It is worth mentioning that due to the design decisions around disk I/O in KVMs, for good performance (which itself is subjective) to be realised, you *really really* need a good SSD based SLOG in the zpool.
My personal favourite is the HGST SSD800MH.B The reasoning here is that every write operation inside a KVM is synchronous to the zpool. A zpool that is not optimised to cater for synchronous writes will not yield very good results. - Dave > On 31 Mar 2016, at 11:09 AM, Ian Collins <[email protected]> wrote: > > On 03/31/16 13:32, Humberto Ramirez wrote: >> >> Is anybody familiar with any recent perfomance comparison of KVM vs VMWare >> ESX? Or perhaps someone who had been running ESX and migrated to SmartOS on >> the same hardware? >> >> I found an old a paper from 2012 on which ESX outperformed KVM by 30% and >> sometimes for as much as 45%, of course 4 years is an eternity and that KVM >> system was not running SmartOS nor ZFS. >> > > I haven't compared like for like for a while, but of all the systems I've > moved from VmWare to SmartOS KVM, there has been a performance gain, > especially I/O. > > The KVM overhead for Ubuntu tested by building gcc compared to bare metal is > about 8-9%. For windows I've seen SmartOS KVM return better than bare metal > numbers (ZFS?) running MS-SQL benchmarks. > > On like for like hardware, I can't see how VmWare+NAS can best KVM+ZFS. > > You should try a comparison with your expected workload. > > -- > Ian. > ------------------------------------------- smartos-discuss Archives: https://www.listbox.com/member/archive/184463/=now RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00 Modify Your Subscription: https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb Powered by Listbox: http://www.listbox.com
