13 Replies Latest reply on Jan 17, 2011 5:10 AM by

    Introducing DuckieHo, Intel SSD Communities' first Moderator!

    SSDelightful

      Everyone,

       

      I wanted to take a quick second to introduce our communities site (and Intel's first ever!) moderator, DuckieHo. He's a well-seasoned SSD expert and is here to help answer your solid-state storage related questions. DuckieHo will be meeting periodically with myself and Intel engineers to make sure he's got the best possible information and that users' concerns will be heard. If you've got any questions/concerns, please voice them here, but if not, come say what's up to DuckieHo!

       

      -Scott, Intel Corporation

        • 1. Re: Introducing DuckieHo, Intel SSD Communities' first Moderator!
          DuckieHo

          Thanks Scott.  I am not an Intel employee but I am here to assist everyone with their SSD questions and guide discussions.  I am pushing Scott and Intel to disclose more information about their SSDs.  Hopefully, this information will help us understand the benefits, the workings, and optimal setup of SSDs.

           

           

          A little background about myself...

          • Financial System Developer by day, technology/science enthusiast by night
          • Moderator at Overclock.netand EVGA.com
          • Former EVGA support contractor
          • BS in Computer Engineering

           

          If you have any questions or concerns that the community cannot address, I hope to escalate Intel to be resolved.

           

          Thanks again!

          • 2. Re: Introducing DuckieHo, Intel SSD Communities' first Moderator!
            redux

            Hi DuckieHo, congratulations. I look forward to seeing more of your posts.

             

            I have a question for you to get you started

             

            My understanding is that when the OS sends a TRIM command if the command is not successfully completed Win 7 will stop sending TRIM commands. If that is right what happens if you have a HDD and then create an image and install the image on a SSD? Will Win 7 restart the TRIM commands?

            • 3. Re: Introducing DuckieHo, Intel SSD Communities' first Moderator!
              DuckieHo

              The OS does not know if the TRIM command is completed by the SSD anyways.  It does not know if the driver supports TRIM nor does the SSD respond that TRIM was successful.  Windows sends enables the TRIM command when ever it detects a storage device with 0RPM.  This is what "fsutil behavior query DisableDeleteNotify" tells you.  This is why the only way to know if TRIM is actually working is through testing.... hitting it with massive volume of random data with TRIM and benching shortly after to see how the drive performs.

               

               

              We hope to have a technical TRIM-related article up soon.  You will be interested so stay tuned!

              • 4. Re: Introducing DuckieHo, Intel SSD Communities' first Moderator!
                mistermokkori

                redux wrote:

                 

                My understanding is that when the OS sends a TRIM command if the command is not successfully completed Win 7 will stop sending TRIM commands. If that is right what happens if you have a HDD and then create an image and install the image on a SSD? Will Win 7 restart the TRIM commands?

                 

                by default, windows 7 always sends trim commands regardless of what kind of drive you install it on.  disabledeletenotify is always 0 by default unless you change it.

                • 5. Re: Introducing DuckieHo, Intel SSD Communities' first Moderator!
                  PeterUK

                  Hi their DuckieHo.

                   

                  I was wondering do Intel SSD have garbage collection thing (Intel might have it called something else and this is not to do with TRIM or the OS this is what the SSD does) that other SSD have? From what I can work out all SSD do realtime garbage collection that erases blocks where data has been deleted to add new data and this works on a single SSD or in a RAID. Then there is idle/background garbage collection that free up more blocks ahead of time when the SSD is idle.

                   

                  So can you confirm what and if garbage collection is part of Intel SSD? And if idle/background garbage collection is part of or can be supported on Intel's SSD?

                   

                  Thanks

                  • 6. Re: Introducing DuckieHo, Intel SSD Communities' first Moderator!
                    DuckieHo

                    Intel SSDs do not have idle time garbage collection (ITGC) or background garbage collection (BGC)  Some SSD controllers like Indilinx Barefoot, SandForce SF-1200, or Samsung S3C29RBB01-YK40 do have this feature.  This is an internal feature of the drive.  The controller determines when it is idle for long enough and then does a file system compare against what is actually valid on the drive.  It then cleans up the dirty pages.  The actually implementation of garbage varies between controllers though.  The downside to idle/background garbage collection is write amplification as it may perform unnecessary writes.

