>
>
>
>
> Tuesday July
> 30, 1996 10:00
> pm PDT -
> Wednesday July
> 31, 1996 2:00
> am PDT
> News Server Downtime
>
> Hi Usenet Users,
>
> Last Friday we added another RAID system to
>our news architecture. We
> attempted to move information from the
>original RAID system to the new
> RAID system. Due to the vast amount of
>information that has to be moved
> and the fact that the original RAID system
>is already saturated, we could not
> move all the information originally
>targeted. To minimize large downtime we
> have to move information incrementally. We
>need to minimize downtime, not
> only to minimize the discomfort of users,
>but also to minimize the time it take
> our system to recover from the back load.
>That is, the longer the system is
> "down", the longer it takes to catch up on
>receiving articles from our
> upstream feeds, the longer it takes to
>expire articles, etc, which exhausts even
> more system resources.
>
> Given that, we will move the overview
>database from our original RAID
> system to the new RAID system. The move
>should take approximately 4
> hours and will be done on Tuesday, 7/30,
>from 10pm to 2am. After this
> move we will re-evaluate the load
>distribution to see what else needs to be
> moved.
>
> FYI, here are the current loads of each of
>the RAID systems:
>
> Original RAID:
>
> CPU NFS Net kb/s Disk kb/s Tape kb/s Cache
> in out read write read write age
> 100% 3750 576 3360 2729 232 0 0 0
> 100% 3586 545 3503 2601 86 0 0 0
> 100% 3036 461 3238 2777 0 0 0 0
> 100% 3470 529 3190 2692 0 0 0 0
> 100% 3884 604 2861 2457 49 0 0 0
> 100% 3962 669 3325 2730 53 0 0 0
> 100% 3867 586 2934 2784 102 0 0 0
> 100% 3889 590 3157 3196 2 0 0 0
> 99% 3878 672 3293 2737 0 0 0 0
> 100% 3599 715 3448 3277 325 0 0 0
> 100% 3652 559 3458 3390 257 0 0 0
> 100% 4033 611 3152 2768 81 0 0 0
> 100% 3962 613 3540 2912 116 0 0 0
> 100% 3955 613 3519 3088 60 0 0 0
> 100% 4174 637 3310 3216 272 0 0 0
> 100% 4265 659 3136 2918 189 0 0 0
> 100% 3705 566 3546 3004 17 0 0 0
> 100% 4129 644 3636 2784 85 0 0 0
> 100% 4450 682 3565 2934 109 0 0 0
> CPU NFS Net kb/s Disk kb/s Tape kb/s Cache
>
> New RAID:
>
> CPU NFS Net kb/s Disk kb/s Tape kb/s Cache
> in out read write read write age
> 9% 272 42 397 245 0 0 0 15
> 14% 333 163 504 338 0 0 0 15
> 8% 253 38 314 142 0 0 0 15
> 13% 387 60 383 201 249 0 0 15
> 11% 574 87 492 334 0 0 0 15
> 15% 585 88 436 261 0 0 0 15
> 15% 563 153 386 225 146 0 0 15
> 12% 541 97 415 160 0 0 0 15
> 16% 464 70 465 193 0 0 0 15
> 14% 692 106 787 190 0 0 0 15
> 18% 703 107 1115 228 84 0 0 15
> 29% 606 92 755 254 0 0 0 15
> 22% 837 127 904 221 0 0 0 15
> 14% 671 118 523 174 0 0 0 15
> 16% 632 172 496 257 84 0 0 15
> 22% 665 101 927 193 0 0 0 15
> 20% 528 80 641 293 0 0 0 15
> 33% 657 101 1210 348 210 0 0 15
> 36% 550 85 975 236 2 0 0 15
> 28% 505 79 757 181 0 0 0 15
> CPU NFS Net kb/s Disk kb/s Tape kb/s Cache
>
> As you can see, we still have some tweaking
>to do to balance the load. After
> the move on Tuesday, hopefully we will be
>closer to a 60 - 40 split. We will
> keep you informed.
>
> Another FYI, once we have load balanced the
>RAIDs, we have equipment in
> house to upgrade each one of our news
>reader machines. This upgrade will
> decrease that load on each news reader
>machine and allow for more client
> connections per machine.
>
> We are trying to stay ahead of the curve,
>but sometimes get behind.
>
> News System Administrator
---= -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- Being a smartass beats being a dumb ass... you should try it sometime. -=-=-=-=-=-=-= -=-=-=-=-=-=-=-=- tbyars@earthlink.net