John Hill

Systems Engineer

New Rack Options With Enhanced Value!

On April 14, IBM announced the release of the IBM TS7700 R5.0.1. The capabilities and enhancements described within this article are expected to be generally available in the U.S. on Friday, May 15, 2020. As with all pre-release announcements, the date for general availability is subject to change.

So what can we look forward to with the TS7700 R5.0.1 Virtual Tape Library? For the most part these announcements are hardware-centric but before we get into it, let’s review and take a little trip back in time before Y2K and the “new millennium”.

History of IBM VTS

IBM’s mainframe virtual tape solution (VTS) is now in its 7th generation, going way back to 1997 (but who’s counting? Certainly not me!) I remember implementing the IBM TotalStorage Virtual Tape Server Model B16 back in those days and I can tell you it was a humbling experience, yet customers recognized it as a revolutionary product in the mainframe space. These early units boasted a 72GB or a 144GB disk storage behind an RS/6000 processor. It’s laughable how inadequate this storage capacity seems when you look at today’s massive data requirements, but back in the day if you had a terabyte or more of data in your mainframe environment you were one of the big dogs on the block.

In the early days, the B16 evolved into the B18, then the next generation B10 and B20. Somewhere between the B18 and B10/B20 rollout we got the first disaster recovery (DR)-capable unit, known as the Peer-To-Peer VTS. This was the first DR unit in the industry that could replicate virtual tape data between two units at any distance. If memory serves, there were separate frames called CX0s to hold special virtual tape controllers (VTCs) to manage the replication of the data. All of these early units were always backed by a tape library and physical tape cartridge since spinning disk was an expensive and coveted commodity. Back in these early days you could never justify spending your budget on spinning disk to support your tape storage environment. The best we could do is buffer the tape writes to a modest disk subsystem and be patient when it came time to read the virtual volumes. I give credit to IBM for coming up with the pre-migration process to get the data to tape and retain a copy on disk, albeit for a limited time period, using a least recently used (LRU) algorithm. These simple processes and algorithms made writing to tape much faster and reading from tape no slower than legacy tape drives and media.

Entering the “modern era” (2006), IBM introduced the TS7700. The earliest units (TS7740) still required a physical tape backend. Many of these units survive today with the currently supported controllers and disk expansion drawers. However, the TS7740 days are numbered, as the end-of-service date recently passed on April 30, 2020.

In 2008, IBM introduced the first totally tapeless unit, the TS7720. By this time, competition had arrived on the scene and, if you’re like me, you know competition is a good thing. It makes us work harder, innovate faster, and do the things we resist so we don’t lose out to the competition.

Introduction of the TS7760

In 2016, the TS7760 unit debuted. This marked the convergence of the tapeless and tape-attached units, hence the marketing model designation of the TS7760, which referenced the TS7740 and TS7720 combined. This innovation gave customers a single model to work with and if you wanted to add tape to your tapeless unit you could do so with the appropriate hardware and feature codes. At this point in time, the TS7760 was using the IBM P8 platform, which is the N-1 generation of pSeries. The P8 processor provided two 10-way processor cards, 3600MB/s of host throughput, and 32GB of RAM. The TS7760 supported both 8GB and 16GB FICON and enhanced compression algorithms (LZ4 and zSTD) with compression ratios up to 7:1 depending upon the data being compressed. With an additional 32GB of memory (total of 64GB), the R4.2 level of code provides for the cloud attachment feature.

The IBM TS7700 supports a binary choice of tape or cloud attachment, although neither is required. For large installations where you do not want to store rarely accessed archive data on spinning disk, however, the addition of a cloud or a tape tier of storage makes perfect sense. When tiering to tape or cloud, the same LRU algorithm introduced back in 1997 will release inactive tape volumes from local disk cache and retrieve them from tape or cloud automatically when requested.

Do you have some data associated with applications that cannot tolerate a delay in a tape mount such that cloud or physical tape would not work? No problem. The TS7760 introduced disk partitioning so that application data requiring fast mounts can be retained on disk, referred to as cache resident data or Cache Partition 0 (CP0) data. Data expected to have infrequent access and a long shelf life is retained in the tape or cloud partitions (CP1 – CP7) and is automatically pushed out and, if necessary, back into the disk storage. Now that the TS7760 is entering maturity, the overall solution can support up to 2.4PBs of usable capacity on a single cluster prior to compression and can expand to an 8-way grid (Hydra) with no geographic limitations.

