RSS
  Push Plumbing |
|
By Barry Nance
Getting the right data to the right people at the right time is the basic, fundamental purpose of a network. Wrong data? Wrong people? Wrong time? Your network has failed in some way. One way to resolve the problem is to give an administrator responsibility for the integrity of the network's information and equip that administrator with server-based tools for distributing that information. Wresting the responsibility from the clients and giving it to the server is the essence of push.
When a server pushes and a client pulls, theory says, push middleware--software that links the push server to the client--will pipe data through the network consistently, reliably, securely and economically. In practice, however, some products implement push better than others.
This test bed for push consisted of a 10BASE-T Ethernet network with three Pentium II 300-MHz machines running Microsoft Windows NT Server 4.0 and 25 Windows95, Apple Macintosh and IBM OS/2 clients. Three of the clients--one for each of the client operating systems--were laptops we could undock, use offline or connect remotely with slow dial-up links, then reconnect to the network.
 We chose products that provide the plumbing within a network to implement data distribution via push, not services (such as PointCast) that ride on top of push technology. We tested Marimba's Castanet, StarBurst Communications Corp.'s StarBurst Multicast, TIBCO Software's TIB/Rendezvous, Wayfarer Communications' INCISA and XcelleNet's RemoteWare Express SoftWare Manager. Another vendor, Modulus Technologies, declined to let us evaluate its InterAgent Toolkit product, saying it did not want the publicity. During our tests, we used Chariot, Ganymede Software's simulation and performance testing tool, to help us gauge the impact of injecting push middleware into an already-busy network. Chariot helped us conclude push middleware saves little or no bandwidth. However, we also found scheduled distribution of information can off-load network traffic to less busy times of the day. Our test criteria were simple yet comprehensive: We looked at the different kinds of content we could push, such as programs, data files, Web pages, corporate missives, team collaborations and company benefit reports targeted at specific groups of employe es; how configurable each product was (for example, whether we could throttle the product to emit data at a slower, less network-intrusive rate); what security features each product offered (keyed encryption or digital certificates); and, with the help of a Dolch Computer Systems' PAC 63C portable running Network Associates' Sniffer software, how much bandwidth each product used (we measured network utilization during each period of data distribution).
|
|
|
|
Complete survey results
The Push middleware features chart
, in Acrobat format.
The Current State Of IP Multicast The Nuts and Bolts of Business-to-Business E-Commerce By Brian Walsh Web Middleware Glue Binds Web Apps A Grand Opening For Virtual Storefronts With Middleware Pushing Past The Hype: Delivering To The Desktop The Ups And Downs Of Analyzing Middleware |
To view the Report card.
|
|
 |
|
|
