System z: Optimizing DASD I/O Subsystem Performance

Historically there was a very simple synergy between the IBM S/370 Mainframe and its supporting disk I/O (DASD) subsystem, allowing for Mainframe host to physical and logical disk device (I.E. 3390) connectivity. The analysis and tuning of this I/O subsystem has always been and continues to be supported by the SMF Type 7n records via IBM RMF and the BMC CMF alternative. However, over the years, major advances in DASD subsystems and the System z Mainframe server have delivered many layers of technology resources (E.g. Cache, Memory, FICON Channels, RAID Storage, Proprietary Microcode, et al) and this has introduced complexities into highlighting DASD I/O subsystem performance problems.

The focus of technology based metrics (E.g. I/O Rate Response Time, I/O MB/S Bandwidth, et al) have also been complemented with more meaningful business focussed Service Level Agreements (SLA). Therefore today’s System z I/O Performance Analyst must gather and act upon proactive meaningful information from the ever-increasing amounts of performance data available. Put another way, too much data can deliver not enough information! As previously stated, it was forever thus, RMF and CMF have always collected the requisite performance data available and arguably no other data source is required (E.g. OMEGAMON/TMON/SYSVIEW Performance Monitor, SAS/MXG/MICS/WPS Performance Database). RMF/CMF is the ideal data source for thorough and timely System z I/O performance management, where intelligent analytics and expert knowledge are required to present this “Golden Record”.

However, today’s System z Support Teams need simple and timely presentation of the data, highlighting potential challenges, graphically presented for their Management, allowing for simple tracking of SLA agreements and technology changes (I.E. Software/Hardware Upgrades).

Additionally, Workload Manager (WLM) can control non-paging queued DASD I/O requests, based upon device busy conditional processing. Therefore the z/OS system can manage I/O priorities in a Sysplex, based on WLM service class goals. WLM dynamically adjusts the I/O priority based on service class goal performance and whether a DASD device can influence the overall performance objectives. For obvious reasons, this WLM function does not micro-manage I/O priorities, only changing a service class period’s I/O priority infrequently. WLM is deployed by many System z users to assist in the automated management of system resources (E.g. CPU, Memory, I/O, et al), based upon Service Level goals.

From a DASD subsystem technology viewpoint, there is no longer an obvious one-one direct connection between the Mainframe host and DASD device. An increasing number of technological advances, both microcode and hardware (E.g. Memory, Fibre Channel, Function Assist Processing, et al) have diminished the requirement for data access directly from the physical device. Put another way, in today’s world of System z servers with multiple cache level CPU chips (I.E. Relative Nest Intensity), massive and multiple processor memory resources (I.E. z13 @ 10 TB Memory), high bandwidth Fibre Channel (I.E. FICON, zHPF) subsystem and a hierarchy of DASD memory (I.E. SSD/Flash, Cache), it’s not uncommon to consider an I/O that requires physical device access as a problem! Finally and most importantly, from a DASD subsystem viewpoint, each of the recognized System z DASD providers, EMC (Symmetrix VMAX), HDS (VSP G1000) and IBM (DS8870) have highly proprietary DASD subsystems that provide z/OS plug compatibility, but deliver overall I/O performance using their own unique architecture and internal algorithms.

Of course, an over configured hardware environment will deliver a poor TCO, while an under configured environment will manifest in SLA issues and bad user experiences, where the middle-ground always delivers the optimal environment. Resource optimization always demands proactive day-to-day management, from an internal and indeed external communication viewpoint. With the highly proprietary design features of the IHV DASD subsystems, whether EMC, HDS or IBM, having the right information and identifying the precise problem, simplifies the communication process with the IHV. Such communication might highlight a resource under provision (E.g. Memory Capacity), a subsystem setting tweak requirement, either host or subsystem based, or indeed a hardware failure. In today’s world, these issues need to be fixed in minutes or hours, not days or weeks.

Therefore, where does today’s System z I/O Performance Analyst start to collect the required information to safeguard that their DASD subsystem is optimized, both from a capacity and performance viewpoint?

A simplistic viewpoint of an I/O health-check should consider the following:

  • Service Level Agreements (SLA): Are overall objectives being delivered or missed?
  • User Experience: Are users (customers) complaining of poor service or response times?
  • I/O Metric Performance: Are there obvious signs of abnormal performance statistics?

