Upload
mayank-chaudhari
View
109
Download
1
Embed Size (px)
Citation preview
Comparative study of
VMWare ESX with other similar
products
Monolithic v/s Microlithic• Hypervisors can be classified as monolithic or
microlithic based on the kernel organization.• Microlithic (Microkernel based)
• Xen• Hyper - V
• KVM (controversial)• closer to monolithic• VM -> Linux kernel space -> Linux user space
-> Linux kernel space -> VM
VMWare ESXi• Type 1 monolithic hypervisor• VMkernel is responsible for virtualization• The VMkernel provides support for :
• Resource scheduling• File System• Users and Groups
• Users : Virtual Infrastructure Client, Remote Command Line Interface,VIM API
• Groups : To combine multiple users and assign privileges as a group
VMWare ESXi
Type 1 Hypervisors• VMWare - Proprietary
• Xen Hypervisor - Open Source
• KVM - Open Source, (free)
• Microsoft Hyper-V - Proprietary
• WIND RIVER - proprietary
• LynxSecure Separation Kernel Hypervisor - Proprietary
• Proxmox VE - Open Source, (free)
• nuxi -Open Source , (free)
• RTS Real Time Hypervisor - Proprietary
• Vembu VMBackup - Proprietary
Microsoft Hyper - V• Hypervisor based virtualization technology for X64
based Windows Server• Two variants
• Standalone• Role based
• Requires a processor with hardware assisted virtualization functionality.
• Implements isolation using partition.• Parent partition manages resources and create
child partition.• Child partition do not have direct access to
resources
Microsoft Hyper - V• Host types
• Enlightened• Unenlightened
• Parent runs (VSP)• Child run (VSC)• Communication between partition takes place
through VMBus (logical channel)
Xen Hypervisor• Developed at University of Cambridge• Only available bare-metal Open Source hypervisor• Components
• Domain 0 (Dom0)• Domain U (Dom U)
• Lightweight due to delegation of management guest.
• Dom 0 is a linux kernel with special privileges.• Dom U do not have any direct access to physical
resources.• Uses CREDIT scheduler with round robin
scheduling of VCPUs.
Xen Hypervisor• Uses emulation to provide support for disk,
network, motherboard, and PCI devices• Shadow code to virtualize psagetables• Uses emulated interrupt controllers• Integrated ROMBIOS to provide virtual BIOS to the
guest
KVM Hypervisor• Type 1 hypervisor• Uses virtualization technology support for full
virtualization• Uses Shadow page tables for memory
management• Guest sees CPU, RAM, disk, etc like on real
machines• Processes can create virtual machines• Guest physical memory part of creating process'
address space• Proximity of guest and user space hypervisor• Massive Linux kernel reuse• Pass-through PCI adapters, disks, etc also possible
Feature wise comparison
Computing for Sustainable Global Development (INDIACom), 2016 3rd International Conference on
Feature wise comparison
https://www.proxmox.com/en/proxmox-ve/comparison
Comparison based on various tests
• Components to be tested• CPU• Disk I/O• Memory • Network I/O
Performance comparison• CPU test (Bytemark)
• All hypervisors performed quite similarly• CPU operations require no help from
hypervisor• Memory test (ramsmp)
• with single VCPU similar results• with all VCPU Xen and KVM shows
performance drop• KVM and Xen uses heavy management
systems
2013 IFIP/IEEE International Symposium on Integrated Network Management (IM 2013)
Performance comparison
• Disk I/O test (Bonnie++ & FileBench)• Xen has poor performance• KVM beats all • Hyper-V experience thrashing with multiple
VCPU• Network test (Netperf)
• Xen has poor performance due to network transmission using network backend driver
• More indirection
2013 IFIP/IEEE International Symposium on Integrated Network Management (IM 2013)
References1. Benchmarking the Performance of Microsoft Hyper-V
server,VMware ESXi and Xen Hypervisors,Hasan Fayyad-Kazan,Luc Perneel,Martin Timmerman,Journal of Emerging Trends in Computing and Information Sciences,Vol. 4, No. 12, December 2013
2. A component-based performance comparison of four hypervisors,Written by: Jinho Hwang; Sai Zeng; Wu, F.Y.; Wood, T,2013 IFIP/IEEE International Symposium on Integrated Network Management (IM 2013)
3. Performance Comparison of Commercial VMM: ESXI, XEN, HYPER-V & KVM,Varun Kumar Manik,Deepak Arora,2016 3rd International Conference on Computing for Sustainable Global Development (INDIACom)
4. https://wiki.xen.org/wiki/Virtualization_Spectrum#Full_virtualization
5. https://technet.microsoft.com/library/hh831531.aspx6. https://pubs.vmware.com/vsphere-4-esx-vcenter/index.jsp?
topic=/com.vmware.vsphere.bsa.doc_40/vc_admin_guide/migrating_virtual_machines/c_nx_xd_considerations.html
7. http://wiki.prgmr.com/mediawiki/index.php/Chapter_12:_HVM:_Beyond_Paravirtualization
8. http://www.stratoscale.com/blog/compute/hypervisor-comparison-of-io-virtualization-models/
9. http://www-archive.xenproject.org/files/Marketing/HowDoesXenWork.pdf