# ClearFog CX LX2 - SATA bandwidth is "shared" with network, reducing performance **URL:** \ **Category:** NXP LX2160\ **Created:** [October 25, 2022, 8:50am UTC](https://community.solid-run.com/t/clearfog-cx-lx2-sata-bandwidth-is-shared-with-network-reducing-performance/502 "2022-10-25T08:50:52Z")\ **Posts on this page:** 12\ **Page:** 1
**Author:** ![robert\_sv](https://avatars.discourse-cdn.com/v4/letter/r/e0b2c6/32.png) [@robert\_sv](https://community.solid-run.com/u/robert_sv)\ **Post date:** [October 25, 2022, 8:50am UTC](https://community.solid-run.com/t/clearfog-cx-lx2-sata-bandwidth-is-shared-with-network-reducing-performance/502/1 "2022-10-25T08:50:52Z")
Hello, I’ve noticed that bandwidth, or perhaps some kind of kernel resource, is shared between the SATA ports and the SFP+ ports. Running two iperf3 servers for two of the SFP+ ports, two parallel iperf3 clients from a separate ClearFog board gets close to a consistent 9.9Gb/s on each port. However, writing to a SATA SSD (e.g. simply with `cp`) at the same time decreases performance to about 8-9Gb/s on each SFP+ port. Writing to multiple SATA disks has only a minor additional effect. SATA write performance does not appear to be affected. There is plenty of CPU to spare (i.e. 1 core for each iperf3 instance, and 1 core for writing to disk). Is this as expected? Or is there perhaps some kind of kernel setting I can adjust to resolve this issue? ---
**Author:** ![jnettlet](https://dub1.discourse-cdn.com/flex005/user_avatar/community.solid-run.com/jnettlet/32/62_2.png) [@jnettlet](https://community.solid-run.com/u/jnettlet)\ **Post date:** [October 25, 2022, 8:58am UTC](https://community.solid-run.com/t/clearfog-cx-lx2-sata-bandwidth-is-shared-with-network-reducing-performance/502/2 "2022-10-25T08:58:45Z")
The only physically shared hardware between them are the SMMU and the DDR. Most likely the performance decrease is being caused by interrupt contention on CPU core 0. ---
**Author:** ![robert\_sv](https://avatars.discourse-cdn.com/v4/letter/r/e0b2c6/32.png) [@robert\_sv](https://community.solid-run.com/u/robert_sv)\ **Post date:** [October 26, 2022, 2:32am UTC](https://community.solid-run.com/t/clearfog-cx-lx2-sata-bandwidth-is-shared-with-network-reducing-performance/502/3 "2022-10-26T02:32:58Z")
From my understanding, interrupts may not be the case of the issue. Running two instances of iperf3 and a disk-writing benchmark I wrote, there doesn’t appear to be a high load on the ksoftirqd processes. ```auto top - 12:14:33 up 11 min, 4 users, load average: 13.82, 4.85, 1.98 Tasks: 266 total, 10 running, 256 sleeping, 0 stopped, 0 zombie %Cpu(s): 8.8 us, 42.3 sy, 0.0 ni, 8.0 id, 27.4 wa, 2.2 hi, 11.3 si, 0.0 st MiB Mem : 29969.5 total, 206.9 free, 12871.8 used, 16890.7 buff/cache MiB Swap: 0.0 total, 0.0 free, 0.0 used. 16742.5 avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 2826 user 20 0 12.3g 9.9g 7988 R 531.2 34.0 4:28.44 DiskWriteBenchm 1663 user 20 0 5248 2060 1588 R 93.8 0.0 2:47.95 iperf3 1672 user 20 0 5248 2140 1668 R 93.8 0.0 2:43.90 iperf3 187 root 20 0 0 0 0 S 50.0 