Mar-012 STORAGE MANAGEMENT by DINESH DATTANI The IBM Family of Data Facility Products Dinesh Dattani is a Senior Technical Analyst at C-I-L Incorporated in Toronto, Canada. This article examines IBM's offerings in the storage management arena and takes a look at some of the positive aspects of these products, along with its shortcomings. Until recently storage management was seldom discussed seriously. The issue was relegated to the back-burner while the technical gurus tackled the more important tasks of CPU tuning, performance management and capacity planning. This trend continued well into the late seventies. Then things began to change. The mainframe industry - both the manufacturers and users - saw that there were potential savings in optimizing disk usage, that effective storage management was an idea whose time had come. It is not difficult to see why. Faced with an average DASD growth rate of 40%, not to mention the escalating staffing costs, installations had to find ways to curb the explosion in disks. For software vendors there was money to be made in the disk management software market, and they were quick to jump on the bandwagon. Many of them flourished - among them Sterling Software (DMS/OS), Innovation Data Processing (FDR/COMPAKTOR), UCCEL (UCC-3), and Cambridge Systems (ASM/2) - to mention a few of the early entries into DASD management (UCC-3 and ASM/2 are now part of Computer Associates product line). Somewhat belatedly, IBM entered the market with its DF family of products. The DF Family IBM unveiled its Data Facility (DF) products gradually. In typical IBM fashion, there was evolution rather than revolution. First came Data Facility/Data Set Services (DF/DSS), the high speed dump/restore utility which replaced this function of the IEHDASDR utility; and Device Support Facility (DSF) which replaced the rest of the IEHDASDR functions. The IBM sort/merge product soon was renamed to DF/SORT and vast improvements were made in its efficiency and function over the previous SORT product. The Hierarchical Storage Manager (HSM) product was made available within this time frame (and later took the name of DF/HSM). Together, these products were supposed to cure all the DASD ills, addressing backup, recovery, automated storage management and efficiency in sorting. But is this so? Do the DF products answer all the storage management needs? Let's take a closer look. DF/DSS Data Facility/Data Set Services is a backup and restore utility. Its most notable feature is its speed, an important consideration when talking of disk backups. It also performs many other DASD related utility functions, including: o Release of idle space o Compress of PDS's o Volume and dataset copy operations o Device conversion o Volume defragmentation o Print functions DF/HSM The Hierarchical Storage Manager is an automated DASD management system that addresses inactive datasets. Based on user criteria, DF/HSM moves (migrates) inactive datasets to Level-1 storage, usually on-line DASD, where the data is stored in compacted format and thus reducing real disk space. Further inactivity of the data results in the movement from Level-1 storage to Level-2 storage, usually off-line tape. Provided the systems programmer has implemented DF/HSM correctly, migration of inactive data to Level-1 and Level-2 is completely automatic. Reference to a migrated dataset will result in automatic recall of the dataset, the only intervention being that of the computer operator mounting a tape should the dataset reside on Level-2. DF/HSM includes other automated functions such as automatic backup of changed datasets to tape, recycling of both backup and migration tapes once the number of valid datasets on a tape reaches a low threshold, and automatic expiration of both migrated and backup datasets based on user criteria. ISMF The Interactive Storage Management Facility (ISMF) is not a separate product, but is included with the Data Facility Product (DFP) component of MVS/XA. It uses the familiar ISPF panels and dialogs to perform on-line data management functions with interfaces to DF/DSS and DF/HSM. It provides a convenient umbrella over these products and users need not get bogged down with syntax. This also implies that ISPF is becoming involved with the overall integration of the DF family of products. DFP, DF/SORT, and RACF The Data Facility Product (DFP) is an integral component of the MVS and MVS/XA operating systems which includes the access methods and standard IBM utilities. The DF/SORT product, as its name implies, is the sort/merge utility. The Resource Access Control Facility (RACF) is IBM's security system. We can expect these three products to play a greater supportive role in IBM's overall storage management strategies. By this I mean that we can expect further integration of these products, and the above mentioned products, in an effort by IBM to provide a complete storage management picture. Significance of the DF Family Any installation evaluating disk management software must consider the DF family seriously. Not for what the products do now, but rather for what they might do in the (near) future. Why? Because IBM has proclaimed that the DF Family will play a leading role in its storage management philosophy. I do not know what final form this philosophy will take, nor do I know the implications in the long term. But I do know that IBM is driving home its point loud and clear. In particular, IBM has indicated that