Miklos, Figuring out how busy all the channel processors are on the Host is easy. It is reported in the Type 73 RMF record for the LPAR and the CEC. This does not tell you how busy the channel is. A FICON Express Four channel maxes out at approx 14,000 normal IOPS, but the FICON channel itself will handle more than double that IO rate. The channel is not 100% busy when the FICON MP on the host runs out of steam.
Also realize that this does not tell you how busy the channel processors are on the Front End Director boards connecting the storage to the channel. There is not a one to one correspondence because the storage typically has a very different architecture and processing cost. For instance IBM storage has a single microprocessor on each board which is shared by four channels, while HDS has four micro-processors that are shared by all four channels on the Board. RMF has no idea how busy these MP are in IBM and HDS, and there is no way to estimate it. This becomes even more complex and confusing when you have channel fan-in, meaning multiple CEC connect to one storage port through a switch, or multiple CEC connected to one board. The only way to measure channel busy behavior is to understand channel path busy using RMF and the MP measurement provided by the storage vendor. For volume busy you are going to get some very unexpected results if you run two or more LPARs, or PAV/HypePAV. The traditional measure for Device busy is (Connect + Disconnect)/ Interval, but because Multiple Device Allegiance and Parallel Access Volumes allow concurrent IO to the volume you can easily end up with a volume greater than 100% busy. Because volume busy has little to do with queuing for MDA and PAV what is the purpose of a Volume busy measure? Note that a LUN Busy measurement on other platforms is just as (in)accurate unless the Q-DEPTH is set to one. If you really want disk and not volume busy, then again there is nothing in RMF, and you have turn to the Storage vendor's internal performance numbers for the Array Groups. Of course the nice charts will be important for Capacity Plan trends, but I'd be sure your colleagues understand how to apply the information they are looking at. I still remember showing my manager a volume running at greater than 100% busy on a 7980-3 because of Logical Device Allegiance and two CEC access, and while I thought it was great he thought it was a problem. Ron > -----Original Message----- > From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf Of > Miklos Szigetvari > Sent: Wednesday, September 15, 2010 3:40 AM > To: [email protected] > Subject: Re: [IBM-MAIN] Channel types > > Hi > On 9/15/2010 12:24 PM, Ron Hawkins wrote: > > Miklos, > > > > By "the disk" I assume you mean the volume. How can you figure out how busy > > a volume is? > > > Not a volume, the complete system, all the DISK channel's > (if we can figure out this) > > RMF does not report how busy a FICON channel is, only the FICON Channel MP. > > > > What is the tool that reports FCP Channel busy and LUN busy on other > > platforms. > I don't know if any tool can report this. > > My colleagues want to get four nice charts about CPU , STORAGE , DISK > and NET activity. > On the other side, it has some sense to ask how active was the complete > DISK or NETWORK system > in a period. > > > Ron > > > >> -----Original Message----- > >> From: IBM Mainframe Discussion List [mailto:[email protected]] On > > Behalf Of > >> Miklos Szigetvari > >> Sent: Wednesday, September 15, 2010 2:27 AM > >> To: [email protected] > >> Subject: Re: [IBM-MAIN] Channel types > >> > >> On 9/15/2010 10:57 AM, R.S. wrote: > >>> Miklos Szigetvari pisze: > >>>> Hi > >>>> > >>>> We would like to categorize the channels as in the Mainframe Concept > >>>> book > >>>> > > (http://publib.boulder.ibm.com/infocenter/zos/basics/index.jsp?topic=/com.ib > > m. > >> zos.znetwork/znetwork_74.htm) > >>>> CCW channel or Coupling channel or QDIO/OSA. > >>>> Can I make this from the channel type codes ? > >>> In fact you just did it. > >>> > >>> What do you want to achieve? > >>> I'm asking, because there are different point of view: > >>> CHPID type - as in HCD, i.e. CNC, FC, FCV, CTC, etc. A physical > >>> channel can be defined as one of allowable types. For ESCON you have 4 > >>> choices. > >>> Channel card type - i.e. FICON Express2, Express4, Express8 - all > >>> those cards support same set of CHPID types. > >>> Channel media - related to Channel card type. > >>> CPC - machine model sometimes decides wht CHPID can be used, i.e. CFR > >>> cannot be defined on z9, even if the same card do support CFR. > >>> > >>> CCW channels: > >>> Bus&Tag, ESCON, FICON (no-Express, Express, Express2,4,8) > >>> CHPID: BL, BY, CTC, CNC, CVC, CBY, FC, FCP, FCP > >>> > >>> CF link channels: > >>> ISC(fiber optic 2-3 generations), ICB (copper, several incompatible > >>> generations, ICB, ICB-3, ICB-4), PSIFB (Infiniband), IC (internal, > >>> emulated). > >>> CHPID: CFP, CFR, CFS, CBP, CBR, CBS, ICP, ICR, ICS, CIB > >>> > >>> Network channels: > >>> Variuos OSA cards, IQD - emulated. > >>> CHPID: OSA, OSE, OSD, OSC, OSN, IQD (and OSM, OSX for z196) > >>> > >>> Remark: I intentionally excluded channel types available in Multiprise > >>> machines (EIO, DSD, ISD) and machines older than 9672 (IOC). > >>> > >>> > >>> HTH > >> Hi > >> > >> Thank you very much. > >> We want to say, from the RMF channel activity reports, that > >> the DISK was active x% and the NETWORK was active y% in the last > >> period, for a performance monitor tool from us. > >> (I feel it has not too many sense, but this tool exits on other > >> platforms ... ) > >> From the RMF I got the acronyms and channel types. > >> > >> ---------------------------------------------------------------------- > >> For IBM-MAIN subscribe / signoff / archive access instructions, > >> send email to [email protected] with the message: GET IBM-MAIN INFO > >> Search the archives at http://bama.ua.edu/archives/ibm-main.html > > ---------------------------------------------------------------------- > > For IBM-MAIN subscribe / signoff / archive access instructions, > > send email to [email protected] with the message: GET IBM-MAIN INFO > > Search the archives at http://bama.ua.edu/archives/ibm-main.html > > > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to [email protected] with the message: GET IBM-MAIN INFO > Search the archives at http://bama.ua.edu/archives/ibm-main.html ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