                     

                    While Intel SSDs do not idle/background garbage collection, they have are very resilent to dirty pages.  I believe this is what you are refering to when you say realtime garbage collection.  Then impact of deletes varies between controllers and has to do with how the drive decides manages data.

                     

                    If TRIM is working, you do not want/need idle/background garbage collection.  For Intel G2 drives not in an array, the SSD Toolbox Optimization is a manual garbage collection (and can be scheduled).  The scenario where you want/need idle/background garbage collection is SSD in RAID.

                    • 7. Re: Introducing DuckieHo, Intel SSD Communities' first Moderator!
                      PeterUK

                      I guess Intel sees garbage collection as an inefficient way of freeing up space then TRIM is in any SSD configuration when fully compatible.

                       

                      Thanks for answering I look forward to that TRIM-related article and if possible how Intel deals with dirty pages when TRIM is not used.

                      • 8. Re: Introducing DuckieHo, Intel SSD Communities' first Moderator!
                        redux

                        That appears to be correct. Windows 7 issues TRIM commands at the physical volume level regardless of whether you use a HDD or a SSD

                        • 9. Re: Introducing DuckieHo, Intel SSD Communities' first Moderator!
                          redux

                          From what I have observed the OS issues a TRIM command (of proportional size to a file deletion) as soon as it occurs. Temporary files also seem to trigger a TRIM command when they expire.

                           

                          Is it possible to give any information about how an Intel SSD deals with a TRIM command from the OS? Does it implement it immediately or does it store the command and execute it when the drive is not busy? Thanks. 

                          • 10. Re: Introducing DuckieHo, Intel SSD Communities' first Moderator!
                            mechbob

                            So what you are saying , if I understanding you correctly . If I wanted to set up a computer with two  SSD in RAID 0 , I would have to use Western Digital Because they use Advanced Wear - Leveling , with a combination of dynamic and static wear - leveling algorithms. So the drives don't degrade in a RAID enviroment. My Next question is when is Intel going to solve this problem. Because there alot of us that would like to be able to RAID or Intel SSDs??

                            • 11. Re: Introducing DuckieHo, Intel SSD Communities' first Moderator!
                              redux

                              The “only” real performance benefits of Raid 0 is higher sequential throughput and scaling at higher queue depths.  This type of performance improvement capability is not much use for normal desktop users and raid 0 can actually reduce performance where it matters for most desktop users – latency.

                               

                              Enterprise use is another issue.

                               

                              Having used hard and hybrid raid 0 (2x X25-E) for some time for desktop use I saw no noticeable performance decrease when I switched to a single drive (X25-M). Boot up was quicker because the raid rom did not have to load, but even if you took that out the equation there was no difference. The reason I saw no noticeable drop in performance was because (with my usage patterns) I was not able to utilise the performance that was available.

                               

                              4K random read/ write IOP performance for a single SSD is already way above what most people will ever need. Saturating SATA II speeds is hard to do without placing a high load on the rest of the system. (i.e. the flood of data from a saturated SATA II has to typically be processed at some stage, so something other than the storage system is likely to become a bottleneck when that happens).

                               

                              My 2 cents anyway. If you can use the performance in raid 0, great, but if not it’s just not worth it.

                               

                              For the reasons above my hope for the G3 drives is lower latency, something that has not yet been revealed. Oh, the suspense.

                               

                              BTW, if you don’t write excessively you will not see much degradation with Intel drives in raid 0. People I know that use raid 0 in a desktop environment tend to run a secure erase every 6 months or so and restore the data afterwards from an image. Not idea but no big deal either to get optimum performance back.

                               

                              You could also run a pure software raid 0 if you wanted TRIM to work. Users on other forums have reported that it works but I haven't tried it myself yet.

                              • 12. Re: Introducing DuckieHo, Intel SSD Communities' first Moderator!
                                DuckieHo

                                redux has pretty much covered the primary points of why RAID0 does not benefit desktop work patterns.  

                                 

                                 

                                Robert, all modern SSDs employ wear-leveling.  If you want SSDs in RAID and maintain long-term write performance, you need a SSD that performs garbage collection.  However, like redux mentioned, Intel SSDs are very resilient to dirty pages and most users won't see noticable write performance degradation even after months of use.

                                • 13. Re: Introducing DuckieHo, Intel SSD Communities' first Moderator!

                                  When we meet intel g3 in supermarkets?