the DF products are strategic in its plans. At a minimum, one can take this to mean that the DF family will one day provide a total storage management solution, at least in the IBM sense. Moreover, advances in other areas like MVS/XA can be expected to enhance, rather than take away from, the DF products. Using the DF Family the system itself will perform the tasks of storage management, reducing staffing costs. This will be a boon to everyone, especially the storage administrators who will be relieved of many of the tedious and routine tasks involved with storage management. Integration of DF At the present, integration among the DF products is all but nonexistent. They are about as integrated as MVS is with DOS. The command syntax within the DF family is inconsistent, and functions overlap between products. But this could (and probably will) all change. The new releases of the products will bring them a step closer toward integration. In May, 1987, IBM announced Version 2 Release 3 of DFP, DF/DSS and DF/HSM. IBM has hinted that these announcements are merely the first phase of a three stage process. IBM has also hinted that the second stage will be unveiled about May, 1988. The new releases are seen as a major step toward product integration, ease of use and system managed storage. Some of the highlights of these new releases are discussed below. Volume Pooling Support DF/DSS and DF/HSM will be able to handle pools of disks as a single, logical entity. A welcome feature, no doubt. What is surprising is that it took IBM so long to deliver it. IBM has always advocated its clients to aim for centralized storage, where the goal is to pool volumes rather than allow individual groups to own volumes. Until now IBM had done little to provide volume pooling functions within its products. Volume Application Currently, ISMF only supports functions at the dataset level. The new releases will expand the functions to cover volume-level operations. That is, users will be able to see volume summaries, free space information, fragmentation index, and the like. A sure sign that IBM intends to enhance ISMF. DF/HSM Tape Copies With the 3480's, users will be able to make copies of DF/HSM-owned backup and migration tape volumes. A necessary feature for off-site storage and disaster recovery. DF/DSS and DF/HSM DF/DSS will be invoked from DF/HSM to perform full volume backups. This is a good example of product integration, DF/HSM is better suited to perform the task - it has a control dataset for recovery operations - and DF/DSS has the better backup processing. A single command will now restore a damaged volume to its current level. Previously, recovery of a single volume via DF/HSM was a manual process involving the restoration of the last full volume backup, and then applying the incremental backups. Deficiencies Within the DF Family In their current state, even considering the new releases, the DF products fall short of fully satisfying the customers. If one were to draw up a "wish list" of desirable features in the DF family, it would include many of the following items. Perhaps the most annoying aspect of DF/HSM has been its inability to duplex tapes under its control, both the migration and backup volumes. Duplexing is needed to create off-site copies for disaster recovery, and in the event of an I/O error on a tape. The new release of DF/HSM addresses this problem by introducing duplexing functions, but only for the 3480 tape cartridge system. It will be many years before every shop has 3480 drives, and this duplexing should be made available for 3420's. The DBRC function of IMS only works with backups created using its own image-copy utility. It will not work with backups created by DF/DSS or DF/HSM. An interface between DBRC and the DF family is desperately needed to achieve common ground among these systems. VSAM cluster allocation requires the specification of a specific volume serial number. The IDCAMS utility needs to be upgraded to allow the specification of an esoteric unit name or some other method of nonspecific volume request. This is necessary to implement the concept of volume pooling. IBM maintains that a system-managed storage environment does not require hardcopy reporting features. While there is some truth to this, the facility should be there for those who wish to produce listings and reports related to the storage environment. Speed-wise, DF/HSM is slow as a tortoise, and DF/DSS is as fast as the proverbial hare. This being the case, DF/HSM should invoke DF/DSS for all backup and restore operations. The new release has taken a step in the right direction by having DF/HSM invoke DF/DSS for full volume backups, but this is not enough. The idea should be extended to cover incremental backups and tape migrations as well. The stand-alone restore program will not restore from a compressed dump. This may or may not be a serious problem, but it should be addressed. The defragmentation function of DF/DSS should be extended to allow the rearranging of datasets on a volume in an order specified by the user. This will allow for easier DASD tuning in the area of dataset placement on a volume. Summary In the race toward system-managed, integrated, disk management software, IBM is considered a late starter. But it is running fast to catch up with the competition. And the speed with which it is running should be sufficient to convince anyone that it will be among the first at the finish line. /* 1844