0.0 0:35.18 kswapd0 2245 root 20 0 0 0 0 R 25.0 0.0 0:09.32 kworker/u32:0+flush-8:0 154 root 20 0 0 0 0 D 18.8 0.0 0:09.36 kworker/u32:1+flush-8:48 202 root 20 0 0 0 0 R 18.8 0.0 0:09.59 kworker/u32:4+flush-8:16 206 root 20 0 0 0 0 D 12.5 0.0 0:09.16 kworker/u32:6+flush-8:32 2939 user 20 0 6116 2892 2272 R 12.5 0.0 0:00.03 top 10 root 20 0 0 0 0 R 6.2 0.0 0:00.23 rcu_preempt 55 root 20 0 0 0 0 S 6.2 0.0 0:02.28 ksoftirqd/9 327 root 20 0 0 0 0 R 6.2 0.0 0:00.21 kworker/2:2-events 1 root 20 0 166796 6164 3332 S 0.0 0.0 0:07.15 systemd ``` Looking at /proc/interrupts, however, ahci-qoriq[3200000.sata], ahci-qoriq[3210000.sata], ahci-qoriq[3220000.sata] and ahci-qoriq[3230000.sata] were experiencing about 500 interrupts/second all on core 0. The network interrupts (dpio.0 to dpio.15 I believe?) were distributed across all cores. Regardless the SATA interrupts seemed like a lot, so I installed irqbalance, and confirmed that the SATA interrupts were being rebalanced. However, the issue of reduced network performance when using the SATA disks remained. Is there anything else I should look at? ---
**Author:** ![jnettlet](https://dub1.discourse-cdn.com/flex005/user_avatar/community.solid-run.com/jnettlet/32/62_2.png) [@jnettlet](https://community.solid-run.com/u/jnettlet)\ **Post date:** [October 26, 2022, 4:12am UTC](https://community.solid-run.com/t/clearfog-cx-lx2-sata-bandwidth-is-shared-with-network-reducing-performance/502/4 "2022-10-26T04:12:28Z")
Can you test and collect output from `mpstat -P ALL 1` this should provide a better per core usage of resources. It is also possible that you are hitting a bandwidth limitation within the CCN of the LX2160a. You could test this by temporarily running your system bus overclocked to 800Mhz, the default is 700. ---
**Author:** ![robert\_sv](https://avatars.discourse-cdn.com/v4/letter/r/e0b2c6/32.png) [@robert\_sv](https://community.solid-run.com/u/robert_sv)\ **Post date:** [October 26, 2022, 4:58am UTC](https://community.solid-run.com/t/clearfog-cx-lx2-sata-bandwidth-is-shared-with-network-reducing-performance/502/5 "2022-10-26T04:58:31Z")
Running both iperf3 instances along with the disk benchmark, I get the following output from `mpstat -P ALL 1`. It doesn’t look bottlenecked to me, but I could be missing something. ```auto 00:33:53 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %gnice %idle 00:33:54 all 6.56 0.00 42.61 31.71 2.41 11.44 0.00 0.00 0.00 5.28 00:33:54 0 0.00 0.00 1.04 0.00 13.54 66.67 0.00 0.00 0.00 18.75 00:33:54 1 0.98 0.00 89.22 0.00 2.94 3.92 0.00 0.00 0.00 2.94 00:33:54 2 3.16 0.00 54.74 29.47 1.05 2.11 0.00 0.00 0.00 9.47 00:33:54 3 3.16 0.00 38.95 51.58 0.00 6.32 0.00 0.00 0.00 0.00 00:33:54 4 6.74 0.00 42.70 38.20 2.25 4.49 0.00 0.00 0.00 5.62 00:33:54 5 12.22 0.00 40.00 40.00 1.11 5.56 0.00 0.00 0.00 1.11 00:33:54 6 7.78 0.00 38.89 45.56 0.00 0.00 0.00 0.00 0.00 7.78 00:33:54 7 7.61 0.00 32.61 58.70 0.00 0.00 0.00 0.00 0.00 1.09 00:33:54 8 9.89 0.00 43.96 41.76 1.10 0.00 0.00 0.00 0.00 3.30 00:33:54 9 7.78 0.00 41.11 48.89 1.11 1.11 0.00 0.00 0.00 0.00 00:33:54 10 3.30 0.00 