{"id":862,"date":"2012-04-25T17:24:18","date_gmt":"2012-04-25T17:24:18","guid":{"rendered":"http:\/\/www.syslog.cl.cam.ac.uk\/?p=862"},"modified":"2012-04-26T18:47:42","modified_gmt":"2012-04-26T18:47:42","slug":"nsdi-2012-session-1","status":"publish","type":"post","link":"https:\/\/www.syslog.cl.cam.ac.uk\/2012\/04\/25\/nsdi-2012-session-1\/","title":{"rendered":"NSDI 2012 Day 1"},"content":{"rendered":"

\"\"<\/a>NSDI 2012@San Jose Fairmont.<\/p>\n

Other info from here:-<\/p>\n

https:\/\/www.usenix.org\/conference\/nsdi12\/tech-schedule\/technical-sessions<\/p>\n

<\/p>\n

Session 1 - Big Data.<\/p>\n

Corfu - clusers of raw flash units
\n(also island in greece near Paxos)....
\nproduces orderd list of events...<\/p>\n

Network attached Flash - aves power&latency
\nThen do log-structured stuff on it....
\nperf with emulated unit does 1M event per sec....v. impressive<\/p>\n

comparison - paxos is i\/o bound...corfu isn't...<\/p>\n

one good question about what happens if Flash guys actually deliver on some
\npromises - how would that change corfu - basically, not a lot.<\/p>\n

nice work - a lot of effort<\/p>\n

Resilient Distributed Datasets (RDDs):<\/p>\n

A Fault-Tolerant Abstraction for In-Memory Cluster Computing
\nmapreduce paper:- replacement for HDFS
\nRDDs are really just another level of indirection with a query API
\nto the aggregator function -<\/p>\n

implemented in Spark in Scala + dryadlinq like api (open sourced)<\/p>\n

eval on example of log mining
\nfull text search on wikipedia (50Gig) in seconds.,,<\/p>\n

Seemed rather obvious to me - language looks quite nice though..
\n{see Nextar@OSDI10 - does cite - also cites CIEL:-)<\/p>\n

Brave presenter did a live demo...which worked...<\/p>\n

Question - what happens when not enough ram - thrash to disk...
\nspeaker had backup slide that showed performance hit...fixing this
\nmaybe distributing memory load\/work was future work...<\/p>\n

Camdoop: Exploiting In-network Aggregation for Big Data Applications- Paolo Costa<\/p>\n

More mapreduce throughput speedup work
\nThis looks at how data reduction maps to actual network topology (viz fattree etc)<\/p>\n

Observe (as others have) that final results in many typcal apps (taken from facebook examples)
\nare 5% of intermediate data<\/p>\n

So use this to do some aggregation in an intermediate stage - now try and do this
\nin a distributed way - problem is how to map logical disributed aggregation
\nonto physical topology - creates hot spots in the aggregation tree....<\/p>\n

mismatch between logical&physical topology..
\nCamdoop targets reducing shuffle time by packet aggregation
\ntakes advantage of CamCube topology
\nSo map tree to camcube -see cute diagrams in paper<\/p>\n

eval - built a 3^3 camcube with quad core servers + have same code in simulator with 8^3 eval
\nlook at worst case perf (good for them) in bisection bandwidth - gain is good exactly where they want it -
\nshow all-to-one and all-to-all - looks like about a 5 fold perf gain in common case (fb workload)<\/p>\n

Wireless Session<\/p>\n

1. WiFi-NC : WiFi Over Narrow Channels from MSR India<\/p>\n

Related idea to Bogdan's thesis
\n(nice to see yet another talk not doing the powerpoint thing too:)
\nbasically doing clever stuff with subcarriers\/narrow band to get effiecient coding\/use of channels -
\nthis works within ISM and OFDM, but also for Whitespace stuff, with fragmented spectrum. Essentially
\nget more parallelism with les interference due to independence of lots of narrow channels than 1 wideband channel (kind of obvious, but they've built it, and the applicability\/flexibility to also work in whitespace with fragmented spectrum is sweet)...tradeoff really is about guardbands and multipath.<\/p>\n

Question about 1 pb is preabmble dilation. yes, is an issue<\/p>\n

2. Catching Whales and Minnows Using WiFiNet: Deconstructing Non-WiFi Interference Using WiFi Hardware<\/p>\n

This is about capturing what kind of non-WiFi source is causing us (a WiFi transceiver) grief.
\nThere are some devices\/sources that are minor problems (not really disruptive) and others which are a big problem (hence, repsectively, minnows & whales) - is extension of NSDI11 PIE paper (passive interference estimation)<\/p>\n

runs airshark on multiple APs, which can pull samples. These are reported to a combiner which then identifies sources
\nanalogue phone, zigbee or bluetooth...this system (WiFiNet Controller) uses the multiple sources of idenified interferrers and infers how many of each type there may be.<\/p>\n

Also uses a model-based mechanism to find the probable localaton of the devices<\/p>\n

Evalation is empirical - the errors in locationation is median 2-4 meters (worst case was 15)
\nLots of interference estimation work in the paper - looks useful and solid piece of nice engineering.<\/p>\n

[Actually, as Dr Hand pointed out, after a while the slides in this session
\nbecame somewhat overwealming:)<\/p>\n