Capacity on Demand

Last October (2019), IBM announced the latest version of mainframe tape hardware, the TS7770 and the latest in microcode, R5.0. This latest release of TS7700 is one based on a Capacity on Demand (COD) model. The TS7770 uses 10TB HDDs and with this capacity increase comes a need for more disk parallelism as disk performance is not keeping pace with capacity. For this reason, a minimum of two disk enclosures are shipped with every unit and, as capacity requirements increase, disks are added in increments of two enclosures. This would appear to lock out smaller mainframe accounts as no one wants to overbuy, especially when alternative solutions are available. The CoD model solves some of this problem by licensing capacity in 20TB increments. Since the base minimum installation of 157TBs of usable capacity can easily exceed the needs of small mainframe accounts, the CoD model provides a minimum licensing capacity of 20TBs. Applying compression of 3:1 (and likely much higher) can easily get you to 60TB of usable capacity. Need more capacity? Add another 20TB license and you can get to 120TB of usable capacity applying the 3:1 compression. In total, the very base configuration can go to 140TB (seven 20TB increments) of usable capacity prior to compression. Overall, the single frame TS7770 has a maximum usable capacity of 789TBs. Add a fully populated expansion frame to the base and you sit at 2.37PBs of usable capacity prior to compression.

Encryption, Power, and Object Offload

Release 5.0 of the TS7770 is not all about capacity. It also deploys the new P9 processor and new functionality such as Secure Data Transfer which provides for encryption over the replication wire and a new method for local and external AES 256-bit encryption that is KMIP compliant. With the new hardware bundle comes the option for single-phase or three-phase power for those who require it. DS8000 Object Offload is a new feature that provides a certain synergy between IBM storage products which ultimately saves MIPS in the mainframe environment and can reduce those MLCs that can drive up mainframe compute costs. The DS8000 Object Offload feature converts the TS7770 into a combination object store and virtual tape device. In the previous version of DS8000 Offload, the IBM TS7700 was a choice between a tape device or an object store. Now that customers no longer have to choose between the two, perhaps z/OS can begin to get out of the data movement business. For now, only DFHSM and DB2 archives are supported but it is anticipated that more software applications will join the fray of DS8000 offload candidates.

April 2020 Announcement

On April 14, IBM announced R5.0.1 of TS7700 microcode, also called R5 PGA1. With this latest code release you have the ability to add a second expansion frame to the TS7770. A fully populated three-frame cluster provides an incredible 3.9PBs of usable capacity prior to compression. The CoD increments are now at a strictly 20TB clip level. With the initial R5.0 code release, only the first 200TBs of CoD were available in 20TB chunks with a requirement of 100TB increments thereafter. IBM quickly realized the fallacies associated with this CoD model and resolved the issue.

Here is the “Out-of-the-Box” part of this announcement, and maybe we should call it “Out-of-the-Frame” thinking. For those who have been following this product for as long as I have, you may want to have a seat before reading any further. You can now have a fully functional TS7770 in your own 19’ rack!!! (somebody queue the Gregorian Chants). Yes. It is true. Of course, there is a process with strings attached:

1. You must get your Mainline representative to apply for an RPQ to allow for the shipment of the equipment without the standard 3952-F06 frame/rack.
2. You must provide the targeted rack make and model along with any already-installed gear within the target rack.
3. The rack must meet the IBM design specs and airflow requirements for heat dissipation.
4. There are minimum height requirements for placement of the TS3000 console and the minimum configuration will be a full 18u of contiguous space within the rack.

The minimum configuration could provide up to 140TBs of uncompressed capacity and, best of all, no compromise when it comes to functional capability. The maximum configuration would be equivalent to a fully populated single frame TS7770, providing the full 789TBs of usable capacity and occupying a full 42u of rack space. I find the smaller units much more exciting, as the possibilities for designing your own high-availability system in as little as 36u of rack space can be very appealing, in particular to those who provide and support IaaS and DRaaS as part of their business model.

I need to note that today this offering from IBM does not allow for the physical MES to add capacity. This will eventually be available; just know that you need to anticipate any growth or expansion when ordering, as you will not be able to add capacity until December 2020 at the earliest. Regardless of when you may need it, you must reserve space in the rack for anticipated growth as there is a requirement that expansion capacity must be contiguous within the rack.

So that I may leave you on a high note, IBM has finally recognized the market share they have been losing over the years to their competitors. In some cases, these competitors have become IBM partners. As an end-to-end IT solution provider, Mainline offers these competitive products to our customers to meet their needs. Currently, IBM is making a hard push to regain some of this lost market share with a very aggressive pricing structure for the smallest minimum configurations. There are certain features that are prohibited from these configurations and other features that are required, so please contact your Mainline representative for details. Keep in mind that the functional feature codes that may be prohibited within this aggressive pricing model can be added afterwards, hardware capacity excluded.

More Information
To learn more about solutions for your compute, storage, and networking needs, please contact your Mainline Account Executive directly, or click here to contact us with any questions.

You may be interested in:

Blog: IBM FlashSystem Storage Use Cases Span Large Enterprises to Small Lines of Business

Video 2:05: IBM Storage Insights

Video 1:50: IBM Storage Insights Pro

Blog: 2020: The Year for Private Cloud on IBM Z