Researchers discover major roadblock in alleviating network congestion: Algorithms designed to ensure multiple users share a network fairly can’t prevent some users from hogging all the bandwidth.

when usrs wanna send data ‘oer the internet faster than the network can handle, congestion can occur — the same way traffic congestion snarls the morning commute into a big city.

computers and devices that transmit data ‘oer the internet break the data down into liler packets and use a spesh algorithm to decide how fast to send those packets. these congestion control algorithms seek to fully discover and utilize available network cap while sharing it fairly with other usrs who maybe sharing the same network. these algorithms try to minimize delay caused by data w8in in queues inna network.

‘oer the past decade, researchers in industry and academia ‘ve developed several algorithms that attempt to achieve high rates while controlling delays. some of these, s'as the bbr algorithm developed by g, are now widely used by many websites and applications.

but a team of mit researchers has discovered that these algorithms can be deeply unfair. in a new study, they show there will always be a network scenario where at least one sender receives almost no bandwidth compared to other senders; that is, a problem known as starvation cannot be avoided.

“wha’ is really surprising bout this paper na results s'dat when you take into account the real-realm complexity of network paths and all the things they can do to data packets, tis basically impossible for delay-controlling congestion control algorithms to avoid starvation using current methods,” says mohammad alizadeh, associate professor of electrical engineering and computer sci (eecs).

while alizadeh and his co-authors weren’t able to find a traditional congestion control algorithm that ‘d avoid starvation, there maybe algorithms in a ≠ class that ‘d prevent this problem. their analysis also suggests that changing how these algorithms work, so t'they allo for larger variations in delay, ‘d help prevent starvation in some network situations.

alizadeh wrote the paper with 1st author and eecs graduate student venkat arun and senior author hari balakrishnan, the fujitsu professor of computer sci and artificial intelligence. the research ll'be presented atta acm spesh interest group on data communications (sigcomm) conference.

controlling congestion

congestion control is a primordial problem in networking that researchers ‘ve been trying to tackle since the 1980s.

a usr’s computer does not know how fast to send data packets ‘oer the network cause it lacks information, s'as the quality of the network connection or how many other senders are using the network. sending packets too sloly makes poor use of the available bandwidth. but sending them too quickly can overwhelm the network, and in doin’ so, packets will start t'get dropped. these packets must be resent, which leads to longer delays. delays can also be caused by packets w8in in queues for a long time.

congestion control algorithms use packet losses and delays as signals to infer congestion and decide how fast to send data. but'a internet is complicated, and packets can be delayed and lost for reasons unrel8d to network congestion. for instance, data ‘d be held up in a queue along the way and then released witha burst of other packets, or the receiver’s ackment mite be delayed. the authors call delays tha're not caused by congestion “jitter.”

even if a congestion control algorithm measures delay perfectly, it can’t tell the difference tween delay caused by congestion and delay caused by jitter. delay caused by jitter is unpredictable and confuses the sender. cause of this ambiguity, usrs start estimating delay ≠ly, which causes them to send packets at un= rates. eventually, this leads to a situation where starvation occurs and some1 gets shut out completely, arun explains.

“we started the project cause we lacked a theoretical cogging of congestion control behavior inna presence of jitter. to place it na' firmer theoretical fting, we built a mathematical model twas' simple enough to think bout, yet able to capture somd' complexities of the internet. it s'been very rewarding to ‘ve math tell us things we didn’t know and that ‘ve practical relevance,” he says.

studying starvation

the researchers fed their mathematical model to a computer, gave it a series of comm1-ly used congestion control algorithms, and asked the computer to find an algorithm that ‘d avoid starvation, using their model.

“we ‘dn’t dweet. we tried every algorithm that we're aware of, and some new ones we made up. nothing worked. the computer always found a situation where some pplz get all the bandwidth and at least one person gets basically nothing,” arun says.

the researchers were surprised by this result, espeshly since these algorithms are widely believed to be reasonably fair. they started suspecting that it may not be possible to avoid starvation, an extreme form of unfairness. this motivated them to define a class of algorithms they call “delay-convergent algorithms” t'they proved will always suffer from starvation under their network model. all existing congestion control algorithms that control delay (that the researchers are aware of) are delay-convergent.

the fact that such simple failure modes of these widely used algorithms remained unknown for so long illustrates how difficult tis to cogg algorithms through empirical testing alone, arun adds. it underscores the importance offa solid theoretical foundation.

but all hope aint lost. while all the algorithms they tested failed, there maybe other algorithms which aint delay-convergent that mite be able to avoid starvation this suggests that one way to fix the problem mite be to design congestion control algorithms that vary the delay range + widely, so the range is larger than any delay that mite occur due to jitter inna network.

“to control delays, algorithms ‘ve tried to also bound the variations in delay bout a desired equilibrium, but thris nothing wrong in potentially creating gr8r delay variation t'get better measurements of congestive delays. tis just a new design philosophy you ‘d ‘ve to adopt,” balakrishnan adds.

now, the researchers wanna keep pushing to see iffey can find or build an algorithm thall eliminate starvation. they also wanna apply this approach of mathematical modeling and computational proofs to other thorny, unsolved problems in networked systems.

“we're increasingly reliant on computer systems for very crit things, and we nd'2 put their reliability na' firmer conceptual fting. we’ve shown the surprising things you can discover when you put inna time to come up with these formal specifications of wha’ the problem actually is,” says alizadeh.

original content at:…