Content Networking - Session 3<\/p>\n

RPT: Re-architecting Loss Protection for Content-Aware Networks - from CMU<\/p>\n

This is the idea of repeated transmission, with de-duping to suppress redundency
\n[Note to self - PGM did this for reliable multicast - retranmsissions were filtered
\nin subtrees that had received early copies of same data packet - RPT is slightly
\nmore expensive in the sense that you don't just keep a sequence number, but a cache of all the actual recent
\ndata - PGM also had a mode like that, since retransmits (or first transmits for new receivers) could be from local subtree cache only:)<\/p>\n

Eval is through measurement of Click implementation. Workload is video, and comparison is FEC v. RPT
\nThe implemenation is (surely) very like what you need for CCN\/ICN\/NDN (i.e. blackadder:)<\/p>\n

I personally think this is a mad idea:)<\/p>\n

Serval: An End-Host Stack for Service-Centric Networking - Erik\/Princeton<\/p>\n

Now this is nice - its a shim that makes things like mobility of ends and the like much easier -
\nwell thought through, and architected...
\nSo if you do migration, serval's abstraction hides this nicely..<\/p>\n

Essentially, service name+id provides level of indirection - uses late binding.
\nHost s\/w is a classic API\/Shim.
\nThe lookup\/mapping looks quite a lot like _signpost_...
\nCode is in kernel and zero copy, so overheads are very low - so their
\nperformance is very good (140,000 connection setups per second, throuhgput and latency overhead is within noise!)
\nLOC for delta is (for example in Firefox) 70 lines of code...
\nAside on SDN to the edge of the net!
\nSDN works on layer 2\/3 abstraction - Serval generalises this, Erik argues.<\/p>\n

A _good_ thing. Dr Hand points out its a lot like the middleware bits in DCE\/Corba\/DCAN etc, which is fair.<\/p>\n

Reliable Client Accounting for P2P-Infrastructure Hybrids - MPI+Akamai<\/p>\n

This is the next phase of evolution for Akamai, so is v. interesting - they are building out hybrids of CDN-classic + p2p - according to bruce maggs (personal communication) they have 35000 peers to assist the core CDN setup.
\n[one nice result is that one can do torrent-style optimally load balanced delivery within a single AS\/ISP, but then use configuration of the CDN-classic system to respect the border peering\/traffic agreements _across ASs\/ISPs.<\/p>\n

This paper is maily about verifying the peers don't freeload (etc).
\nThey make use, essentially of MPI's PeerReview (SOSP 07). That sort of makes it a fairly simple trick<\/p>\n

A NetSession system contains rules that are what are checked by the peerreview scheme<\/p>\n

Eval:- Client & CDN's perspective - net and CPU cost and log storage - all reasonable (as you'd expect from PeerReview)<\/p>\n

Network Robustness - Session 4<\/p>\n

Header Space Analysis: Static Checking for Networks, stanford, UCSD<\/p>\n

This is a formal methodish paper which then is used a bit predicate-routinglyishly
\nto determine what should be where and when. Seems really quite useful and practical. The use of transfer functions and the header space transforms they perform is applied over the topology to compute what can do what.<\/p>\n

The paper was evaluated through use cases and shown to be useful:)<\/p>\n

A NICE Way to Test OpenFlow Applications, EPFL\/Princeton<\/p>\n

This is from Marco et al (was here for a bit) and is quite nice - maybe we should look at how we'd do this similarly with
\nHaris' Mirage OpenFlow? Quite a lot of this work is about state space reduction to make the checking work tractable.<\/p>\n

Toward Predictable Performance in Software Packet-Processing Platforms, EPFL<\/p>\n

Reminds me of Active Nets work (e.g. safetynet work at Sussex and JMS' work on ML and active packets) where you bound behavioru by limiting computational expressiveness....<\/p>\n","protected":false},"excerpt":{"rendered":"

NSDI 2012@San Jose Fairmont. Other info from here:- https:\/\/www.usenix.org\/conference\/nsdi12\/tech-schedule\/technical-sessions<\/p>\n","protected":false},"author":5,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[1],"tags":[],"_links":{"self":[{"href":"https:\/\/www.syslog.cl.cam.ac.uk\/wp-json\/wp\/v2\/posts\/862"}],"collection":[{"href":"https:\/\/www.syslog.cl.cam.ac.uk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.syslog.cl.cam.ac.uk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.syslog.cl.cam.ac.uk\/wp-json\/wp\/v2\/users\/5"}],"replies":[{"embeddable":true,"href":"https:\/\/www.syslog.cl.cam.ac.uk\/wp-json\/wp\/v2\/comments?post=862"}],"version-history":[{"count":13,"href":"https:\/\/www.syslog.cl.cam.ac.uk\/wp-json\/wp\/v2\/posts\/862\/revisions"}],"predecessor-version":[{"id":897,"href":"https:\/\/www.syslog.cl.cam.ac.uk\/wp-json\/wp\/v2\/posts\/862\/revisions\/897"}],"wp:attachment":[{"href":"https:\/\/www.syslog.cl.cam.ac.uk\/wp-json\/wp\/v2\/media?parent=862"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.syslog.cl.cam.ac.uk\/wp-json\/wp\/v2\/categories?post=862"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.syslog.cl.cam.ac.uk\/wp-json\/wp\/v2\/tags?post=862"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}