Icccn'17 Trip Notes, Days Ii As Well As 3

Keynote 2 (Day 2)

Bruce Maggs gave the keynote on Day 2 of ICCCN. He is a professor at Duke academy as well as Vice President of Research at Akamai Technologies. He talked nearly cyber attacks they direct hold seen at Akamai, content delivery network (CDN) as well as cloud services provider.

Akamai has 230K servers, 1300+ networks, 3300 physical locations, 750 cities, as well as 120 countries. It slipped out of him that Akamai is thus big, it tin select downward internet, if it went evil, but it would never become evil :-) Hmm, should nosotros say "too large to become evil?". This, of course, came upwards equally a inquiry at the halt of the talk: how prepared is the Internet for ane of the biggest players, such equally Google, Microsoft, Netflix, Yahoo, Akamai, going rouge? Bruce said, the Internet is non prepared at all. If ane of these companies turned bad, they tin melt internet. I followed upwards that inquiry alongside rouge employee as well as insider threat question. He said that, the Akamai arrangement is thus large that it doesn't/can't travel alongside manual instruction. They direct hold a real large operational room, but that is mainly to print investors. Because at that scale, the human monitoring/supervision does non work. They direct hold autonomous systems inwards place, as well as thus the direct chances of screw-up due to manual didactics is real low. (This nevertheless doesn't say much of the direct chances of a computerized screw up.)

OK, dorsum to his talk. Akamai has customers inwards eCommerce, media as well as entertainment, banks (16/20 of the global banks), as well as almost all major antivirus software vendors. He gave some daily statistics: 30+ TB/s traffic served, 600 1000000 IPv4 addresses, three trillion http requests, as well as 260 terabytes compressed logs.

Then he started talking nearly DDOS attacks, where the attackers desire to produce harm to the provider past times its overwhelming resources. The attackers oftentimes recruit an regular army of compromised drone/bot machines, as well as they human face for amplification of the requests sent past times this drone army.

Bruce showed a graph of largest DDOS attacks past times year. The attacks were exponentially growing inwards size inwards GB/s. 2017 saw the largest assault past times a element of two, where it reached 600Gbps gigabit per minute at some bespeak during the attack. WOW!

In 2016, xix attacks exceeded 100 Gbps. The March 12, 2016, DNS reflection assault reached 200 GB/s. Thursday most pop attacks are the ones alongside the largest amplification, which is defined equally the charge per unit of measurement of asking to response. DNS reflection assault has 28 to 54 amplification. The machinery used for blocking this assault was built past times "prolexic" IP anycast scrubbing centers. In this setup the rootage server had dozens of scrubbing centers/servers that filter the requests commencement as well as let only goodness ones to become the rootage server.

It looks similar these CDN guys are faring wars alongside the attackers on the Internet on a daily basis. It turns out attackers mostly perform pre-attack reconnaissance using curt outburst of attacks/probes, as well as the CDN companies also monitor/account for these tactics.

Bruce gave some statistics nearly DDOS assault frequency. The surprising affair is the gaming manufacture is the target of bulk of attacks at 55%. It is followed past times Software technology scientific discipline at 25%, Media at 5%, as well as Finance at 4%. Why target the gaming houses? A DDOS slows the online game, as well as upsets the gamers. So the attackers produce this to extort the gaming companies.

Bruce also talked nearly the assault on krebsonsecurity.com, the spider web log for the safety researcher Jay Krebs.  Akamai hosted this page pro bono. But this site got a huge assault stemming from IOT bots. This was to a greater extent than than twice inwards bulk of whatever assault they direct hold seen. Akamai held up, but afterward a twosome days of rigid attacks, this started costing honey money to Akamai, who was doing this pro bono. After September 26, Google took over hosting the Krebs site pro bono.

Bruce talked nearly many other attacks, including the  SQL attack: select * from employees where lname= '' or '1'='1'. The lesson is you lot should sanitize your SQL input! Akamai scrubs bad looking SQL.

Another assault type is bot-based concern human relationship takeover. The attackers commencement assault as well as obtain password dumps. And as well as thus they exploit the fact that many people job same username as well as password across services. The attackers as well as thus accept the large password file, intermission it into pieces, ship it to compromised routers, as well as these routers attempt these combinations alongside banking corporation accounts, as well as human face for lucky matches. This is non a DDOS attack, inwards fact they attempt to produce this inconspicuously at rates equally tedious equally twosome per hour.

My takeaway from the presentation is whatever these CDNs are charging their client companies is nevertheless a goodness deal. Because the manner things are setup currently, it is difficult for a pocket-size companionship similar a bank, media site, etc. to withstand these attacks alone. On the other hand, I hope these CDN companies remain on the superlative of their game all the time. Because they direct hold huge responsibility, they are equally good big/dangerous to fail. It is scary to mean value that it is Akamai who is serving the https, non the banks. In other words, Akamai has the someone keys for the banks, as well as serve https on their behalf.

Panel 2 (Day 2)

Panel 2 was on "Cloud Scale Big Data Analytics". The panelists were: Pei Zhang(CMU); Vanish Talwar(Nutanix); Indranil Gupta (UIUC); Christopher Stewart (Ohio State University); Jeff Kephart (IBM).

Indy Gupta talked nearly intent-based distributed systems harking dorsum to the "intent-based networking" term coined past times Cisco. He cautioned that nosotros are non catering to our existent users. We invented the internet, but missed the web. We developed p2p, but missed its applications. He cautioned that nosotros are dangerously unopen to missing the boat for large information analytics. The typical users of large information analytics are non CS graduates, but rather physics, biology, etc. domain experts. And they don't empathize "scheduling", "containers/VMs", "network as well as traffic virtualization". And neither should they live on forced to learn/understand this inwards an ideal world. They know what performance they need, such equally latency, throughput, as well as deadlines, as well as nosotros should pattern our large information systems to live on able to serve them based on these halt goals/metrics.

