Hi Ron, I agree with everything that you say about cu tiering, except for the fact that a migration tier is no longer necessary. CU tiering is an exciting new storage opportunity for ILM. I'd like to discuss this topic with you the next time that you attend SHARE or the Technical University. I've discussed this topic with your colleagues from HDS and also those from EMC. Let me go into more detail on some of the questions that were raised.
- I don't recall (pun intended) specific individuals that I've discussed tape/disk with, but HDS presentations on Tiering at SHARE show data moving to ML2 for archive after it has gone through the L0 - Ln disk tiers. These presentations also highlight the value of software / hardware tiering on a slide or two. Industry charts that show cost comparisons of cost/GB of storage show the clear cost value of tape. - I didn't mean to say that cu tiering is only for small environments. I was communicating that I believe that only small environment could eliminate a migration tier. CU tiering is of value for all environments. There is a finite amount of data that can be online to z/OS because of the UCB constraints. To eliminate the migration tier, you have to uncompress and return all migration data to online UCBs. That may be an option for smaller environments, but I don't see value in having uncompressed, archived data sitting on online disk, and large environment physically can't do that because of the limit to the amount of data that can be online. - DFSMShsm Transitioning is not a 'kludge' but rather a strength. Strengths of cu tiering 1) transparent 2) works on open data 3) no host MIPS. Strengths of Transitions 1) Data set level 2) Business policy based 3) Works across CUs. Weakness of cu tiering 1) movement done on heatmap with no understanding of data's business value. Weakness of transitions 1) data must be quiesced. Cu transitining weaknesses: I reorg a DB2 object. The data that had been fine-tuned to the correct tier is now scrambled across multiple tiers and until the cu relearns the correct tiers, I suffer subpar performance. Also, in the presentations that I've seen, CU tiering is appropriate for data that can be learned, like database data. It is not good for data like Batch data. (Once again, look at the HDS presentations on tiering). - The example that I have used for combining the two technologies that HDS has included in their Tiering presentation: 3 Tiers: T0 is SSD and Enterprise disk. T1 is Enterprise and SAS. T3 is Migration. Newly allocated data goes to T0. Cu tiering moves the data between SSD and Enterprise based on heat map. After a policy-based amount of time, the data has diminishing business value and HSM transitions it to T1. Data remains on the lower cost T1 while it is still active and the cu moves the data between Enterprise and SAS based on heat map. After the data goes inactive and should be archived, HSM migrates it to the migration tier. The migration tier can be all ML1, ML1/ML2, ML2 Virtual with all disk, ML2 Virtual with a combo of disk and tape, or all tape... whatever is best for each client's environment, as each has its strengths. - If you reference my presentations on tiering, I have been a proponent of eliminating multiple migration tiers for years. I have been recommending that customers use CU tiering for online tiering and that they don't migrate data until it really goes inactive, and then send it to a single migration tier. Until recently, I ML2 was the best choice because you get the compression for free on the tape cu (virtual or real). In this quarter, HSM is shipping support for the new z compression engine. That provides very high compression ratios for data on ML1, without using MIPS for compression. So, that now makes ML1 attractive also, for those customers who want a tapeless environment, like those on this thread. - There is a clear cost value to tape. If a client can afford to have all of their data uncompressed on online disk and don't have to worry about the UCB constraints, then more power to them. But, I suspect that most clients are still looking to keep their storage costs to a minimum. Our strategy is to provide all the options so that clients can select the ILM strategy that best needs of their data. Integrating the strengths of CU Tiering with Software Tiering provides the best of both worlds. Glenn ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
