This is our first pass at a NetApp FAQ. We have a number of additional questions on the way, but wanted to make this available. Many of these questions are the result of questions we have received about previous posts. If you have any questions that are not answered here, please enter it at the bottom and we will try to get you an answer.

Thanks,
– Jesse

NetApp FAQ – Questions

1. How do I monitor latency for IO to a LUN?
2. Is it possible to see CPU utilization by individual CPU or core?
3. What is NetApp Maintenence Center?
4. What is A-SIS?
5. What is the granularity of NetApp deduplication?
6. Is NetApp deduplication inline or post-processed?
7. Is NetApp cache "dedup aware"?
8. What are Predictive Cache Statistics?
9. What are evil-twin files? What is zombie filesystem space?
10. I have read that minra should be set to on for application X. Is this correct?
11. How do I monitor the performance of a Performance Accelerator Module (PAM)?

NetApp FAQ – Answers

1. How do I monitor latency for IO to a LUN?

   The command is "lun stats". I use "lun stats -o -i 10" which provides extended statistics and reports every 10 seconds. Latency and queue length can be very powerful information when analyzing performance issues with LUNs.

2. Is it possible to see CPU utilization by individual CPU or core?

   Yes. The sysstat -m command will provide this information. This is a powerful command as it can reveal a performance issue due to a single saturated core.

   my_filer> sysstat -m 5
   ANY AVG CPU0 CPU1 CPU2 CPU3
   39% 17% 14% 12% 20% 23%
   44% 19% 13% 17% 22% 26%
   54% 21% 10% 13% 30% 30%
   56% 22% 12% 8% 31% 34%
   54% 21% 15% 14% 27% 30%
   my_filer>

3. What is NetApp Maintenence Center?

   NetApp Maintenence Center is a mechanism in Data ONTAP that attempts to verify a disk is truly failed before replacing it. It will only work if you have at least 2 spares for each drive type in your system. Each disk drive runs it's own firmware and like all software, it has bugs. When a drive appears to be failed, it can often be fixed by power cycling it or reloading the firmware. These are the types of activities Maintenence Center attempts when a drive fails. This prevents the drive from needing to be physically replaced. If a drive fails hard or fails more than once, it will replaced as expected.

4. What is A-SIS?

   A-SIS stands for Advanced Single Instance Storage. It was the NetApp name for deduplication before deduplication became the official buzz word.

5. What is the granularity of NetApp deduplication?

   NetApp deduplication operates at the 4KB block level.

6. Is NetApp deduplication inline or post-processed?

   Deduplication is a scheduled operation and does not happen inline.

7. Is NetApp cache "dedup aware"?

   No, the primary memory cache is not deduplication aware. The Performance Accelerator Modules (PAM) are dedup aware. If you have a block that is referenced by 10 files and all 10 files access the block, there will be 10 copies of the block in cache. The PAM is dedup aware and only needs to store a single copy of that block.

   For more information on how NetApp deduplication works with cache, check out this article.

8. What are Predictive Cache Statistics?

   Predictive Cache Statistics (PCS) are a mechanism to allow a system to predict the benefit of additional memory. Additional cache can be added by adding PAM cards to a NetApp system and PCS can predict the effectiveness of that cache before it is added.

   This article explains PCS in more detail.

9. What are evil-twin files? What is zombie filesystem space?

   Traditional filesystems can take a long time to do things like delete files. NetApp uses the zombie filesystem to make the deletion process appear instantaneous. My understanding is the file is moved into the zombie filesystem and then removed asynchronously. You can observe this on a filer if you create a large file and then remove it. The remove command will appear instantaneous. Sysstat will show CPU load and disk IO as the file is removed asynchronously in the background.

   You can find more detailed patent information here.

10. I have read that minra should be set to on for application X. Is this correct?

    In the past, the NetApp read ahead algorithms were less than perfect and at times could negatively impact performance. NetApp documentation often recommended setting minra to on. As of 7.2.x, the read ahead algorithms have been significantly improved and minra should be set to off. Of course, there could be some pathological use cases that work better with it on, but that is no longer the general case.

11. How do I monitor the performance of a Performance Accelerator Module (PAM)?

    The stats command can provide details on the performance of the PAM in a controller.

    ntap> stats show -i 5 -p flexscale-access
     Cache                                               Reads       Writes      Disk Reads
     Usage    Hit   Meta   Miss Hit Evict Inval Insert Chain Blocks Chain Blocks  Replaced
         %     /s     /s     /s   %    /s    /s     /s    /s     /s    /s     /s        /s
        96   7936     20   3098  71     0     8     38  3952   7879     0     38      3952
        96   8231      0   3049  72    89  6240   3055  4088   8174    47   3055      4088
        96   7852      0   3003  72    85     0   3097  3897   7791    48   3097      3897
        96   7933      0   3103  71    89  6233   3072  3938   7872    48   3072      3938