Jeff Kephard from IBM TJ Watson talked nearly embodied knowledge as well as symbiotic cognitive computing, but inwards a twist of fate had to attend the panel equally a disembodied Skype participant.

Yunqiao Zhang from Facebook talked nearly disaggregated storage as well as mapreduce at Facebook. The watch hither is to carve upwards the compute as well as storage resources thus that they tin evolve/sprawl/and acquire utilized independently. The 2 disaggregated systems, i.e., the compute as well as storage systems, are tethered together past times real fast Ethernet. The network speed as well as capacity today is thus good, it is possible as well as economical to produce this without worrying nearly traffic. This was real interesting indeed. I found a beak on this which I volition head to larn to a greater extent than nearly the disaggregated MapReduce at Facebook.

Pei Zhang from CMU at Silicon Valley talked nearly collecting IoT information from the physical world.

Chris Stewart from The Ohio State University talked nearly the demand for becoming transparent for large information systems from information collection, management, algorithm design, to the information translation/visualization layers.

The inquiry as well as respond session included a inquiry on the gap betwixt the information mining as well as cloud systems communities. The panel members said that to a greater extent than collaboration is needed, field it is inevitable as well as fifty-fifty useful to human face at the same problems from dissimilar perspectives. Couple panel members remarked that today the best house these communities collaborate is within companies similar Facebook as well as Google.

Keynote three (Day 3)

Henning Schulzrinne talked nearly "Telecom policy: competition, spectrum, access as well as technology scientific discipline transitions". He has been working at the authorities at the final vii years on as well as off as well as thus was able to give a dissimilar perspective than the academic. He talked nearly opportunities for enquiry that become beyond classical conference topics.

He listed the fundamental challenges as:
+ contest & investment poorly understood
+ spectrum is no longer merely bookkeeping
+ rural broadband is nearly finding the correct levers
+ emergency services nevertheless stuck inwards pre-internet

He talked at length nearly network economics. What nosotros equally CS guys direct hold been optimizing turned out to live on a real pocket-size sliver of the network economics: equipment 4%, construction 11%, operations 85%. We the CS researchers direct hold been optimizing only equipment as well as direct hold been ignoring economics! We should focus to a greater extent than on facilitating operations. Operations is non efficient, if nosotros tin figure out how to brand networks to a greater extent than easily operable, as well as require less human resources, nosotros volition direct hold larger touching than tweaking protocols.

He talked also nearly rural broadband, as well as mentioned that the drones/balloons are inapplicable equally their capacity is non enough. The terms of deployment inwards rural is high, as well as the incentive for companies to deploy is low. But, pretty much everyone has wired telephone service, how did that happen? There was an unspoken bargain: the authorities said to ATT we'll give you lot monopoly, as well as you'll give us universal service. This was never stated but understood. He said to solve the rural broadband problem, policy levers demand to pulled.
+ decrease terms of serving: dig once: bury cable during street repair & construction
+ render funding: universal service fund (US $8 billion from revenue enhancement money).

He talked nearly recycling TV broadband spectrums as well as how this is a real active termination now. He also talked nearly serving the disabled via the subtitles requirements, text-to-911, voip emergency, as well as wireless 911 services.

To conclude he asked us to mean value nearly the network job holistically, including economic science as well as policy inwards the equation. Many of the problems are incentive problems. And at that topographic point is a demand to mean value inwards decades non conference cycles! The network performance is rarely the fundamental problem; academics travel on things that tin live on measured, fifty-fifty when they are non that important.

Panel three (Day 3)

The Panel three was on "Federal Funding for Research inwards Networking as well as Beyond". The panelists were Vipin Chaudhary (US NSF); Richard Brown (US NSF); as well as Reginald Hobbs (US Army Research Lab).

Rick Brown talked nearly NSF CISE divisions:
+ CNS: estimator network systems
+ CCF: computing as well as communication foundations
+ IIS: information & intelligent systems
+ OAC: Office of Advanced Cyberinfrastructure

He mentioned that NSF was established past times congress inwards 1950 alongside the yearly  budget of $3.5 billion, alongside the post service ww2 agreement of importance of scientific discipline to the country. NSF promotes bottom upwards basic enquiry civilization inwards contrast to NIH NASA DARPA which tells you lot what to travel on as well as build.

The full NSF 2017 budget 7.8 billion. NSF gets around 50K proposals, funds 10K of them. 95% of budget goes to the grants, only 5% goes to operational costs.

Reginal Hobbs talked nearly the funding & enquiry collaboration opportunities at the Army Research Laboratory.

Vipin Chaudhary talked commencement nearly NSF broadly as well as and thus specifically nearly the  office of advanced cyberinfrastructure at NSF. He said that the CISE budget is merely about 840M, as well as inwards estimator scientific discipline 83% of academic enquiry inwards CS is covered past times NSF. (I didn't await this ratio to live on this high.)

He described the NSF I-Corps program at the halt of his talk, which was real interesting for its back upwards for entrepreneur activities. This programme helps you lot to figure out if you lot are facing valley of decease or dark hole inwards your enquiry commercialization process. Most academic spinouts neglect because they railroad train something no ane cares about. I-Corps provides back upwards for you lot to come across alongside customers as well as exam your hypothesis nearly what your production should live on based on their feedback.

Share this post