Electrical computers and digital processing systems: multicomput – Computer-to-computer data routing – Centralized controlling
Reexamination Certificate
2000-04-07
2004-06-08
Luu, Le Hien (Department: 2758)
Electrical computers and digital processing systems: multicomput
Computer-to-computer data routing
Centralized controlling
C709S238000, C709S219000
Reexamination Certificate
active
06748447
ABSTRACT:
FIELD OF THE INVENTION
This invention is directed towards data communication, and more particularly towards reliable and efficient distribution of data to large numbers of network locations.
BACKGROUND OF THE INVENTION
Digital content creators are users who utilize workstations or other computers to create or digitize information in preparation for publication as “content.” When such content is to be shared with or published to a number of other computer users using a wide area network (WAN), such as the World Wide Web (“the Web”), reliability, latency, security, and efficiency become major issues. Reliability refers to the ability to ensure that the data was received without debilitating errors. Latency, the measure of how much time it takes to deliver data, suffers when finite resources become overloaded, whether in the respective processors, intermediate storage or a communications link. Inefficiency may arise because multiple copies of data have to be retransmitted between the same source(s) and destination(s) due to lost or garbled messages. As the number of recipient sites grows, issues of latency and efficiency complicate the architecture.
Inefficient communication protocols for reliable data exchange amplify problems in real-time systems where latency directly determines user satisfaction.
Historically, manual or customized operations were the only solutions available for distributing new or modified content, as networks expanded and data-distribution needs changed.
However, such solutions have the disadvantage of not being flexible enough to handle real-time load balancing. Temporary outages of system components can also cause havoc in a statically defined distribution method. Similarly, manual or customized actions become increasingly labor-intensive as data files proliferate and the number of servers increases exponentially, as seen in the recent growth of the Internet. In particular, the operation of the “Web” requires massive data management and distribution. Many users expect instantaneous access, worldwide, to the fastest source of the best data available at any given moment. This puts a heavy burden on service providers for better information control and infrastructure management.
One well known solution to reduce access latency by large numbers of users is to distribute content to file servers at numerous remote sites, and then direct user access requests to those servers. Multiple copies of content must then be tracked and synchronized in order to provide uniformity and consistency of data among all users. Many network content publishers obtain network file server services from a variety of geographically dispersed service providers. Manual coordination with each service provider for content distribution increases complexity and creates more room for error and delay.
To manage the problem of rapid content distribution from a master copy, several companies have experimented with or proposed semi-automated systems for streamlining the distribution process. These solutions are typically targeted at one of three critical points: “content management;” reliable and efficient distribution across WANs; or the local replication and synchronization across multiple servers within a Local Area Network (LAN). Content management refers to the methods of ensuring that only the necessary data is sent, that the remote copies are synchronized, and that file transmission is properly compressed and encrypted, as necessary.
One example of a content management system is the Content Delivery Suite (CDS) product distributed by Inktomi Corporation of Foster City, Calif., as described at www/inktomi.com/products
etwork/traffic/tech/cdswhitepaper. According to the available documentation, CDS management components determine when data content changes within file systems on a “staging server,” and then send updated files to “CDS Agents” on distributed web-servers. Once the updated files are received at the web servers, the CDS triggers all web servers to take the updated files “live” simultaneously. This particular solution suffers from numerous disadvantages. Sending entire files for an update is relatively inefficient, when only a small amount of data may have actually changed out of millions of bytes in the file. File transmission to each remote server originates from a single, central point, and all remote servers must wait for the others accessing the same central source to receive and acknowledge the correct data before the new content goes “live.” The referenced implementation lacks the ability to intelligently schedule distribution or replication of pertinent content to different parts of the network according to the user's needs.
Another example of a system for managing content distribution is the global/SITE product of F5 Networks, Inc., of Seattle, Wash., as described at http://www.f5.com/globalsite/index.html. The available documentation indicates that global/SITE is an additional computer appliance that is added to a LAN and a central site. The specialized hardware and software at the central site automatically replicates and transfers only those files that have changed (i.e., new, updated, or deleted). Changes to updated files include only the changed portions, thus reducing the wasted transmission load. However, disadvantageously, the addition of separate hardware and software at each site inherently reduces reliability, since there are more components subject to maintenance and potential failure. In fact, the global/SITE system becomes a single point of failure which could cripple an entire site if the unit is rendered inoperable, whether accidentally or maliciously. Installation, configuration and maintenance of these additional units will also require on-site support and customized spare parts.
One approach to schedule management is proposed in U.S. Pat. No. 5,920,701 (“the '701 patent”), issued Jul. 6, 1999. The '701 patent teaches a system in which data transfer requests and schedules from a content source are prioritized by a network resource scheduler. Based upon the available bandwidth and the content priority, a transmission time and data rate is given to the content source to initiate transmission. The scheduler system requires input that includes information about the network bandwidth, or at least the available bandwidth in the necessary content path. This has the disadvantage of requiring additional complexity for determination of network bandwidth at any given moment. It also requires a method for predicting bandwidth that will be available at some transmission time in the future. Furthermore, a content distributor is required to provide a “requested delivery time deadline,” which complicates content management by requiring each content distribution requester to negotiate reasonable transmission times for each piece of content. This approach is focused entirely on bandwidth allocation, and fails to address issues of network dynamics, such as regroupings of the target servers for load-balancing. Whatever efficiency may have been derived from the '701 is substantially completely lost when the entire content must be retransmitted to an additional server, making a huge waste of bandwidth for every node in the multicast path which already received the file.
Each of these alleged management and distribution solutions relies upon file replication and transmission techniques that remain closely tied to one-on-one file transfers to each individual server. The problem grows geometrically as the number of servers increases and multiple copies of selected files are required at each remote web site.
The ubiquitous Internet Protocol (IP) breaks messages into packets and transmits each one to a router computer that forwards each packet toward the destination address in the packet, according to the router's present knowledge of the network. Of course, if two communicating stations are directly connected to the same network (e.g., a LAN or a packet-switching network), no router is necessary and the two stations can communicate directly using IP or a
Basani Vijay R.
Bestavros Azer
Crovella Mark E.
Karge Leroy R.
LaRosa Domenic J.
Brown Rudnick Berlack Israeis, LLP
Luu Le Hien
Michaelis Brian L.
Network Appliance Inc.
LandOfFree
Method and apparatus for scalable distribution of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for scalable distribution of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for scalable distribution of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3330893