Value-added services at Internet exchanges

[Andreas Schachtner <]

Hello,

at the last meeting on European Internet Exchacnges at the RIPE meeting in 
September last year, I was appointed for writing a paper regarding value-added 
services at exchange points.

I turned out to be more a collection of possibles VAS as a detailed discussion 
how to install and operate VAS, but even (or in particular) this might set the 
basis for discussions how to proceed.

It is attached as plain ASCII.

Regards,
	Andreas Schachtner

Value-added services at Internet Exchanges Points
Andreas Schachtner, DE-CIX/IntraNet GmbH

Abstract

Internet exchanges (IXes) flourish all over the world. In particular in Europe,
with its history of national telecom markets, there are IXes in almost every 
country.
In this paper, possible value-added services of IXes are listed, sorted in 
certain classes and discussed with regard to their impact on ISP operations.


What is an IX

An Internet exchange (IX) is a central piece in intra-ISP infrastructure.

An IX provides a place for neutral interchange of data amongst connected ISPs.

IXes have different models for their organisation

- run by a neutral body
- run by a neutral self organised body
- run as a service among others
- run by a commercial service of an ISP

Genuine services at IXes

ICXes offer some genuine services coming from the nature of an IX. Those 
genuine services are

- data interchange over a common (LAN) medium
- housing of equipment to perform the data interchange
- basic computing centre environment

On top of those genuine services, a variety of value-added services can be 
installed.

VAS classification

In this paper, possile value-added services are sorted into several classes:

Technical and organisatoral value-added services. These services provide 
improvement in usability of IXes.

ISP tuning. These value-added services allows ISPs to tune their infrastructure 
by means of synergies in deployment of central services.

Quality assurance services. These services allows for quality control and 
management by monitoring ISP performance. These services are primarily oriented
towards end users e.g. customers of ISPs.

Enabling services. These services are only meaningful or reasonable by 
placement at central places, thus enabling the deployment of these services.

Other servies, that don't fit into above classes.

VAS listing

Dial-In - T
Dial-in facilities for out-of-band access to IX equipment. This dial-in can be 
further refined into
- shared dial-in to the common interchange medium
- dediated dial-in lne to specific equipment

This service is of general interest to all ISPs. The relation of cost to 
usefullness might be doubted.

Remote control - T
Remote control of ISP equipment (remote power cycle, remote reset, ...).

This service is of general interest to all ISPs. The relation of cost to utilty might be questioned.

Route reflector - T
Operation of a central BGP4 route reflector. A route reflector simply reflects 
all received routes to connected peers.

This service brings down the number of pering sessions from o(n*n) to o(n). 
It constraints implemenmtation of routing policies, though.

Route server with Routing-Policy - T
Operation of a central BGP4 route server. Additional to a route reflector, 
a route server maintains policy information provided by connected peers. 
Routing policies may be submitted by several ways
- out-of-band configured
- generated from route repositories (RIPE DB)

As for the route reflector, the number of peering sessions is reduced. 
Flexibility of policy impmentation/enforcement is secured. Formulation and 
publishing of a routing poliy is required.

Redundancy of central infrastructure - T/Q

Provision of redundancy of central infrastructure, in particular
- location
- exchange medium (Ethernet switch, ...)
- route server, route reflector
- power supply
- ISP spare equipment

This is in general interest of all ISPs. Parts of the redundancy can only be 
achieved at high cost.

Monitoring of hardware configuration - Q
Monitoring of configured/measured status of visible central and ISP owned 
equipment. Notification of operator of malfunctioning equipment.

This service - despite of being in general interest - will be achived by most 
ISPs in their internal NOCs.

"Trust server" - E
Operation of a server for generation and management of tickets and certificates
for secure transactions.

A central data excange is a good place for ticket servers. A central 
implemetation might interfere with provider-based services.

News server - I
Operation of a central news server. This server will provide news server of 
connected ISP with a newsfeed.

Despite of being ideal for tuning, a central news feed opens issues of 
settlement between feeders and feedees.

MBONE-router - I
Co-location of a multicast router at an IX.

The same problems as for news will be faced here. The problem of an adaquate 
upstream is even more severe.

Name server - I
Operation of a master name server for certain zones, most prominent for 
servers for
- . (root)
- national TLD
- gTLDs

In principle, the same issues as top news apply. Provision of a central name 
server is far more accepted, though. The upstream problem is present as well, 
but to a lesser extend.

WWW cache - I
Operation of a web cache server.Cache fetches pages for connected ISPs. 
The cache can be engineered to fetch pages for connected ISPs only or provide 
upstream fetching as well.

In principle, the same issues as to news apply. Provision of a central cache 
server might be more accepted, in particular if the cache can be integated 
into a larger caching farm / caching hierarchy.

Content-Monitoring - E
Monitoring and supervision of Internet contents (e.g. News and WWW) according 
to legal and/or accepted principles.

Only the data interchange and selected upstream provision allows for cost 
efficient monitoring of contents as in the WWW, News or other, despite being 
discussed for applicability.

EDI-Clearing - E
Central services as archive and conversion of EDI messages

A central data excange is a natural place for clearing as well. Technical 
implemetation and - more important - operational procedures and contractual 
obligations are subject for further studies.

Access to equipment/IX location - T
Access to IX location, on demand, during business hours or 7x24h.

This service is of general interest to all ISPs and to some degree, a basic 
"must" for all IXes. Higher levels of access are subject to cost-usage 
relations.

Data collection and reporting - Q/O
Regular collection and reporting of central operational data in internal and 
public reports (peering matrices, traffic exchange statistics, ISP 
installation status, ...)

In principle of interest of all ISPs, some reservations, in particular to 
public reports, can expected from ISPs in competition.

WWW interface for publishing of reports - Q/O
Reports and operational information (trouble tickets, scheduled outages and 
maintenance, scheduled installations and upgrades) in public and private 
(password protected) web pages.

Information can optionally provided by ISPs themselves.

Only cost/usage relation will drive this process, as soon as reporting itself 
is decided upon.

Network performance evaluation - Q
An IX is an ideal place for comparative performance evaluation (RTTs, 
one-way-delays, throughput) amongst a group of ISDPs.

To a greater extend as data collection and reporting, this will put ISPs into 
being evaluated. Some reservations might come from this fact. This service 
might be the ground (reason?) for distinction between competitors in the 
future and serve for unbiased guidance for customers of ISPs.

One-way delay measurements - Q
As a corrolar from Network performance evaluation, IXes are ideal places for 
one-way delay measurements. A workstation enabled for one-way delay 
measurements can inject and measure one-way delays for a whole group of 
connected ISPs, thus easily enabling large measurements matrices with onle a 
couple of measurement points.

The cons of Network performance evaluation apply.

WebCam - T
Visual monitoring of IX premises via web. This can by used in adition of NMS 
based monitoring.

This service will be judged "far out" compared cost of implementation to 
usability. It might serve for marketing gags, though :-)

Co-locating of ISP housing - I
Co-location of ISP networking or POP equipment.

Colocating heavily drives implictions for neutrality of IXes. It certainly 
needs clearly defined interfaces and distinction between the IX sphere and 
ISP-controlled equipment.

Data transit - I
Provision of transit capacity to other IXes.

This service directly intervenes with the enbling business of cerain ISPs and 
carriers. Not much acceptance can be exected for this.