46.15 41.76 0.00 8.79 0.00 0.00 0.00 0.00 00:33:54 11 5.32 0.00 51.06 32.98 0.00 10.64 0.00 0.00 0.00 0.00 00:33:54 12 6.32 0.00 57.89 29.47 2.11 1.05 0.00 0.00 0.00 3.16 00:33:54 13 27.96 0.00 23.66 39.78 1.08 0.00 0.00 0.00 0.00 7.53 00:33:54 14 1.04 0.00 3.12 0.00 10.42 65.62 0.00 0.00 0.00 19.79 00:33:54 15 3.12 0.00 72.92 16.67 1.04 3.12 0.00 0.00 0.00 3.12 00:33:54 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %gnice %idle 00:33:55 all 7.08 0.00 41.35 31.45 2.29 11.27 0.00 0.00 0.00 6.55 00:33:55 0 0.00 0.00 7.29 4.17 11.46 63.54 0.00 0.00 0.00 13.54 00:33:55 1 3.09 0.00 62.89 27.84 1.03 3.09 0.00 0.00 0.00 2.06 00:33:55 2 4.17 0.00 46.88 23.96 0.00 1.04 0.00 0.00 0.00 23.96 00:33:55 3 6.38 0.00 37.23 45.74 0.00 3.19 0.00 0.00 0.00 7.45 00:33:55 4 7.45 0.00 46.81 39.36 0.00 6.38 0.00 0.00 0.00 0.00 00:33:55 5 5.38 0.00 60.22 26.88 1.08 6.45 0.00 0.00 0.00 0.00 00:33:55 6 9.28 0.00 32.99 50.52 0.00 5.15 0.00 0.00 0.00 2.06 00:33:55 7 6.38 0.00 40.43 47.87 1.06 3.19 0.00 0.00 0.00 1.06 00:33:55 8 16.33 0.00 37.76 28.57 3.06 1.02 0.00 0.00 0.00 13.27 00:33:55 9 7.45 0.00 32.98 52.13 0.00 1.06 0.00 0.00 0.00 6.38 00:33:55 10 9.38 0.00 48.96 32.29 1.04 6.25 0.00 0.00 0.00 2.08 00:33:55 11 8.51 0.00 38.30 41.49 1.06 9.57 0.00 0.00 0.00 1.06 00:33:55 12 20.21 0.00 38.30 31.91 1.06 0.00 0.00 0.00 0.00 8.51 00:33:55 13 8.60 0.00 36.56 53.76 0.00 0.00 0.00 0.00 0.00 1.08 00:33:55 14 0.00 0.00 2.08 0.00 12.50 66.67 0.00 0.00 0.00 18.75 00:33:55 15 1.00 0.00 90.00 0.00 3.00 3.00 0.00 0.00 0.00 3.00 00:33:55 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %gnice %idle 00:33:56 all 7.35 0.00 42.36 31.17 2.25 10.99 0.00 0.00 0.00 5.89 00:33:56 0 0.00 0.00 4.17 1.04 12.50 63.54 0.00 0.00 0.00 18.75 00:33:56 1 7.69 0.00 53.85 35.16 1.10 1.10 0.00 0.00 0.00 1.10 00:33:56 2 37.37 0.00 20.20 25.25 1.01 1.01 0.00 0.00 0.00 15.15 00:33:56 3 7.69 0.00 39.56 49.45 1.10 2.20 0.00 0.00 0.00 0.00 00:33:56 4 8.42 0.00 52.63 28.42 1.05 7.37 0.00 0.00 0.00 2.11 00:33:56 5 5.43 0.00 45.65 43.48 0.00 3.26 0.00 0.00 0.00 2.17 00:33:56 6 5.38 0.00 49.46 35.48 1.08 5.38 0.00 0.00 0.00 3.23 00:33:56 7 5.26 0.00 51.58 33.68 0.00 4.21 0.00 0.00 0.00 5.26 00:33:56 8 6.32 0.00 49.47 38.95 1.05 1.05 0.00 0.00 0.00 3.16 00:33:56 9 5.26 0.00 40.00 49.47 0.00 4.21 0.00 0.00 0.00 1.05 00:33:56 10 4.12 0.00 67.01 22.68 1.03 5.15 0.00 0.00 0.00 0.00 00:33:56 11 6.38 0.00 40.43 44.68 0.00 5.32 0.00 0.00 0.00 3.19 00:33:56 12 4.12 0.00 64.95 11.34 4.12 2.06 0.00 0.00 0.00 13.40 00:33:56 13 5.32 0.00 45.74 38.30 1.06 1.06 0.00 0.00 0.00 8.51 00:33:56 14 1.03 0.00 7.22 1.03 10.31 64.95 0.00 0.00 0.00 15.46 00:33:56 15 6.67 0.00 47.78 44.44 0.00 1.11 0.00 0.00 0.00 0.00 ``` How do I overclock the system bus? ---