Several decades ago, an overall I/O health check might have been a periodic (E.g. Weekly or longer) activity, whereas today it’s undoubtedly a Business As Usual (BAU) and 24*7 activity. Therefore a fully automated solution is required, built upon the tried and tested System z performance fundamentals, namely RMF or CMF. The ideal solution will perform analytics based data reduction, presenting the right information, at the right time, allowing for intelligent business based communication, both internally, to customers and end users from an SLA viewpoint, and externally, with IHV DASD suppliers, safeguarding optimal performance and TCO.

EADM (Easy Analyze DASD Mainframe) is a solution from Technical Storage that performs automated performance analysis of the z/OS I/O subsystem, delivering predictive analytics for better storage capacity planning and performance measurement. The Technical Storage EADM architects have in excess of 40 years IBM Mainframe experience, specializing in the I/O subsystem, and so it’s no surprise that EADM delivers expert and timely knowledge via an easy-to-use solution.

EADM is an easy-to-install and easy-to-use plug-and-play solution that has no proprietary considerations, requiring no additional System z resource (E.g. CPU, Memory, DASD, et al) requirements. Installed on Microsoft server platforms, EADM is easily virtualized via VMware, Hyper-V, et al, requiring no target database for performance data storage. EADM performs a daily health check of the entire System z disk subsystem. EADM works around the clock, delivering customized and automatic user friendly GUI type reports. For today’s System z technician, the open and IP architecture base of EADM allows for secure remote access via Mobile, Tablet or Laptop devices, as and when required.

Operations and performance teams are alerted as soon as performance variances occur, typically in minutes, assisting in the identification of underlying root problems, causing changes in system behaviour. Incorporating intelligent and meaningful I/O performance indicators, with drill-down and zoom-in ability, storage technicians can determine if the problem is temporary, permanent, local or global. By simplifying the data reduction process (E.g. RMF/CMF data from numerous LPAR/Sysplex environments), EADM safeguards that the internal technical team can efficiently manage their ever increasingly complex and large DASD environment, for intelligent and timely communications with internal business teams and external suppliers alike.

EADM simplifies the System z I/O subsystem capacity and performance management process, delivering expert reports and timely historical analysis, for example:

  • Automatic daily (24 Hour) analysis of Sysplex wide workload (On-Line TP & Batch) I/O response times
  • Systematic intelligent alerts of early performance variances with exact occurrence time indicators
  • Identification of I/O performance hot-spots with DASD volume and data set level granularity
  • Performance trending at DFSMS Storage Group, Subsystem LCU and DASD volume level
  • DR (E.g. PPRC) simulations to prevent data loss and forecast Data Centre failover scenarios
  • I/O subsystem WLM indicators to determine exactly what impacts performance objectives
  • Full FICON channels and zHPF analysis, incorporating typical I/O throughput indicators
  • HyperPAV and associated LCU indicators to easily balance volumes, optimizing PAV alias allocation
  • Performance monitoring and balancing via intelligent LCU, SSID and I/O analytics
  • DASD capacity usage via DCOLLECT data, comparing assigned vs. allocated vs. actual disk utilization
  • EADM supports entry-level several LPAR and complex multiple CPC/LPAR System z configurations

A well provisioned and performing System z I/O subsystem is of vital importance for safeguarding today’s ever increasing storage requirements of mission critical business applications. A poorly performing I/O subsystem will generate unnecessary and extra CPU overhead, with potential and tangible TCO impact, in conjunction with potential business impact. Although the advances of the System z server and underlying DASD I/O subsystem can compensate for many application code or data placement issues, the fundamental concepts of analysing and tuning the I/O subsystem remain.

Therefore the savvy and proactive System z customer will safeguard that they find a solution to deliver optimal DASD I/O performance. Without doubt, such an analysis could be performed by a highly-skilled individual, but today’s 21st Century world demands a hybrid of technical and commercial skills. Therefore a solution that incorporates the diagnostic knowledge of the most highly trained technician, performs intelligent analytics on a plethora of Sysplex wide performance data sources and presents the information required, is one that will deliver benefit each and every day. EADM is an example of such a solution, delivering demonstrable System z TCO optimization benefits, while safeguarding a short-term ROI, with simple deployment and resource utilization attributes.

Revisiting The zSeries Mainframe Storage Hierarchy

Recommendation: The next time you perform a zSeries Mainframe server upgrade, consider adding Flash Express cards, for an extra 1.4-5.6 TB of memory speed storage. Similarly, the next time you perform a zSeries Mainframe DASD subsystem upgrade, consider adding as much SSD (flash memory) capability that you can afford and justify. Both upgrades will deliver significant performance and business benefits, arguably for minimal cost, when considered as a several year TCO investment.

