Security Aware and Energy-Efficient Virtual Machine Consolidation in Cloud Computing Systems Farhad Ahamed, Seyed Shahrestani and Bahman Javadi School hool of of Comput omputing, ing, Engineer ngineering ing and and Mathema hematics ics Wes ester ern n Sydney dney Univ Univer ersit ity, , Aus ustralia alia
Secur ecurit ity Concer oncerns ns in in Cloud loud Comput omputing ing Attac ack k types pes – – cons consider idering ing res esour ource ce shar haring ing is issue ue
VM consolidation – What and why ? How ¡do ¡we ¡ensure ¡that ¡our ¡data ¡in ¡the ¡Cloud ¡is ¡not ¡ living ¡with ¡a ¡malicious ¡neighbour? ¡
Propos oposed ed Appr pproac oaches hes a) The Compartment Isolation Approach of Secure Virtual Machine Consolidation b) Security Profiles for the Virtual Machines PAGE 4 PAGE 4
Secur ecure e VM Cons onsolida olidation ion The Compartment Isolation Approach of Secure Virtual Machine Consolidation SIR model • Susceptible • Infected • Recovered Using the isolated compartment strategy; separation of VMs reduces the probability cross computer malware spreading. PAGE 5 PAGE 5
Secur ecure e VM Cons onsolida olidation ion Security Profiles for the Virtual Machines PAGE 6 PAGE 6
Secur ecure e VM Cons onsolida olidation ion Security Profiles for the Virtual Machines PAGE 7 PAGE 7
Int nter ernal nal vulner ulnerabilit bility Internal vulnerability and Intrusion analysis score PAGE 8 PAGE 8
Ranking of the security profiles of the VMs
High le High level el pr procedur ocedure e of of the he VM mig migration ion VM Mig igration ion Select election ion VM Mig igration ion Placement lacement 1. 1. Random andom Select election ion (RS) 1. 1. Local Local Regres ession ion (LR LR) 2. Minimum 2. inimum Mig igration ion Time ime (MMT) 2. Local 2. Local Regres ession ion Robus obust (LR LRR) 3. Maximum 3. aximum cor correla elation ion (MC) 3. Median 3. edian Abs bsolut olute e Devia iation ion (MAD 4. 4. Minimum inimum Ut Utiliz ilization ion (MU) U) 4. 4. Static ic Thr hres eshold hold (THR HR)
Secur ecure e VM Select election ion Algor lgorit ithm hm (SBS)
Secur ecure e VM Placement lacement Algor lgorit ithm hm (SBP)
Simula imulation ion Set etup up a) CloudSim Simulator b) PlanetLab Workload c) Power Consumption Profile PAGE 13 PAGE 13
VM and PM initial configuration for simulation Virtual Machine Details Virtual Machine Details Total MIPS of VM 2500 Total PES (Processor unit) of VM 1 Total RAM of VM 1024 MB Network Bandwidth of VM 100 Mbit/s Total Storage size of VM 2.5 GB Physical Machine Details Physical Machine Details Total MIPS of PM 2660 Total PES (Processor unit) of PM 2 Total RAM of PM 8192 MB Total Storage size of PM 80 GB PAGE 14 PAGE 14
Security aware VM consolidation comparison for low- medium workload for 800 hosts 322.000 ¡ 320.000 ¡ 318.000 ¡ 316.000 ¡ 314.000 ¡ kWh ¡ 312.000 ¡ 310.000 ¡ 308.000 ¡ 306.000 ¡ 304.000 ¡ 2 ¡ 3 ¡ 4 ¡ 5 ¡ 6 ¡ 7 ¡ 8 ¡ 9 ¡ 10 ¡ Security ¡level ¡ PAGE 15 PAGE 15
nergy consumption versus security graph when the number of PMs are 1000, 2000 and 3000 1000 Hosts 5100 2000 Hosts 3000 Hosts 5000 4900 kWh kWh 4800 4700 4600 4500 1 2 3 4 5 6 7 8 9 10 Security level Security level PAGE 16 PAGE 16
Security level vs. energy consumption in multiple security-aware algorithms 340.000 330.000 Secure IQR-MC 320.000 Secure MAD-MMT kWh kWh Secure LLR-MMT 310.000 Secure LR-MU 300.000 Secure LR-MMT Secure THR-MMT 290.000 280.000 2 3 4 5 6 7 8 9 10 Security level Security level PAGE 17 PAGE 17
Conc onclus lusions ions • Introduced the compartment isolation technique to achieve the security aware VM consolidation. • Di ff erent types of simulation setup and the subsequent result confirms that there are no abrupt changes in power consumption to achieve security aware VM consolidation. • The solution presents an added protection measure with the minimal impact on energy e ffi cient algorithm. • This work could be extended to improve the VM reliability as well as security and energy This work could be extended to improve the VM reliability as well as security and energy consumption. consumption. PAGE 18 PAGE 18
Recommend
More recommend