**Author:** ![jnettlet](https://dub1.discourse-cdn.com/flex005/user_avatar/community.solid-run.com/jnettlet/32/62_2.png) [@jnettlet](https://community.solid-run.com/u/jnettlet)\ **Post date:** [October 26, 2022, 5:31am UTC](https://community.solid-run.com/t/clearfog-cx-lx2-sata-bandwidth-is-shared-with-network-reducing-performance/502/6 "2022-10-26T05:31:44Z")
Looking at mpstat, I don’t believe you are hitting a hardware limitation but instead a kernel limitation. Between that and your top output I see you have many of the cores sitting around in iowait. This is most likely due to the kernel using the multi-queue accesses for the drive accesses. You also have very limited memory available as everything has been allocated to buffer/cache…you can also see this in the CPU utilization with kswapd0 and the kworkers. Overall it looks like you just need to tune your kernel configuration for this specific workload. ---
**Author:** ![robert\_sv](https://avatars.discourse-cdn.com/v4/letter/r/e0b2c6/32.png) [@robert\_sv](https://community.solid-run.com/u/robert_sv)\ **Post date:** [October 26, 2022, 6:59am UTC](https://community.solid-run.com/t/clearfog-cx-lx2-sata-bandwidth-is-shared-with-network-reducing-performance/502/7 "2022-10-26T06:59:29Z")
Running the iperf3 instances alone does not appear to affect iowait or buffer/cache, but I can imagine how their utilization being high could put strain on the kernel and affect networking. Would a high buffer/cache value not indicate sufficient memory availability, given that Linux can drop it from the cache at a moment’s notice if another process needs it? Or perhaps I’m misunderstandings things, and the large size of buffer/cache means that there’s some kind of memory pressure. Which kernel settings should I look at for this workload? ---
**Author:** ![jnettlet](https://dub1.discourse-cdn.com/flex005/user_avatar/community.solid-run.com/jnettlet/32/62_2.png) [@jnettlet](https://community.solid-run.com/u/jnettlet)\ **Post date:** [October 26, 2022, 7:04am UTC](https://community.solid-run.com/t/clearfog-cx-lx2-sata-bandwidth-is-shared-with-network-reducing-performance/502/8 "2022-10-26T07:04:53Z")
the buffers are most likely being created by your disk benchmark. The kernel can unmap and remap them but that adds overhead to the network stack which could limit the throughput. This is something the Google’s MGLRU that was merged in the 6.1 kernel could help with. You can also try different IO policy schedulers for your drives to see if it makes a difference. ---
**Author:** ![robert\_sv](https://avatars.discourse-cdn.com/v4/letter/r/e0b2c6/32.png) [@robert\_sv](https://community.solid-run.com/u/robert_sv)\ **Post date:** [November 15, 2022, 1:49am UTC](https://community.solid-run.com/t/clearfog-cx-lx2-sata-bandwidth-is-shared-with-network-reducing-performance/502/9 "2022-11-15T01:49:23Z")
I’ve installed 64GB of RAM, but there is no change in behavior. buff/cache seems to just consume as much memory as it possibly can. Changing IO policy schedulers also had no noticeable impact. Do you have any other suggestions for how this may be worked around? Here is the output of `mpstat -P ALL 1`: ```auto Linux 5.4.47-00007-g9c7b74fdbb19 (clearfog1) 11/15/22 _aarch64_ (16 CPU) 12:45:41 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %gnice %idle 12:45:42 all 10.22 0.00 26.12 6.37 2.21 8.45 0.00 0.00 0.00 46.62 12:45:42 0 9.00 0.00 28.00 16.00 2.00 5.00 0.00 0.00 0.00 40.00 12:45:42 