Conceptually the zSeries Mainframe storage hierarchy has comprised the same layers for many decades, while performance and capacity attributes have dramatically increased over time. Although System/390 introduced the concept of Expanded Storage (I.E. Hiperspace, Data Space) in 1990 and there have been various implementations of SSD (E.g. StorageTek 4080), the ability to transparently implement significant capacity memory layers has only recently become possible.

Let’s not forget, the closer data is to that most precious and expensive of resources, namely CPU, the faster it will process. When revisiting the traditional storage hierarchy, we can now consider two new layers, namely Flash Express and Solid State Drive (SSD):

zSeries Storage Hierarchy

I have previously written about the Flash Express layer. Flash Express is a new memory layer within the zSeries Mainframe storage hierarchy, which can be considered as either a Solid State Drive (SSD) or Storage Class Memory (SCM) technology. Flash Express is integrated on PCI Express attached RAID 10 Cards, packaged as a two card pair, each with a 1.4 TB capacity per mirrored card pair. A maximum of 4 card pairs can be configured, delivering up to 5.6 TB of memory capacity, assigned to LPAR resources, just like main memory.

The simplest function to benefit from Flash Express memory would be SVC dump processing, substantially reducing dump capture time.

Flash Express can also be deployed to replace z/OS disk paging, substantially reducing the response time associated (I.E. ~5-20 μs vs. ~10 ms). The benefit for z/OS paging is not the replacement of memory paging, but replacing disk paging with Flash Express storage. Flash Express is suitable for workloads that can tolerate paging, but will not benefit workloads that cannot tolerate paging activity. The fundamental z/OS design for Flash Express memory will not completely remove any virtual storage constraints created by a paging spike, although a modicum of scalability relief is expected due to the faster I/O associated with Flash Express memory.

In conjunction with Flash Express, there were advancements in the Real Storage Management (RSM) function, including pageable 1MB Large Page Support. Large Pages (1MB) deliver benefit, with increased performance, decreasing the number of Translation Lookaside Buffer (TLB) misses that an application incurs, reducing time when converting virtual addresses into physical addresses and reduced real storage usage to maintain DAT structures. The use of Large Pages typically deliver Internal Throughput Rate (ITR) performance benefits of ~1% for IMS, ~3% for DB2 and ~5% for Java workloads.

Although SSD (flash) storage might have been selectively deployed in the zSeries Mainframe Data Centre for the last 5 years or so, the ever increasing requirement for increased Quality of Service (QoS) in terms of data availability and ultra-fast transaction response times dictate the increased usage of SSD architectures. Entire DASD subsystems can be built upon SSD technologies, or more likely, hybrid subsystems, containing both SSD and traditional HDD technologies. This storage subsystem evolution allows organizations to gain significant competitive advantages, delivering new services for existing and more importantly, new customers alike.

Using SSD disk subsystems, overcomes the limitations of traditional spinning hard disk drives. However, not every enterprise application needs this ultra-high performance; since flash storage still costs more than spinning drives for the same capacity, organizations must be mindful of expenditure and now much flash memory (SSD) they deploy; as always, flexibility is key.

Complete or hybrid SSD I/O subsystems deliver performance and economic advantages for your mission critical business environment:

  • Green Data Centre: ~25-60% energy reduction (flash memory vs. spinning disk)
  • Data Centre Space: ~20-40% smaller footprint (memory cards vs. Hard Disk Drives)
  • Optimal Performance: Consistent ~1-3 ms access (Hard Disk Drives @ ~10 ms)

The utopia is for a self-tuning disk subsystem, automatically redirecting I/O between SSD and HDD, based on file performance and overridden, as and when required, by storage policies. Whether EMC, HDS (HP OEM) or IBM, this self-tuning ability is evolving, while each disk vendor has their own implementation. However, whatever your choice of disk subsystem, the ability to incorporate SSD into your storage hierarchy, either full or partial is evident.

In conclusion, ~25 years ago, the zSeries Mainframe user benefitted from faster performance via System/390 Expanded Storage and disk subsystems with cache and DASD Fast Write memory buffers. The cost of such memory storage was a major consideration then, but with good I/O tuning disciplines, the savvy zSeries Mainframe user benefitted from these technology advancements. Flash Express and SSD deliver the potential to deliver increased performance, for a relatively low cost, and now is the time to embrace these technologies. Ignore the storage hierarchy at your peril and as I previously documented, optimal I/O performance always delivers significant benefit.

