Power Global

Power Global

A central meeting place for IBM Power. Connect, ask questions, share ideas, and explore the full spectrum of Power technologies across workloads, industries, and use cases.


#TechXchangePresenter
#Power

 View Only

  • 1.  Power 10 to Power 11 comparison

    Posted Tue May 05, 2026 07:58 AM

    We have two Power 10s and one Power 11.  The two Power 10s are in one data center with disk served by a FS7300.  The Power 11 is in a different data center with it's own FS7300.

    The 9105-41B has four LPARs of IBM i.  One of those LPARs supports five Domino servers.  Each of those five Domino servers is clustered over to an LPAR on the 9105-42A and each of those five Domino servers is clustered over to a LPAR on the 9824-22A.

    A dedicated offline Domino compact on one Domino server took 5:15, with one database alone taking 2:21 on the 9105-41B.

    That same process took 4:12 on the 9824-22A with that one database taking 2:02.

    05/02/2026 06:33:04   Recovery Manager: Restart Recovery complete. (0/0 databases needed full/partial recovery)

    05/02/2026 08:18:12   Informational, NIFNSF is enabled in database quotestubarchive.nsf.
    05/02/2026 08:18:12   Compacting quotestubarchive.nsf (Group Dekko Quote Stubs Archive),  -c -i -d
    05/02/2026 10:20:39   Recovery Manager: Assigning new DBIID for /GDDATA2/NOTES/DATA/quotestubarchive.nsf (need new backup for media recovery).
    05/02/2026 10:20:40   Compacted  quotestubarchive.nsf, 3328K bytes recovered (<1%),  -c -i -d

    05/02/2026 10:45:34   Database compactor process shutdown 

    Full disclosure:  There is a little age difference between the two FS7300s.  Both NVMe.  The one serving up the Power 10's might be '3 drives and the Power 11's might be '4 drives.



    ------------------------------
    Robert Berendt IBMChampion
    Business Systems Analyst, Lead
    Dekko
    Fort Wayne
    ------------------------------


  • 2.  RE: Power 10 to Power 11 comparison

    Posted Wed May 06, 2026 02:10 AM

    Robert.

    The faster result on the 9824‑22A is almost certainly due to storage/host throughput rather than Domino itself. A " dedicated offline copy‑style" compact ((-c) rewrites the entire NSF and assigns a new DBIID. 
    That makes the job largely a sequential read/write/metadata operation Your 9824‑22A + its FS7300 **likely with newer FCM4 drives and/or faster links** is delivering more sustained bandwidth and lower latency than the Power10 + the older FS7300 (likely FCM3) so the same work finishes 
    20-25% faster...
    What the log tells us
    NIFNSF enabled: view indexes live in the NSF, not separate .NDX. Does not materially change copy‑style compact time.
    "Compacting ... -c -i -d": -c is copy‑style; it creates a new file and copies all documents, then swaps files. -d discards view indexes (they'll be rebuilt at next open). -i ignores errors.
    "Assigning new DBIID ... (need new backup...)": confirms copy‑style ran. A full backup is required afterward.
    "3328K bytes recovered (<1%)": very little space reclaimed; the runtime was spent rewriting, not cleaning up free space.
    "Restart Recovery complete (0/0 ...)": no recovery work-nothing abnormal.
    Why the 9824‑22A/FS7300 is faster
    Storage generation: FS7300 with FCM4 NVMe drives has better latency and higher sequential bandwidth than FCM3. Even the same model FS7300 can differ notably by drive gen and count.
    Drive count/pool width: more drives (and better striping) → higher aggregate throughput for big sequential copies like -c compact.
    Host I/O path: newer 9824 (Power11) adapters/PCIe Gen5, 64G FC vs 32G FC, more paths, and better queue depth can all raise effective bandwidth.
    CPU helps a little: Domino compact is mostly I/O bound, but Power11's single‑thread performance and faster memory/PCIe can shave time too.
    What to check/measure Storage
    Drive generation and count: Are the P11 LUNs on FCM4 and the P10 LUNs on FCM3? How many drives in each pool/MDisk? Pool extent size and RAID type?
    Controller/port utilization and latency on both FS7300s (use GUI/IBM Storage Insights):
          Host port throughput (MB/s) and IOPS during compact                           Read/write latency (ms) and queue depths
          Node/volume ownership (avoid non‑preferred node access)
          
    Compression/dedup: Confirm settings are identical; different settings affect CPU on the array and latency.
    Host fabric
          HBA link speed and count (32G vs 64G FC) zoning and multipathing policy ensure paths are

     
     
     
    Atte.,
    + de 30 años asegurando continuidad operativa en entornos IBM
     
    Nelson Gaete
    Especialista en Infraestructura Crítica IBM
    Power, Storage & SAN
    ngaete@pure-ti.cl
    +5627566725 | +56942905244
    Pure-TI SpA – Partner Tecnológico en Soluciones IBM
    Merced 838, Santiago, Metropolitana

          





Related Content