Printable View
I read somewhere that Intel will soon be able to pass TRIM commands thru to RAID arrays with their RST yaddayadda... can LSI pass TRIM commands thru yet? I'm comtemplating 2x 64GB C300/M4 on a 9240 in RAID0 :D
or should I not be concerned about no TRIM in RAID? (is the onboard GC in the C300/M4 good enough?)
TIA! :up:
been running 2 C300 64Gb in raid for about 6 months now the GC seems to be up to the task just fine. Have had no slow down in performance at all.
No TRIM or SMART passthrough for LSI in RAID yet.
Their SAS3/SATA4 (12Gbps) controllers will hopefully bring this
from speaking directly with the VP and GM of the LSI Accelerated Solutions Division (ASD) during AIS 2011 interview, it is on the roadmap, and with existing controllers. There is no official timeline. The key is the correct implementation. As testing here has shown, there are TRIM lags/hangs and other issues that can arise from TRIM. A big thing also is predictability of performance, and in massive usage flash environments, with tons of devices, the extra I/O traffic. latency concerns..etc.
and you dont want to schedule it to happen at certain times...that can lead to unpredictable performance. OR even worse, hangs and lags like the type weve seen demonstrated here.
In the data center predictability is the #1 concern. No deviation in performance is the goal. The introduction of TRIM brings in alot more problems than one might think jsut looking at the surface. Different write loads could lead to different levels of I/O overhead...etc...3
most would actually prefer NO TRIM with flat line (lower) performance, than a performance gain that results in any type of deviation from SOP.
runing 3 x intel X25 G2 80GB on RAID0 since months and still same amazing performance
yes that is very true! in RAID you arent going to see that much of a difference in RAID 0, even if you are doing moderate loads. Heavy loads, YES there will be a big difference, but your talking server loads. sure you wont be doing that.
with RAID 5/6 parity sets, etc, there will be more noticeable degradation *in my experience*
parity sets in general also complicate the I/O overhead for TRIM. many factors there...
I always really enjoy your input CT! :) :toast:Quote:
Originally Posted by Computurd
what would you think if, say yourself, were to contact/liaison with LSI & suggest they release TRIM support for desktop users only? - so TRIM for standalone SSDs & small RAID0/10 arrays (say ~4 SSDs max)? :shrug:
Desktop users that use LSI controllers will be a tiny drop in an ocean of Servers, I doubt TRIM will be high on the priority list.
LSI has bought OCZ who have bought Indilinx (Octane), so LSI will have direct access to how the controller operates or is operating or is going to operate.
No matter what happens there are going to be very interesting times ahead.
^^ LSI acquired SandForce.
yup i agree. Even if i begged they wouldnt do it :)Quote:
Desktop users that use LSI controllers will be a tiny drop in an ocean of Servers, I doubt TRIM will be high on the priority list.
alot of it is also validation, they would never release anything to the public that hadnt been fully validated, if it got out in the wild...