cisc 849 : applications in fintech mengyuan ren dept of computer & information sciences university...

Download CISC 849 : Applications in Fintech Mengyuan Ren Dept of Computer & Information Sciences University of Delaware Are Clouds Ready to Accelerate Ad Hoc Financial

If you can't read please download the document

Upload: william-manning

Post on 18-Jan-2018

219 views

Category:

Documents


0 download

DESCRIPTION

CISC 849 : Applications in Fintech Aggregate Risk Analysis  A Monte-Carlo like simulation performed on a portfolio of risk held by a reinsurer.  Aggregate Risk - The exposure of a bank, financial institution, or any type of major investor to foreign exchange contracts from a single counterparty or client.

TRANSCRIPT

CISC 849 : Applications in Fintech Mengyuan Ren Dept of Computer & Information Sciences University of Delaware Are Clouds Ready to Accelerate Ad Hoc Financial Simulations? CISC 849 : Applications in Fintech What Financial Simulations? CISC 849 : Applications in Fintech Aggregate Risk Analysis A Monte-Carlo like simulation performed on a portfolio of risk held by a reinsurer. Aggregate Risk - The exposure of a bank, financial institution, or any type of major investor to foreign exchange contracts from a single counterparty or client. CISC 849 : Applications in Fintech What is the Challenges? 1.Input Data v.s. Memory Bandwidth 2.Memory Architecture CISC 849 : Applications in Fintech Parameters YET={T i ={(E i,1,t i,1),,(E i,k,t i,k)}} Where i=1,2, and k=1,2,,800,,1500 YET ELT PF L CISC 849 : Applications in Fintech Parameters EL i ={E i,l i } ELT= FT 1 =(FT 11,FT 12,) Where i=1,2,,100000,,30000 YET ELT PF L CISC 849 : Applications in Fintech Parameters PF={P1,P2,,Pn} where n=1,2,,10 YET ELT PF L CISC 849 : Applications in Fintech Parameters ={ELT1,ELT2,...,ELTj} L= FT2=(FT21,FT22) FT3=(FT31,FT32) Where i=1,2,,100000,, YET ELT PF L CISC 849 : Applications in Fintech Algorithm 1for each Program, P, in PF do 2for each Layer, L, in P do 3for each Trial, T, in YET do 4for each Event, E, in T do 5for each ELT covered by L do 6Lookup E in the ELT and find corresponding loss, lE 7Apply Financial Terms FT1 to lE 8lT lT + lE 9end 10Apply Financial Terms FT2 to lT 11Apply Financial Terms FT3 to lT 12end 13end 14end 15endInput:YET, ELT, PF 16Populate YLT using lTOutput:YLT CISC 849 : Applications in Fintech Platform Instance TypeNo. of VirtualMemoryProcessor TypeClock Speed CPUs (VCPU)(GiB)(GHz) m1.xlarge415.0Intel Xeon E m2.xlarge217.1Intel Xeon E m2.2xlarge434.2Intel Xeon E m2.4xlarge868.4Intel Xeon E m3.xlarge415.0Intel Xeon E m3.2xlarge830.0Intel Xeon E cr1.4xlarge Intel Xeon E Amazon cloud VMs cc1.4xlarge1623.0Intel Xeon X cc2.8xlarge3260.5Intel Xeon X hi1.4xlarge1660.5Intel Xeon E hs1.4xlarge Intel Xeon E cg1.4xlarge1622.5Intel Xeon X NVIDIATesla0.575 M2050 vm Intel Xeon E vm Intel Xeon E Private VMs NVIDIATesla0.575 M2050 VMs employed for Aggregate Risk Analysis CISC 849 : Applications in Fintech Implementation on the Cloud Time taken when number of Trials are varied on the fastest public and private VMs CISC 849 : Applications in Fintech Implementation on the Cloud Parallel execution (single thread per virtual core) of Aggregate Risk Analysis on public (a) to (e) and private (f) VMs CISC 849 : Applications in Fintech Implementation on the Cloud Parallel execution (multiple threads per virtual core) of Aggregate Risk Analysis on public (a) to (e) and private (f) VMs CISC 849 : Applications in Fintech Implementation on the Cloud Parallel execution of Aggregate Risk Analysis on public and private VMs CISC 849 : Applications in Fintech Implementation on the Cloud Execution of Aggregate Risk Analysis on a single GPU on public and private VMs CISC 849 : Applications in Fintech Implementation on the Cloud Performance of multiple GPUs CISC 849 : Applications in Fintech Implementation on the Cloud Aggregate Risk Analysis on public and private VMs with multiple GPUs CISC 849 : Applications in Fintech Conclusion 1.It is noted that the private VMs perform better than the public VMs. 2.Acceleration could not be run most efficiently on the public cloud since there was wasted computational time for every dollar spent. CISC 849 : Applications in Fintech Thanks for your time!