1 8.00 0.00 35.00 23.00 1.00 0.00 0.00 0.00 0.00 33.00 12:45:42 2 7.14 0.00 32.65 0.00 1.02 0.00 0.00 0.00 0.00 59.18 12:45:42 3 7.22 0.00 36.08 0.00 1.03 0.00 0.00 0.00 0.00 55.67 12:45:42 4 3.96 0.00 23.76 11.88 0.99 1.98 0.00 0.00 0.00 57.43 12:45:42 5 7.84 0.00 39.22 16.67 0.98 3.92 0.00 0.00 0.00 31.37 12:45:42 6 25.77 0.00 14.43 0.00 1.03 0.00 0.00 0.00 0.00 58.76 12:45:42 7 18.37 0.00 40.82 0.00 1.02 1.02 0.00 0.00 0.00 38.78 12:45:42 8 17.35 0.00 17.35 19.39 3.06 4.08 0.00 0.00 0.00 38.78 12:45:42 9 28.43 0.00 18.63 3.92 0.98 2.94 0.00 0.00 0.00 45.10 12:45:42 10 7.92 0.00 40.59 0.00 0.99 0.00 0.00 0.00 0.00 50.50 12:45:42 11 10.00 0.00 35.00 0.00 0.00 0.00 0.00 0.00 0.00 55.00 12:45:42 12 1.04 0.00 4.17 0.00 10.42 58.33 0.00 0.00 0.00 26.04 12:45:42 13 1.05 0.00 3.16 0.00 10.53 58.95 0.00 0.00 0.00 26.32 12:45:42 14 5.94 0.00 27.72 4.95 0.99 0.00 0.00 0.00 0.00 60.40 12:45:42 15 4.04 0.00 19.19 5.05 0.00 3.03 0.00 0.00 0.00 68.69 12:45:42 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %gnice %idle 12:45:43 all 12.59 0.00 24.86 5.40 1.53 8.52 0.00 0.00 0.00 47.11 12:45:43 0 9.09 0.00 25.25 20.20 1.01 5.05 0.00 0.00 0.00 39.39 12:45:43 1 11.11 0.00 39.39 0.00 0.00 1.01 0.00 0.00 0.00 48.48 12:45:43 2 31.31 0.00 17.17 0.00 0.00 1.01 0.00 0.00 0.00 50.51 12:45:43 3 15.00 0.00 26.00 0.00 0.00 1.00 0.00 0.00 0.00 58.00 12:45:43 4 19.19 0.00 21.21 0.00 0.00 0.00 0.00 0.00 0.00 59.60 12:45:43 5 30.93 0.00 18.56 9.28 0.00 0.00 0.00 0.00 0.00 41.24 12:45:43 6 5.15 0.00 16.49 0.00 0.00 1.03 0.00 0.00 0.00 77.32 12:45:43 7 20.20 0.00 31.31 14.14 0.00 2.02 0.00 0.00 0.00 32.32 12:45:43 8 7.14 0.00 44.90 0.00 0.00 0.00 0.00 0.00 0.00 47.96 12:45:43 9 5.00 0.00 12.00 6.00 0.00 2.00 0.00 0.00 0.00 75.00 12:45:43 10 16.33 0.00 28.57 17.35 0.00 5.10 0.00 0.00 0.00 32.65 12:45:43 11 14.14 0.00 51.52 11.11 1.01 1.01 0.00 0.00 0.00 21.21 12:45:43 12 0.00 0.00 5.26 1.05 11.58 60.00 0.00 0.00 0.00 22.11 12:45:43 13 0.00 0.00 4.17 0.00 11.46 58.33 0.00 0.00 0.00 26.04 12:45:43 14 13.27 0.00 33.67 7.14 0.00 2.04 0.00 0.00 0.00 43.88 12:45:43 15 3.00 0.00 21.00 0.00 0.00 0.00 0.00 0.00 0.00 76.00 12:45:43 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %gnice %idle 12:45:44 all 10.54 0.00 23.24 3.43 2.03 8.00 0.00 0.00 0.00 52.76 12:45:44 0 7.07 0.00 36.36 0.00 6.06 3.03 0.00 0.00 0.00 47.47 12:45:44 1 2.02 0.00 11.11 19.19 1.01 0.00 0.00 0.00 0.00 66.67 12:45:44 2 30.61 0.00 17.35 0.00 0.00 0.00 0.00 0.00 0.00 52.04 12:45:44 3 15.15 0.00 35.35 0.00 1.01 0.00 0.00 0.00 0.00 48.48 12:45:44 4 11.46 0.00 18.75 0.00 0.00 0.00 0.00 0.00 0.00 69.79 12:45:44 5 10.89 0.00 20.79 0.00 0.99 0.00 0.00 0.00 0.00 67.33 12:45:44 6 10.10 0.00 21.21 0.00 1.01 0.00 0.00 0.00 0.00 67.68 12:45:44 7 5.94 0.00 41.58 0.00 0.00 0.99 0.00 0.00 0.00 51.49 12:45:44 8 12.12 0.00 39.39 0.00 1.01 1.01 0.00 0.00 0.00 46.46 12:45:44 9 8.16 0.00 24.49 0.00 1.02 0.00 0.00 0.00 0.00 66.33 12:45:44 10 11.34 0.00 21.65 30.93 0.00 6.19 0.00 0.00 0.00 29.90 12:45:44 11 17.00 0.00 28.00 5.00 0.00 4.00 0.00 0.00 0.00 46.00 12:45:44 12 2.06 0.00 6.19 0.00 9.28 55.67 0.00 0.00 0.00 26.80 