Uncategorized

PlanAlyzer: assessing threats to the validity of online experiments

~ Adrian Colyer ~ 1 Comment
PlanAlyzer: assessing threats to the validity of online experiments Tosch et al., OOPSLA'19 It’s easy to make experimental design mistakes that invalidate your online controlled experiments. At an organisation like Facebook (who kindly supplied the corpus of experiments used in this study), the state of art is to have a pool of experts carefully review ... Continue Reading

Taiji: managing global user traffic for large-scale Internet services at the edge

~ Adrian Colyer ~ 1 Comment
Taiji: managing global user traffic for large-scale internet services at the edge Xu et al., SOSP'19 It’s another networking paper to close out the week (and our coverage of SOSP’19), but whereas Snap looked at traffic routing within the datacenter, Taiji is concerned with routing traffic from the edge to a datacenter. It’s been in ... Continue Reading

Scaling symbolic evaluation for automated verification of systems code with Serval

~ Adrian Colyer ~ 2 Comments
Scaling symbolic evaluation for automated verification of systems code with Serval Nelson et al., SOSP'19 Serval is a framework for developing automated verifiers of systems software. It makes an interesting juxtaposition to the approach Google took with Snap that we looked at last time out. I’m sure that Google engineers do indeed take extreme care ... Continue Reading

The inflection point hypothesis: a principled approach to finding the root cause of a failure

~ Adrian Colyer ~ 1 Comment
The inflection point hypothesis: a principled debugging approach for locating the root cause of a failure Zhang et al., SOSP'19 It’s been a while since we looked a debugging and troubleshooting on The Morning Paper (here’s a sample of earlier posts on the topic). Today’s paper introduces a root cause of failure detector for those ... Continue Reading

File systems unfit as distributed storage backends: lessons from ten years of Ceph evolution

~ Adrian Colyer ~ 1 Comment
File systems unfit as distributed storage backends: lessons from 10 years of Ceph evolution Aghayev et al., SOSP'19 Ten years of hard-won lessons packed into just 17 pages (13 if you don’t count the references!) makes this paper extremely good value for your time. It’s also a fabulous example of recognising and challenging implicit assumptions. ... Continue Reading

Optimized risk scores

~ Adrian Colyer ~ 3 Comments
Optimized risk scores Ustun & Rudin, KDD'17 On Monday we looked at the case for interpretable models, and in Wednesday’s edition of The Morning Paper we looked at CORELS which produces provably optimal rule lists for categorical assessments. Today we’ll be looking at RiskSLIM, which produces risk score models together with a proof of optimality. ... Continue Reading