Special Features:
GRID DELIVERY: A NEW TECHNIQUE FOR
CONTENT IN THE ENTERPRISE
By Daniel Sapir, Bandwiz, Inc.
1. The Need
Efficient and rapid delivery of information of all sorts to employees,
distributors, and corporate partners that can be anywhere in the world is
essential to meeting business goals. Employees, partners and distributors need
to be trained on products, especially at the time of new product launches.
Support personnel need to get detailed training on products if they are
expected to effectively support these products. New regulations need to be
distributed to make sure employees know about them, e.g. sexual harassment
policies and OSHA safety regulations. Organizational goals need to be clearly
distributed to all employees to make sure the organization is moving together
as a team to meet corporate goals.
No organization can be successful with factions moving in different
directions. One Texas CEO, upon being asked what his toughest challenge was,
replied "keeping the herd moving approximately west."
Customers also need to be courted to gain loyalty and repeat sales. The way
to
do that today is largely by electronic delivery of news, market programs, and
other material as well as with personal contact.
To be effective, the information content that is delivered is getting
richer,
is updated more frequently, and is sent to ever larger numbers of recipients.
However, rich content means larger sizes. Our computers can usually cope as
memory available on PCs is growing rapidly and PCs typically have much
underutilized memory.
However, enterprise networks are usually not up to the task. If any two of
the
characteristics above are true (i.e. large content plus frequent delivery or
frequent delivery to large numbers of recipients, or large content delivery to
large numbers of recipients even if infrequent), there will be a content
delivery problem in that organization.
At busy times of the day, the corporate network is often congested leading
to
long delays. Corporate IT personnel put in place policies that forbid
downloading large content in attempts to prevent congestion in the corporate
network. Upgrades to the network are generally out of the question; budgets
are flat to down and IT staff is stretched thin.
2. Solutions
There are three basic approaches that can be taken to improve this
situation:
1 Upgrade the links with more bandwidth – this is generally too expensive
and
is virtually never done.
2. Adapt an Internet Content Delivery Network (CDN) solution to the
enterprise
by adding proxy caches to store content at the edges closer to consumers. This
has merit as it trades off bandwidth for storage; a good tradeoff given
storage is about 100X cheaper than bandwidth. But, this approach was designed
for the Internet environment, which is very different than the corporate
Intranet.
3. Use a "Grid Delivery"* approach, a new technique specifically targeted
at
providing solutions for the enterprise.
The CDN and Grid Delivery approaches are discussed in more detail.
2.1 The Internet CDN Approach (using proxy caches)
Before discussing the use of proxy caches in the enterprise, it is worth
examining the differences between the public Internet and the corporate
Intranet. Proxy caches, the primary component used in Internet Content
Delivery Networks, were designed for the Internet environment to speed up Web
experiences.
The table below lists some of the primary differences between the Internet
and
corporate intranets:
- Corporate Intranet
- Low bandwidth WAN links
- High bandwidth local networks
- One or a small number of domains
- Centralized control and management
- Security and authentication is important
- Public Intranet
- High bandwidth backbones
- Low bandwidth last mile links
- Thousands of domains
- Distributed control and management
- Security and authentication only needed in some applications, e.g.
transactions
Content Delivery Networks basically speed up Web traffic by trading off
storage for bandwidth by storing popular Web content "closer" to the user in a
proxy cache. Closer, in Internet terms, means within the same domain since the
largest sources of congestion in the Internet aside from last mile
constrictions are the junction points between domains. It is not unusual for
traffic to traverse many domains to reach the recipient.
A mechanism is put into place (usually using the WCCP protocol in the local
router) that redirects http and ftp requests from browser users to the local
proxy cache. If the local cache has the content requested, it delivers it to
the requestor. If it does not have it, it fetches it either from another cache
further up in a hierarchy of caches or from the origin. It delivers it to the
requestor and stores it locally so that it is immediately available to the
next requestor.
Although cache capabilities have been upgraded to support streaming media
and
to be able to pre-populate them before a request is made for the content, the
fact remains that they operate basically on the pull paradigm, i.e. content is
stored at the edge based on a request from a consumer located at that edge.
Additionally, they were designed for Web traffic and other content is dealt
with as an afterthought if at all. Finally, there is no centralized
management; they operate autonomously, a requirement for operation in the
Internet.
However, proxy caches can be positioned within an enterprise at remote
offices
to speed up content delivery to users of the corporate portal at that site.
However, this is a limited solution for the enterprise that is also expensive,
as each site needs to add a proxy cache, a new piece of hardware. A summary of
these limitations is shown below:
1. Proxy cache hardware costs about $2000-$5000 each, prohibitive when
there
are thousands of remote sites.
2. Addition of remote hardware adds support costs in the field, very
undesirable when enterprise support personnel are typically overworked and
centrally located.
3. Caches by nature store content in the clear open to all. Enterprises
often
want to secure the content.
4. Caches are autonomous devices, whereas the enterprise desires
centralized
management and control of remote devices.
5. Enterprises want a "push" solution, where content can be sent to
recipients
without requiring a request for content from the consumer. They can do this to
some extent with e-mail attachments but content is getting too large and
another way is desired. Caches do not address this need at all.
6. Enterprises often require usage reporting, i.e. reports of content
actually
received and opened. This is also not addressed by caches.
