HMP software - you've served your apprenticeship; you're hired
Features
Since the first white papers and early builds of software grabbed by some eager innovators, we're now seeing a broader range of HMP products available in the market with better targeted features and stability as the classic software 'version 2s' hit the streets from leading vendors.
But first let's start by reviewing what media processing is about and go on to recap what HMP is and where it fits. A variety of media processing functions need to be taken care of as part of designing and integrating a communications platform, depending on the target application. They range from recording and playing back speech using an appropriate compression or encoding scheme to suit the telephony devices being used, through DTMF recognition to fax transmit and receive. The amount of mathematical processing power required to support each function is very much dependent upon its complexity.
Historically, the only commercially viable method of performing the high number of computations required to run the different algorithms was using digital signal processors or DSPs. These devices are still the mainstay of media processing today and have the unique ability to perform both multiply and accumulate functions on a single clock pulse - unlike general purpose processors. This efficiency enables high density designs to be achieved, without causing material impacts upon the host processor, which is usually running the main application. And this sound engineering logic stands good today.
No-one has missed that general purpose processors on our PCs and servers seem to creep up in power every year. And this point hasn't been lost on software engineers looking to work on newer versions of algorithms for media processing. While at any point in time we would expect a DSP to out-perform a general purpose processor executing like-for-like media processing functions, the more important question is, "Can the general purpose processor support enough channels of media processing required to do the job in the communications system?" This fundamental question can get lost sometimes and reminds me of a debate I heard comparing two sporty looking high-powered cars - each had some special attributes like road holding on corners, or raw straight-line speed, but when you boiled it down to a question of can it get me from A to B safely in reasonable time? The simple answer was yes, of course!
Few would deny the market trend is to increasingly see technology change to IP instead of TDM. In this context it means that no specialist telephony hardware interface is required any more, and the traffic will arrive at the server via a low cost NIC. Now we're talking for the first time about having a media processing solution with no specialist hardware! This holds appeal for organisations with strong software application abilities and have found the prospect - or even the experience - of integrating traditional telecoms somewhat daunting. So maybe we're looking at a slightly different type of developer who will eagerly embrace HMP?
So what is the profile of applications being developed with HMP technology? As with any new technology there was a level of caution from many, who were happy to try the basic functions at modest channel counts. This has led to a number of either embedded IVR functions on the same server as the communication application, or a stand-alone server configuration to complement an all new IP system. These key system/PBX/contact centre solutions have typically been in an enterprise environment where maximum local control is achievable and channel counts are frequently below 40 users. This arrangement has allowed a proper exercising of the functionality to everyone's satisfaction and has allowed feedback to be gained by the HMP vendors such that key improvements can be made.
We mustn't forget that the diffusion of innovation follows a well documented and reproducible path, so there is a high level of certainty that HMP technologies have staked a firm claim and will see high growth. Initially this is likely to be at the expense of DSP boards with analogue interfaces, which are synonymous with low density installations. The T1/E1 density board space will be eroded from the low end in due course. But none of this will happen without the main market players taking appropriate steps to listen and act upon the customer's feedback. This in itself can be a challenge for some vendors who take a simplistic view that if a product's average sell price is below $1000, then it will be sold via channel partners. The simple alternative is to make the product available by the vendor's website - but is technical support going to be provided in the price of the web-based product sale? No-one is claiming that the technology is an order of magnitude easier to understand than a hardware-based DSP circuit card. Whichever route to market is selected, the vendor has to ensure feedback is managed and support provided appropriately.
Another area worthy of note is the licensing of the HMP product. Given the ease of software distribution, and the inevitable concerns about possible illegal copying, there needs to have been an appropriate amount of consideration given to the balance needed to protect the vendor and the flexibility needed by the developer and systems integrator. There is an understandable tension in trying to get this right and given that the majority of HMP vendors have a hardware dominated background, we are witnessing steady improvements across the market. The leading offerings have either an OEM licensing arrangement to enable the developer to embed an appropriate number of channels - at their control - or a software licence key that is generated to match the platform upon which the application is run.
To assist customers who fit the 'majority' profile - and that is the stage of the market we are now in - most HMP vendors are offering a free, time limited version of their product to remove barriers to adoption. With typically 4-6 weeks time to carry out product evaluations, some vendors even offer free technical or pre-sales support, there is clearly a push to capture the mind share of the development community.
Given this is IP-based technology; we anticipate an increasingly wide range of devices that HMP will need to be connected to, possibly with different media and compression schemes. In addition to the standard G.711 audio codec, we're seeing implementations of an interesting range including; G.729, G.726, G.723, Speex and iLBC, the Internet low bit-rate codec. The iLBC codec was specifically developed to accommodate for the 'bursty' transmission characteristics of the Internet, and is used in the widely deployed Skype VoIP communications software most of us are familiar with. It is worth spotting the mixture of audio codecs that cover traditional wired and wireless TDM networks, ensuring a wider applicability for HMP products than at initial launch.
And yes, there still seems to be a solid market need for facsimile - particularly for sectors like financial services, so support for the T.38 standard is common and this ensures the developer can 'tick the box' when asked.
Having touched on audio and fax, now is the right time to mention video too. Those of us watching the newer features coming through on the latest host processors will know that every release of silicon brings us 'new' and 'spectacular' graphic handling and sets the expectation that we should be seeing at least a few channels of video from the HMP products too. While still early days, we can now see the first products becoming available that can connect with 'like-to-like' video devices; that is to say no transcoding is being performed, which would be quite an intensive processing activity.
What of the future? We can reasonably expect a continuous rise in the number of channels and features that may be supported on a single installation, thus leading the system architect to consider HMP as a serious alternative when making technology selections for more solutions.
As familiarity with the technology increases, both from the HMP vendors and the developers' perspective, we would expect more insights into ways of taking advantage of a software-only architecture that lead to experience-based enhancements.
One such area is likely to be in methods of providing distribution amongst different servers to provide both additional resilience and a means of scaling a solution. These areas of development should consolidate and enhance the choice for designers to consider either DSP or HMP technologies in solutions.
All in all you could argue that HMP is just completing its apprenticeship and is ready for a prime time audience.
Bookmark