By William Fellows In the next century IT services will be defined by a myriad of devices making use of data provided by information utilities, according to Hewlett-Packard Co chief scientist Joel Birnbaum. Think of them like today’s utility companies says Birnbaum. An electricity supplier makes its resource available for use by every conceivable type […]
By William Fellows
In the next century IT services will be defined by a myriad of devices making use of data provided by information utilities, according to Hewlett-Packard Co chief scientist Joel Birnbaum. Think of them like today’s utility companies says Birnbaum. An electricity supplier makes its resource available for use by every conceivable type of equipment that has a power plug, he says.
HP is pitching its E-Speak distributed hypervisor technology as both the power plug through which any device can access these information services and as the switch that brokers between different collaborative services to send appropriate data to a device. E-Speak, which is written in 80,000 lines of Java and C++ code, presents information and data through APIs to devices which are pre-installed with a client telling an E-Speak server which services it can receive or is offering.
A hypervisor is built from wrappers placed around system calls for selected system components and can perform various security functions including fine-grained access control and auditing of events. E-Speak sits between the operating system and applications. It’s not platform specific nor does it require kernel changes. Hypervisors have traditionally been used to implement virtual machines and can intercept, buffer and distribute signals from outside the system. If a service is replicated on several nodes of a distributed system, a signal received by one node will be delivered to all of the replicas, and multiple signals will be delivered in the same sequence on each node.
Birnbaum hopes that E-Speak will become the lingua franca, or common language of electronic services on the web, providing the mechanism to dynamically compose, deploy, mediate, manage and access these services. For example it may automatically broker between multiple airline reservation service providers to select the one that best fits a user’s profile. It will also, HP argues, drive the requirement for new devices in addition to enabling internet services to be accessed by cellphones, PDAs and other small devices. For example, says Birnbaum, a medical cardiac monitoring e-service will drive demand for a chest monitor patch that combines GPS, arrhythmia monitoring and a radio transmitter technology in a small lightweight package. E-Speak will be to e-services what Java is to the internet, Birnbaum says.
IDC believes in the next phase of the internet economy, a standard lexicon (APIs) and protocols for accessing and deploying services on the internet need to be added to content-oriented internet standards. These new standards must provide common ways of supporting the core functions required for successful operation of internet-based services, including description/virtualization of internet services. The web enables representations of data, information, and user/organization identities in many forms. These virtual components are key to deployment of services. Today, catalogs, icons, files, and (frequently) entire web pages represent actual products, databases, or organizations. The creation of virtual representations of web objects is proliferating rapidly, yet there is no standard means of doing so. E-Speak fits this bill, HP claims.
The E-Speak core and repository uses a meta language to describe and represent services which register their availability. It registers the attributes of both users and services. E-Speak is able to make all kinds of resources – Java objects, files, printers, legacy applications – available to any device which utilizes E-Speak APIs, by publishing their availability. The repository supports JDBC, the communications stack runs over TCP/IP, supports XML, X.500, SNMP, Corba, HTTP, WAP, SSL and LDAP, has program libraries for writing services in C++, Java and Perl (with Python planned) and EJB and COM are being considered as component environments.
Crucial to E-Speak’s security will be the ubiquity of PKI public key infrastructure, without which E-Speak will run into all sorts of problems admits Birnbaum. He sa
ys there are still many security issues with E-Speak that still have to be worked out.
Through support for these protocols and programming conventions, HP says that E-Speak provides virtual machine intermediation, name space management, service monitoring and management, secure interaction limits and controls, location negotiation, location independence, asynchronous message passing and has advertisement, distributed event and authentication. Transaction support, tunneling and monitoring services are still being developed.
Birnbaum has been in the open systems business too long to be religious about E-Speak. If something else becomes the standard then HP will use it, he says. But the company will push hard to make E-Speak the de facto standard for e-services. Having learned the hard way during the Unix wars and the rise and fall of the Open Software Foundation. HP will pursue the widespread adoption of E-Speak, but not through a standards body. He says that at the time he didn’t care if it wasn’t HP-UX that was adopted a standard Unix for the industry. If agreement had been reached between OSF, Unix International and other groups on use of common Unix APIs, NT wouldn’t be around today he believes. The emergence of the web was the real wake-up call he said. E-services will happen like the web. Users decided to use HTML and HTTP. Vendors would never have agreed on them, he says.
Birnbaum says three engineers started developing and demonstrating E-Speak four years ago but only recently has its practical application has become apparent with advent of electronic commerce. There are now more than 100 engineers on the project.
Having touted E-Speak around several companies he said that AT&T came over and was so enthralled with E-Speak it considered jettisoning its own one million line project along the same lines in favor of E-Speak. At one time it sent up to 100 engineers to HP’s Labs to work on E-Speak but the companies have yet to come to a commercial arrangement. The problem is that equipment makers such as Nortel and Alcatel are simply not interested in developing open systems solutions that will be required to make E-Speak ubiquitous.
Birnbaum believes it will take around four years for E-Speak or something similar to become pervasive. It was 46 years after the invention of electricity before 25% of US households had access to it; 35 years for the telephone; 16 years for PCs; 13 years for cell phones; and seven years for the web, he says.
New generations of information appliances will use biometric sensors to detect who is using any particular device and where they are. Passwords and cards could quickly become antique. As a result ownership of a particular device could become irrelevant except that consumers become very attached to their personal property, cars, cell phones, laptops and so on, Birnbaum observes.
Birnbaum says rather than compete with Sun’s Jini, RosettaNet, or Microsoft Plug and Play, E-Speak will work on top of them as well as HP’s own Chai embedded Java machine. There will be subsets of E-Speak, he explains, which will act as the glue connecting them.
The real competition will come from services over which Birnbaum claims HP has a lead of at least a year, including Microsoft’s BizTalk. IBM is also developing proxy and translation services that will reside on the network with the goal of enabling networks to host any type of client system and allow those clients to take advantage of any service on the network, regardless of what combination of hardware or software is hosted.
Birnbaum says the problem with Microsoft’s approach is that it will make BizTalk a component of the operating system from the word go; it will be proprietary in other words. As more and more services get sucked into operating systems that approach may make sense over time. But to seed and nurture a market, burying the services in an operating system simply doesn’t make sense Birnbaum says. He expects the industry will settle on a common solution or some interoperable superset by the usual smoke and mirrors methods. Let the horsetrading begin.