Flash Express – Back To The Future

It’s just not science fiction films that get rebooted or reimagined, the same thing happens in the technology world, although not always as obviously.  Over the years we have seen Star Trek, Star Wars, and numerous superhero stories be updated for current day audiences, some better than others, and similarly, flash memory or Solid State Drives (SSD) is not a new concept in the Mainframe world.  For me, I wonder whether the Back To The Future films might be rebooted, and if they do, hopefully it’s done well!  Anyway, I digress, so back to the Mainframe flash memory observations… 

Flash Express is a new feature for the zEC12 server, designed to help drive System availability and performance to even higher levels. 

Flash Express cards are delivered as a RAID 10 mirrored pair for superior resiliency and reliability.  In the unlikely event of device failure, Flash Express cards can also be concurrently replaced.  The cards are designed for superior wear levelling and have a long expected lifetime.  From a security perspective, Flash Express stored data remains protected.  Data is encrypted on the Flash Express adapter with 128-bit AES encryption.  Encryption keys are stored on smart cards that are plugged into the SE.  Removing the smart cards renders the data on the card inaccessible. 

In the first instance, IBM seems to be targeting Systems Paging (I.E. Auxiliary Storage Manager – ASM) and System Dumps (E.g. SVC, Standalone) as an introduction to this new level of storage hierarchy. 

Flash Express reduces latency for critical system paging that might otherwise impact the availability and performance.  For system paging flexibility and efficiency, Flash Express is a higher performance option when compared with traditional auxiliary storage (I.E. Disk).  z/OS uses both Flash Express and page data sets for auxiliary storage by paging data to the preferred storage medium first, based on response times, data set characteristics, and other parameters.  Wherever possible, the system will page first to Flash Express resulting in faster performance.  Especially for data intensive applications the use of Pageable Large Pages with Flash Express enables the transfer of large amounts of data at faster speeds, which can result in improved performance for DB2 analytic workloads.

NB. Because Flash Express is not persistent across IPL events, it cannot be used for Virtual I/O or PLPA data used in warm starts.  VIO and PLPA datasets must still be defined on DASD.

During diagnostic collection, as in SVC or standalone dumps, IBM states that systems can become sluggish effectively rendering key systems unavailable.  When data is transferred into main memory as part of a dump event, the fast I/O rates and low latency associated with Flash Express provide decreased first failure data capture time, and faster page-ins of the critical pages needed for dump creation.  This allows the system to return to normal workload performance faster, without incurring extra delays.

So clearly, adding a flash memory layer to the Mainframe storage hierarchy can only deliver benefit, and over time, seemingly Flash Express can be used for other I/O intensive applications, further improving response times and decreasing associated CPU cycles.

No doubt the non-Mainframe technician might say “so what, we’ve been doing flash and SSD for nearly 20 years.  Typical Mainframe, always behind the rest of the IT world”!

Hmmm, perhaps not, hence the Back To The Future reference.  Depending on your viewpoint, and perhaps urban myth, from a release date viewpoint, but in the mid-to-late 1980’s, StorageTek introduced a device called the 4080, AKA Control Data Set Manager (CDSM) for their Host Software Component (HSC) product.  You will still see this 4080 device referenced in the HSC Systems Programmer’s Guide technical manual.  The 4080 is a Solid State Device, utilizing memory storage, emulating 3380 and 3390 device types, allowing z/OS (MVS) data sets to be created and accessed, as and if required.  Arguably if not certainly, this was the first instance of logical and physical device separation for Mainframe disks, and urban myth might dictate that this StorageTek 4080 technology was used or at least borrowed for the first EMC Symmetrix disk subsystem…

StorageTek typically packaged this 4080 device within a NearLine (ATL) solution sale and customers used this SSD device for HSC Control Data Sets, but also other high I/O files, such as the IMS Write-Ahead Data Set (WADS).  Therefore, Flash Express isn’t the first SSD solution for the Mainframe, and similarly, Flash Memory/Cards, USB Sticks, et al, aren’t the first SSD type technology in the IT world.

As somebody (Machiavelli) far wiser than I once said “whoever wishes to foresee the future must consult the past”.  The longevity of the Mainframe, nearly 50 Years of technical innovation, largely dictates that technological ideas are at least borrowed from the Mainframe, whether System Paging, Fibre Channels, Storage Area Networking (SAN), Virtual Storage, Flash/SSD Memory, naming but a few.

Should today’s zEC12 Mainframe customer deploy Flash Express?  Without doubt, but to deliver maximum ROI and benefit, it should be considered as more than a faster paging and dump solution, and we look forward to how IBM will further enhance this product offering.