Persistent Data Metric (PDM) stats are sometimes required to investigate performance issues (particularly CPU, memory and worker count) when the default hourly snapshots aren't granular enough. Specifically, the client and server byte statistics are very useful for troubleshooting.

Here’s how to interpret the PDM stats.

Using HTTP Client Bytes as an example,  here's how to interpret the data. Each value represents the total number of bytes the client sent for the duration in the bracket (). This value is represented in seconds.

(60,60) means there are 60 values and each represents the total client bytes for the duration of 60 sec:
http:client-bytes~hourly@Mon, 09 Nov 2009 11:07:00 UTC[06](60, 60): 390520974 451725151 545296712 751610075 638740761 496456969 429192906 460856158 277671355 344564596 355194142 349898751 640510157 492376091 410721337 318598967 507521956 474781501 534399346 419637051 355057209 489503221 360452542 395076804 409367624 494832934 494574126 347353852 345836879 464010117 436319912 509994301 353067445 393147824 355668118 489472074 570461057 416004963 557006481 412800946 452737023 381296582 360433940 630404796 552660656 390923232 489270132 430768389 913018132 476931942 510630897 540868802 676553812 495913658 545698709 555765563 385108521 499442366 385614715 473668441

(24,3600) means there are 24 values and each value is the total client bytes for 1 hour. Eq; 25255334755 bytes was transmitted between SG and the client in 1 hour
http:client-bytes~daily@Mon, 09 Nov 2009 11:00:00 UTC[10](24, 3600): 25255334755 28780528133 25936179740 28284239329 29533805195 24972925285 20225390922 14181406174 6841495302 1812781365 1079300077 1537681715 2306930251 1186293616 1999503059 7459580371 18277119394 21641552994 24240829431 29032972145 29050111824 33382386688 31045866965 28115300188

21600/3600=6 hours which means 4 values = 1 day. There are 28 values which means you see the 1 week stats here
http:client-bytes~weekly@Mon, 09 Nov 2009 06:00:00 UTC[04](28, 21600): 150062045930 159091359578 35731186890 70343485350 159146788616 154791801830 34351220182 67781981459 149307343616 155666074029 27410248420 67507959452 150556455633 142343922988 24322985986 71508876259 149316810315 158007150841 46216029508 89752068367 171303153976 192743063836 45814236044 90475375636 158096368923 157733068604 27759594884 74804878865

There are 31 values below and each represents the total bytes transmitted in a day (86400/60/60 = 24 hours)
http:client-bytes~monthly@Mon, 09 Nov 2009 00:00:00 UTC[07](31, 86400): 51248228771 58712529065 53654362210 54536592156 173714704087 368228073225 352704896270 406307990168 397636829792 413431459949 379299239689 349394356045 324285786310 328223451683 431173125611 492953551770 432965295897 356268326155 303471042370 327477790374 358669513571 394748900265 450677744624 450528454253 400508900017 418633295978 400165647524 384731324059 425048866923 499612522223 434064408047

Each value represents total client bytes for 1 week (604800/60/60/24=7days). 52 represents the number of weeks in a year
http:client-bytes~yearly@Sun, 08 Nov 2009 00:00:00 UTC[46](52, 604800): 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1954 324201015 2469153177 0 1939794983 242114621817 303230041816 357140161134 387913895799 1806783447908 2718760971057 2624278613256 2979229010977

Using the values above, we can then correlate the stats to that of the CPU or memory to determine if the increase/decrease in CPU/memory is traffic related.

What about the numbers in [square] brackets?  The numbers in the square brackets are a time index indicator.  Each entry (hourly, daily, weekly, monthly, and yearly) have their own unique index.  For example, the hourly index is updated at the top of each hour.  Using the hourly example above, the number in brackets is [06].  The far right side of the list of numbers is 00.  Counting from the farthest right, 06 is the 7th number.  This means that when the value was 495913658, the time was 11:00:00 UTC on Mon, 09 Nov 2009.  This is handy for creating graphs using the actual time of day to correlate with an anomaly in the statistics.