With the VMAX there are some areas of concern. What is your cache hit ratio? What is your CMR time?
In my shop we are either on SSD or not for DB2. So for all production DB2 tables they are on SSD and not tiered. In fact, we do not use tiering from EMC for VMAX for the mainframe. We are using FAST VP for the open system files that share the VMAX with us. Do you use charge back so that those datasets/databases that land on SSD will be charged a higher rate than those on spinning disk? Since you can have mixed devices in a SMS Pool, have you looked to just putting some of your more sensitive DB2 tables on just SSD in the pool rather than tiering? I will see if I can find more notes. Lizette -----Original Message----- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of k Zaf Sent: Sunday, August 11, 2013 11:55 PM To: IBM-MAIN@LISTSERV.UA.EDU Subject: Re: SSD tiering benefits Thanks for your answer Lizette We are using various tools like IBM RMF & ASG TMONZOS in z/OS running SMS & HSM. We actually want to tier some very utilized DB2 v9.1 workload to SSD and keep all the rest running in spinning VMAX 15000 rpm disks. I am very interesting in forecasting any potential CPU benefit beside of course the better IO. How can we project this? On 7 August 2013 12:55, Lizette Koehler <stars...@mindspring.com> wrote: > What tools do you have? Do you have SAS SAS/MXG SAS/MICS, Omegamon > (Tivoli), DISK Magic, etc... > > What version of z/OS, DB2? > > What hardware is used? EMC (VMAX, other), HDS, IBM? > > RMF which creates the Type 70 records, should contain some of the > information you are looking for. Do you have experience in > Performance analysis? If you do, what tools do you normally use? > > What will you be using for tiering? Each hardware vendors have their own > tiering solutions. > > I think DISK magic (fee product) by Intellimagic can do what if analysis. > > My understanding is that Tiering is used to move less active data to > slower devices inside a storage array and put highly accessed data on > faster devices in the array. That tiering is used to smooth out the > hot spots in the array itself. Or move data from more expensive > storage (SSD) to less expensive (SATA) devices within the array. > > So is your issue hot spots in the array and DB2 requires better > performance (1 ms or less per transaction)? Are you mixing SSD to > SATA type devices and want to ensure high access data is on the fast device? > > What is the problem you are trying to solve with SSD drives? > > Do you use DFHSM and DFSMS to manage your storage? Do you have a > mixed DFSMS pool with both Spinning and SSD disks? Is the SMS pool a > pure DB2 table pool or share with non-VSAM? Are your DB2 Tables > currently on SSD or spinning disks? Or on mixed devices? Do you use > your Storage groups to direct your DB2 tables to SSD? > > Lizette > > -----Original Message----- > From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] > On Behalf Of K > Sent: Wednesday, August 07, 2013 1:28 AM > To: IBM-MAIN@LISTSERV.UA.EDU > Subject: SSD tiering benefits > > Dear Listers, > > We are thinking to implement storage tiering using some SSD volumes > for some of our very active DB2 tablespaces. Is there any way to > measure performance improvements for both CPU and IO utilization? Do > you have any experience on tiering with SSD in DB2 environment? > > Kind regards > Kostas > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, send > email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