12:45:44 13 1.04 0.00 8.33 0.00 10.42 58.33 0.00 0.00 0.00 21.88 12:45:44 14 18.56 0.00 31.96 0.00 0.00 0.00 0.00 0.00 0.00 49.48 12:45:44 15 5.05 0.00 8.08 0.00 1.01 1.01 0.00 0.00 0.00 84.85 ``` And here is the output to `top`: ```auto top - 12:45:05 up 18 min, 3 users, load average: 7.01, 3.46, 1.42 Tasks: 266 total, 2 running, 264 sleeping, 0 stopped, 0 zombie %Cpu(s): 11.9 us, 19.9 sy, 0.0 ni, 34.6 id, 22.4 wa, 1.4 hi, 9.8 si, 0.0 st MiB Mem : 62113.3 total, 384.9 free, 8810.2 used, 52918.2 buff/cache MiB Swap: 0.0 total, 0.0 free, 0.0 used. 52694.5 avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 3823 user 20 0 5882772 3.0g 20252 S 218.8 4.9 7:11.86 NetAndDiskTest 3822 user 20 0 5890968 3.1g 20108 S 187.5 5.1 8:25.19 NetAndDiskTest 154 root 20 0 0 0 0 D 25.0 0.0 0:08.48 kworker/u32:1+flush-8:32 5114 root 20 0 0 0 0 D 25.0 0.0 0:10.15 kworker/u32:0+flush-8:16 202 root 20 0 0 0 0 D 18.8 0.0 0:06.67 kworker/u32:4+flush-8:48 206 root 20 0 0 0 0 D 18.8 0.0 0:12.55 kworker/u32:6+flush-8:0 70 root 20 0 0 0 0 R 12.5 0.0 0:10.08 ksoftirqd/12 75 root 20 0 0 0 0 S 6.2 0.0 0:09.88 ksoftirqd/13 805 root 20 0 0 0 0 S 6.2 0.0 0:00.06 jbd2/sdd1-8 5838 user 20 0 6116 2820 2256 R 6.2 0.0 0:00.03 top 1 root 20 0 168840 10128 7232 S 0.0 0.0 0:14.80 systemd ``` ---
**Author:** ![robert\_sv](https://avatars.discourse-cdn.com/v4/letter/r/e0b2c6/32.png) [@robert\_sv](https://community.solid-run.com/u/robert_sv)\ **Post date:** [November 22, 2022, 6:39am UTC](https://community.solid-run.com/t/clearfog-cx-lx2-sata-bandwidth-is-shared-with-network-reducing-performance/502/10 "2022-11-22T06:39:28Z")
@jnettlet Sorry for the bump, but do you have any further theories as to what the issue could be, considering that 64GB of memory did not help? ---
**Author:** ![jnettlet](https://dub1.discourse-cdn.com/flex005/user_avatar/community.solid-run.com/jnettlet/32/62_2.png) [@jnettlet](https://community.solid-run.com/u/jnettlet)\ **Post date:** [November 22, 2022, 6:45am UTC](https://community.solid-run.com/t/clearfog-cx-lx2-sata-bandwidth-is-shared-with-network-reducing-performance/502/11 "2022-11-22T06:45:50Z")
I didn’t expect more memory to solve the issue. The issue isn’t a lack of memory but the overhead in linux of processing the memory, This is what Google has work on with the MGLRU patch. Generally you are going to need to profile the kernel and see what is bottlenecking the system. I would also recommend testing with another filesystem benchmark like fio or stress-ng that allows more customizable workloads and see if that makes a difference. ---
**Author:** ![benoit.donnette](https://avatars.discourse-cdn.com/v4/letter/b/8c91f0/32.png) [@benoit.donnette](https://community.solid-run.com/u/benoit.donnette)\ **Post date:** [November 23, 2022, 6:23am UTC](https://community.solid-run.com/t/clearfog-cx-lx2-sata-bandwidth-is-shared-with-network-reducing-performance/502/12 "2022-11-23T06:23:54Z")
Intel once had a kernel patch + userland tool called “latencytop”, I wonder if they can be ported to ARM. And they’d need to have been maintained too (8+ years old IIRC).