On Fri, Jan 19, 2018 at 10:31 AM, Ian Collins <[email protected]> wrote:
> On 01/20/2018 04:07 AM, Jim Wiggs wrote: > >> I've been told by quite a few folks that splitting the SSD between log >> and cache is "a bad idea" or "suboptimal" but frankly, I don't buy it. It >> may just be my personal experience, but for my use cases, I've been >> operating with limited resources and haven't been able to justify the >> expense of having three or more SSDs to do this. Since I've never needed a >> ZIL with more than 2 GB of space and the smallest SSDs you can buy are more >> than 10x that size, mirroring a pair of SSDs for the ZIL was a huge waste >> of space. I started doing this about 4 years ago when SSDs were much more >> expensive and I couldn't justify that waste, so I'd partition a 1-2 GB >> slice on each SSD and mirror them for my ZIL, and use the remaining space >> on both SSDs, un-mirrored, for cache. Again, in my experience, this has >> always resulted in better general performance than either adding only log >> or only cache. >> >> > You don't need mirrors, so you could have used a small write intensive SSD > for your log and a bigger general purpose one for your cache. > > For KVM workloads on SmartOS, you are better of spending your money on the > best log device(s) you can afford and skip the cache. Spend the money on > more RAM! I haven't looked recently but Joyent's recommended system builds > didn't have cache devices. > > > Cheers, > Ian. > As I mentioned in a previous reply, I was not aware that ZIL is now recoverable. My experience with ZFS for KVM workloads has been solely on Linux, using ZVOLs for my disk images. The word there was always *mirror your ZIL* or you risked losing the last ~5 seconds worth of writes if your ZIL device died. If you don't need to do this any more (or at least, not under Illumos/SmartOS), then obviously the calculation changes. My point is that if I have to mirror my ZIL, I'm forced to shell out for two devices instead of one. My ZIL is never going to need more space than is required to store the last ~5 seconds of write data, so even a 32 GB drive is going to go 90% unused. I wrote scripts back in the early days to monitor the output of "zpool iostat" and track usage of the ZIL. I NEVER saw a backlog of more than about 1300 MB in any of my ZILs, even under the heaviest write-intensive workloads I could put together. During normal day-to-day operation, it's significantly less than *100* MB. Hell, a lot of the time it's under 10! Given this fact, I'd argue that it makes perfect sense to partition out the other 28 GB of each of those SSDs and use that 56 GB of combined space to provide some L2ARC. Can you (or anyone else on this thread) show me *actual benchmark numbers* to *prove* that I've been hurting my overall performance over the years by making that choice? Remember, I've been operating based on the assumption that I *do* need to mirror my ZIL, which was absolutely true back when I started out. best, Jim > > ------------------------------------------- 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
