cost are not optimized. The number of voice calls versus the number of data calls does not remain constant over the course of time, such as a day. A need exists for a communications infrastructure that is flexible, can handle both signal processing and control processing functions dynamically, under software control and can be dynamically reconfigured to a configuration that corresponds to the demand.
In accordance with the invention, a communications infrastructure that can be dynamically reconfigured over a wide range of application combinations from being dedicated to one application to any combination of the first application and a second application (or more). The communications infrastructure includes a plurality of processors coupled to a bus and a shared system memory. Each processor can execute control or signal processing functions efficiently, which allows for truly dynamic reconfiguration. Instructions to implement the first and second function are stored in system memory and available to each processor over the bus.
A communications infrastructure 10 that can be dynamically reconfigured over a wide range of application combinations from being dedicated to one application at a first end of a spectrum, to any combination of the first application and a second application, to being dedicated to a second application at the second end of the spectrum is shown in the block diagram of
Communications infrastructure 10 is capable of processing several calls simultaneously. For example, communications infrastructure 10 may be able to process 64 voice calls. As shown in the block diagram of
Two computation intensive applications provided by communications infrastructure 10 include speech coding and modem functions. The speech coding function is required when a voice call received from a cellular handset, for example, is coupled to the circuit switched network 42. Speech coding is carried out by any known speech coding technique. Instructions to carry out the various speech coding techniques are stored in system memory 22. Modem functions may be carried out by any of the modem standards such as V.90 together with any fallback standard.
Rather than having one or more of the plurality 12 of processors dedicated to providing the speech coding function and another one of the plurality 12 of processors combined with an associated microprocessor host dedicated to providing the modem function, communications infrastructure 10 is dynamically configured to allocate sufficient processor resources to the speech coding and modem functions to meet the current level of demand for those functions over the channels processed by communications infrastructure 10, and to route the calls to the appropriate network dynamically. Instructions executed by processors 14, 16 and 18 are configured for speech processing, modem control and data pump functions, or a combination of these two applications on each processor 14, 16 and 18. In a preferred embodiment, processors 14, 16 and 18 are digital signal processors (DSPs), such as 16-bit SC 140 DSP cores each with a dual multiply-accumulate capability, although the invention is not limited thereto. Also preferably, processors 14, 16 and 18 are of an architecture that permits the data and instructions to be stored in a single memory space, although the invention is not limited thereto. Processors 14, 16 and 18 may be homogeneous, that is of the same architecture, or heterogeneous, that is of a different architecture.
Communications infrastructure 10 provides a versatile single hardware platform that can configure on demand to the type of processing required based on the demand for control or signal processing applications as the profile of usage changes. Communications infrastructure 10 achieves the dynamic configuration by storing instructions to perform each of the control and signal processing functions in system memory 22 and transferring instructions over bus 20 to instruction cache memory in one of processors 14, 16 and 18 the instructions to perform a function as increased demand for that function necessitates.
Bus 20 in the illustrated embodiment is a 128-bit wide split-transaction bus, although the invention is not limited thereto. A split-transaction bus exhibits the property that devices coupled to the bus places a request for information on the bus and releases the bus for use by other devices. At a later time, when the requested information is available, the requesting device is notified to receive the previously requested information over/from the bus.
One of the SC 140 cores, also referred to as processors 14, 16 or 18, can optionally execute control code only to provide the control aspect of the modem function for itself, and other DSP cores should they provide modem functions. Each processor core is programmed to know its capacity to handle control and signal processing functions and to know the portion of that capacity that is not utilized. When one of the functions of communications infrastructure 10 or an increased level of a function is required, a hierarchical order of assignment determines which DSP core is designated to undertake the additional processing. The designated DSP core ascertains whether the capacity to perform the additional processing is available. If not the next DSP core in the hierarchical order is designated to undertake the additional processing until a DSP core is identified with processing capacity to undertake the additional processing or the capacity of communications infrastructure 10 is determined to be full. Once a DSP core is identified that can undertake the additional processing, that DSP core requests, from specific memory locations in system memory 22, instructions to perform the function. The instructions are transferred over bus 20 from system memory 22 to an instruction cache in the designated DSP core. Data is buffered until the architecture of communications infrastructure 10 is reconfigured and the instructions to perform the additional processing are transferred to the designated DSP core. The DSP core taking on the additional processing task interleaves the additional processing into its existing workload. The process is repeated to accommodate additional processing tasks up to the capacity limit of communications infrastructure 10. Of course, tasks may be completed thereby reducing the load on one or more DSP cores and making available the unused capacity for subsequent additional processing tasks.
Number | Name | Date | Kind |
---|---|---|---|
4539637 | DeBruler | Sep 1985 | A |
4731817 | Jonsson et al. | Mar 1988 | A |
4823256 | Bishop et al. | Apr 1989 | A |
5222229 | Fukuda et al. | Jun 1993 | A |
5276828 | Dion | Jan 1994 | A |
5367678 | Lee et al. | Nov 1994 | A |
5515511 | Nguyen et al. | May 1996 | A |
5659701 | Amit et al. | Aug 1997 | A |
5867704 | Tanaka et al. | Feb 1999 | A |
5951683 | Yuuki et al. | Sep 1999 | A |
5999990 | Sharrit et al. | Dec 1999 | A |
6061008 | Abbey | May 2000 | A |
6067615 | Upton | May 2000 | A |
6092174 | Roussakov | Jul 2000 | A |
6098123 | Olnowich | Aug 2000 | A |
6104721 | Hsu | Aug 2000 | A |
6108414 | Ahrndt et al. | Aug 2000 | A |
6295455 | Fischer et al. | Sep 2001 | B1 |
6330603 | Seki et al. | Dec 2001 | B1 |
6363421 | Barker et al. | Mar 2002 | B2 |
6381321 | Brown et al. | Apr 2002 | B1 |
6442682 | Pothapragada et al. | Aug 2002 | B1 |
6459901 | Chawla et al. | Oct 2002 | B1 |
6493776 | Courtright et al. | Dec 2002 | B1 |
6556548 | Kirkby et al. | Apr 2003 | B1 |
6578076 | Putzolu | Jun 2003 | B1 |
6631447 | Morioka et al. | Oct 2003 | B1 |
6690951 | Cuffaro et al. | Feb 2004 | B1 |
7085858 | Fox et al. | Aug 2006 | B1 |
20010003831 | Boland | Jun 2001 | A1 |
20020095565 | Nemirovsky et al. | Jul 2002 | A1 |
20030033411 | Kavoori et al. | Feb 2003 | A1 |
20060039317 | Subramanian et al. | Feb 2006 | A1 |
Number | Date | Country |
---|---|---|
WO 9949623 | Sep 1999 | EP |