Ph.D. Krishna P. Gummadi: Glasnost at the Max Planck Institute – Transparency for the internet

Broadband networks such as DSL or ca­­ble are used more and more widely for accessing the internet. Over 158 million people worldwide use these networks. By 2011, this number is expected to rise to 477 million. In Germany, over 65 per cent of all internet users are connect­­ed to broadband networks at home. In addition, many governments have made it their target to promote universal broad­­band access.

Broadband networks deliver ac­­cess to the internet infrastructure and most of the time, bottlenecks to these access net­­works can be found in the ef­­­ficiency of the internet. Many popular in­­ternet ap­­pli­­cations, such as voice over IP (VoIP), vid­­eo on demand, online games and peer-to-peer networks, greatly de­­pend upon the performance qualities of broadband networks.
However, many private customers are ignorant of the characteristics of their broadband connections.

Cable and DSL providers increasingly use middle boxes, such as for example traffic shaper, blocker or firewalls, for the super­­vision and control of their cli­­ents’ data traffic. Middleboxes categorize and man­ipulate the flow of data from the various applications in accordance with ISP-specific guidelines. And because the guidelines of data traffic management are often in­­fluenced by company interests (for example peering agreements between ISPs, which are mutually regulated by con­­tract), indica­tions to middlebox implementation is not at all made known by many ISPs.
This is why today, many end users are not even aware of the existence of middlebox­­es and can therefore often not imagine their consequences on the performance of internet applications.

Recently, it has been stated that specific access providers in the US secrectly pre­­vented their clients from exchanging data with other users by means of the popular file-sharing protocol BitTorrent. It was determined that these ISPs blocked Bit­­­Torrent data flows by sending false pro­to­­col packages to communication end-points, which leads to an immediate in­­­terruption of the connection. Reports on such blockages started an in­­tensive and far-reaching guideline con­­­tro­versy over acceptable practices in the traf­­fic management of ISPs and the issue of network neutrality among ISPs, us­­er protec­­tors, in­­ternet providers and government authorities.

We do not want to discuss the advan­tag­es of the various traffic management guidelines here. Neither do we wish to plead for specific guidelines, such as net­­work neu­tral­­ity, which does not allow for a different treatment of packages by ISPs, on grounds of the communication end-points, nor the used internet application.

The formulation of optimal traffic manage­­ment guidelines is extremely important and demands the careful consideration of numerous technological, economic and social factors, which are decisive for the future of the internet.

Instead, we want to plead for net­­work transparency. In simple terms, net­­­work transparency demands that end us­­­ers know the characteristics of the access networks they use. This knowledge helps users when choosing their provider. If this transparency existed, end users would know which quality of service they receive, when using different applications such as VoIP or video on demand.
Also, ap­­plication de­­velopers can use such know­­ledge to spe­cial­­ly create their ap­­pli­­cations in such a way that they run better in the home environ­ment.

Ideally, networks could be cre­­at­­ed with more transparency through ISPs, in that they dis­­close how they control their data traffic and where they im­­ple­ment traf­­­fic manage­ment. How­­ever, ex­­perience has proven that providers are rather reluctant to disclose such information, even when they are required to do so by state su­per­­visory authorities. Further­­more, in­­­for­­mation provided by ISPs to their cus­tomers is often vague and at best in­­ter­­pretable in many ways, at worst, near-misleading.

For this reason, end users need appropriate mechanisms, with which they can ex­­amine information received from their ISPs. With this in mind, we at the Max Planck Institute for Software Sys­tems, have initiated the Glasnost pro­ject. Our aim is to enable to draw con­­clusions on the characteristics of their ac­­cess networks through direct measurements. The Glasnost system is cre­­­ated in such a way that it approaches numerous tricky duties: It attempts to find out which data traffic is being ma­­nipulated and where traffic shapers can be found. Interested readers can carry out the test by visiting the following website:


Krishna_Gummondi_009Ph.D. Krishna P. Gummadi is the head of the Networked Systems Research Group at the Max Planck Institute for Soft­­ware Systems in Saarbrücken. Be­­fore this, he worked as a research assistant at Microsoft in Redmond, ISCI and Intel in Berkley after completing his internship at the Polytechnische Hochschule Lausanne. Gummadi studied in Madras, India, and at the University of Wash­ing­­ton (USA), where he also completed his doctorate.