Changes between Version 17 and Version 18 of KVM_vs_Xen
- Timestamp:
- Jul 14, 2010, 5:37:07 PM (14 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
KVM_vs_Xen
v17 v18 6 6 7 7 ---- 8 = 【Xen 和 KVM 8 = 【Xen 和 KVM 雙方辯論會】 = 9 9 * Xen: 我效能很好勒 10 10 * KVM: 我都不用 Patch Kernel 勒 11 (VMware 表示:你們把我放在哪裡 \_/)11 (VMware 表示:你們把我放在哪裡 @@a ) 12 12 13 13 ---- … … 41 41 || '''VMmark''' || || || || 42 42 || '''dbench''' || || || || 43 43 44 * __'''Memory 效能比較:'''__ 44 45 || || '''Xen-PV''' || '''Xen-HVN''' || '''KVM''' || 45 46 || 還沒想到 || || || || 47 46 48 * __'''Graphics 效能比較:'''__ 47 49 || || '''Xen-PV''' || '''Xen-HVN''' || '''KVM''' || 48 50 || '''CUDA''' || || || || 51 49 52 * __'''I/O (Disk) 效能比較:'''__ 50 53 * 註:Xen 可選擇 file、[http://wiki.xensource.com/xenwiki/blktap tap:aio] 和 xen4.0 最新 [http://wiki.xensource.com/xenwiki/blktap2 tap:tapdisk:aio] 來設定 Guest disk 51 54 * 註:KVM 可使用 [http://www.linux-kvm.org/page/Virtio virtio] 來增加 I/O 效能 (KVM >= 60, kernel >=2.6.25) 52 55 * [http://www.scribd.com/doc/14683287/Virtualized-Block-Device-IO-Performance Virtualized Block Device I/O Performance] 53 >> 樓上那位大大的 Disk I/O 測試總結:[[BR]](1) Xen is actually pretty damn good[[BR]] (2) KVM performance is sorry with full device emulation (expected)[[BR]] (3) KVM+VirtIO is better, but nowhere near Xen Paravirt 56 > 樓上那位大大的 Disk I/O 測試總結: 57 >>(1) Xen is actually pretty damn good[[BR]] (2) KVM performance is sorry with full device emulation (expected)[[BR]] (3) KVM+VirtIO is better, but nowhere near Xen Paravirt 54 58 55 59 || || '''Xen-PV''' || '''Xen-HVN''' || '''KVM''' || 56 60 || '''bonnie++''' || || || || 57 61 || '''dbench''' || || || || 62 58 63 * __'''I/O (Network) 效能比較:'''__ 64 * [http://www.linux-kvm.com/content/kvm-developers-achieve-native-network-io-performance KVM developers achieve native network IO performance] 65 > 此位大大的測試,使用 KVM PCI-passthrough 配合 Intel VT-d 可達到 Host 相同的網路效能 66 59 67 || || '''Xen-PV''' || '''Xen-HVN''' || '''KVM''' || 60 68 || '''tbench''' || || || || 69 61 70 62 71 ----