|
Charly, The method of determining whether disk arms are a problem hasn't really changed through the life of the AS/400-iSeries. Most performance characteristics follow a logarithmic curve. The knee of the curve, that place where performance ceases a gradual decline and starts to decline dramatically, is the threshold at which red flags should go up. This number has stayed constant at forty per cent through all of the advances in I/O speed, caching, and disk characteristics. It has to do with what is called a queuing multiplier. What is the likelihood that an I/O request can be fulfilled immediately when that request is issued to a disk drive? If the drive is busy then the request must be queued and reissued later. At forty per cent, this multiplier effect creates the disk arm equivalent of main storage thrashing. For disk activity, this occurs at about forty per cent disk arm utilization as viewed on the WRKDSKSTS display or in many of the performance reports. If someone thinks that the number of disk arms is a problem, then they can use this command to view per cent utilization during a sample interval. If all drives are close to (or above) this threshold, then you have a problem. If a single (or very few drives) are pegged, then the system could be balanced using the ASPBAL commands with the *USAGE parameter. It is really one of the easier performance characteristics to diagnose. Generally speaking, if you have maxed out activity on all drives, your only alternative to fix the problem involves hardware. In your list below, I believe that you may have omitted one of the more important hardware balancing factors. All of the disk and tape controllers (IOA's) are controlled by input/output processors (IOP's). These IOP's do have maximum throughput characteristics so putting multiple high-speed adapters on the same I/O processor can lead to the type of performance bottleneck you describe. If you had two RAID controllers and a high-speed tape controller, probably the most important hardware step you could take to optimize performance would be to have each adapter controlled by a different processor. In many cases, this can be accomplished without adding additional buses. I have found discussions of buses to be more of theoretical importance than practical importance (I confess to being fascinated by system throughput issues). The reason being, when you start to design a system to maximize throughput, all of a sudden you start to hear a "Ka-ching! Ka-ching!" coming from the direction of Rochester. Certainly on a system with multiple buses, high throughput devices should be spread out; but in most cases the idea of adding buses purely for performance reasons is financially out of reach for most companies. On the newer hardware, your disk enclosures tend to come with their own buses. For instance the forty-five unit disk expansion towers will allow for three RAID controllers in each enclosure. Generally you would fill the unit and have six RAID sets. Few installations could afford to not fully utilize that resource. This becomes particularly important when you consider that the addition of a chassis comes with monthly maintenance charges. A second factor which could be added to your list would be the number of high-speed loops (HSL). The 8xx hardware attaches the disk expansion towers via HSL and multiple towers can be daisy-chained in a single loop. If I recall correctly, we can also attach some of the high-speed tape drives via HSL also. Because these loops do have maximum throughput limits, it would be beneficial to install as many loops as possible and to avoid putting tape drives and disk enclosures on the same loop. Even though the HSL cables are expensive, additional loops can be added without adding buses or enclosures which would drive up monthly maintenance costs. Regards, Andy Nolen-Parkhouse > On Behalf Of Charly Jones > Subject: RE: disk arms (was RE: Tips for user ASP) > > > Chris + Scott -- > > How do you know that your disk arms are the problem? What > tools do you use? Which reports (or screens) do you look at? > Would you be willing to talk about how many disk arms are > attached to what disk controllers and whether you have > separate disk pools. I would like to continue this > discussion if anyone is interested. > > Actually the bus is sometimes important as well. If you > have a fast tape drive on the same bus with disk drives, the > disk activity can reduce the maximum throughput of the tape > drive. I can't help thinking that tape operations > would similarly affect disk activity. So to get the full > picture, is anyone willing to post the following? > > 1) How many system buses you have. > 2) How many disk controllers you have. > 3) The distribution of disk controllers on your buses. > 3) What other high speed devices share a bus with a disk controller. > 4) How many disk arms are attached to each disk controller. > 5) How many disk controllers + arms are in each disk ASP. > 6) Whether you think you have a disk arm problem. > 7) How you can tell you have a disk arm problem. > > Thanks a bunch, > > -- Charly
As an Amazon Associate we earn from qualifying purchases.
This mailing list archive is Copyright 1997-2024 by midrange.com and David Gibbs as a compilation work. Use of the archive is restricted to research of a business or technical nature. Any other uses are prohibited. Full details are available on our policy page. If you have questions about this, please contact [javascript protected email address].
Operating expenses for this site are earned using the Amazon Associate program and Google Adsense.