There is a fundamental difference between content delivery in the Internet
and
in the corporate Intranet. Consumers go to the Internet to find content of
interest making it inherently a "pull" model, whereas in the corporate
intranet, the supervisor of employees or the corporation may wish to "push"
content to their employees for use in their work.
2.2 Grid Delivery
Enter "Grid Delivery", a new system first coined by the Gartner Group that
is
designed specifically to address the needs of the Enterprise. The term "Grid
Delivery" is used since many desktop and possibly even laptop computers
cooperate to perform the delivery in similar fashion as many computers can
cooperate to perform "grid computing" tasks.
The intent of a grid delivery system is to utilize network and computing
resources already present in the enterprise more efficiently. Enterprise
networks only become congested at busy hours leaving other times with
capacity. Additionally, there are computing and storage resources in the
desktop and laptop computers throughout the organization that can be tapped.
The key trick is to be able smooth out traffic on the network and tap into the
underutilized computing power and storage available in desktop and laptop
computers.
The key technology is called "mediated peer-peer". Desktop and laptop
computers at remote sites are loaded with a piece of software that operates in
the background and is used to provide four key functions:
1. Application layer routing forwarders. Efficient one-many delivery
(application layer multicast) can be achieved without having to touch
networking devices in the network, while at the same time gaining the
efficiencies promised by network layer multicast routing.
2. Delayed time forwarding, to optimally utilize low network usage times to
deliver content over the corporate intranet.
3. Caching of content on multiple desktop/laptop computers locally, to
provide
rapid delivery of content to local consumers of the content.
4. Reporting of content delivery and usage.
Additionally, since content may reside on desktop/laptop computers and may
be
confidential, security is an inherent requirement of the system. By default,
all content is encrypted during delivery and when under local storage.
Grid Delivery is also inherently cost effective and support friendly since
no
new hardware is deployed in the field, only small software agents that reside
on the organizations’ desktop and laptop computers. This software is all
managed centrally.
A basic system would work with one remote site. Although there are many
client
computers at the remote site, content is only delivered once to that site from
the central site. Within that site, application layer forwarders distribute
the content to the other computers at the site. In some cases, all the
computers may receive the content; in others, only a few and if a computer
without the content wants it, it retrieves if from one or more of the local
clients. Sometimes, there are downstream sites; in this case, the application
layer forwarders send it off to the satellite sites, again with only one
transmission per wide area link.
The remote peers also act as a distributed storage cache for the site,
providing local (read rapid) access for content of interest to consumers at
that site. Also, since content is often delivered to the actual desktop,
reporting of delivery and usage by the consumer can be accomplished, important
for enterprises.
A grid delivery system also has the flexibility to interface to
applications
already operating in the enterprise, removing any need for training by users
of the system.
Some common applications are the following:
1. Pre-loading of content posted on the corporate portal (internal web
site).
In this application, content that management wants to deliver to its
employees, partners, or customers is pre-downloaded to the site(s). When the
consumer goes to the corporate portal to get it after notification of its
availability, he is re-directed locally and fetches it locally providing rapid
response.
2. Offloading of e-mail attachments that cannot be handled by the corporate
e-
mail system. Attachments not able to be carried by e-mail are stripped from
the e-mail system and delivered by the grid delivery system; the result is the
user can use a system for push delivery that he is very familiar with.
3. Secure delivery of sensitive content. By default, all content delivered
is
encrypted providing secure delivery of sensitive information.
4. Local caching of Web content without need for additional hardware.
Desktop
and laptop computers have a browser cache in each computer; by communicating
with other local computers to discover content residing on their local browser
cache, a virtual cache server for the site is created without the need for
additional hardware.
5. Simulated live multimedia streaming media to the desktop. Streaming
media
sourced at corporate headquarters is usually not feasible due to the
constriction of the wide area network links. With simulated live, the media is
delivered as large files at times the network can support it and then is
streamed locally over the high bandwidth local network at the multiple sites.
This can be scheduled for a time to start the streaming, hence the term
"simulated live".
3. Conclusions
"Grid Delivery" is a new type of system for content delivery focused on the
needs of the enterprise. It is a flexible system that utilizes the network and
computing resources already present in the enterprise much more efficiently,
extremely important in this age of tight IT budgets and staffing. It solves
all of the content delivery needs of the enterprise and is centrally managed
and controlled, which is important for enterprise IT staffs. Finally, it can
be used in both intranets and extranets, so that content can be delivered to
partners and customers as well as to internal employees.
Expect Grid Delivery systems to enter the mainstream in the near future.
Its
benefits to the enterprise are too high for it to be ignored.
About the Author:
Daniel Sapir is VP of business development at Bandwiz (www.bandwiz.com), a knowledge logistics
software company focusing on the distribution and management of content
throughout extended enterprise environments. He was formerly a vice president
at Ascential Software, a provider of enterprise data integration solutions.
Previously, he founded iCentric Solutions, Inc., an eBusiness integrator.
Sapir established and served as vice president of the enterprise solutions
business unit of Data General Corporation. Sapir has also served as a director
of sales for McDonnell Douglas Information Systems and Allerion Corporation,
as well as product manager at Pertec Computer Corporation and engineering/QA
project leader for MIS Data Corporation. Reach him at DSapir@bandwiz.com.
|
