Today, IBM has announced the next version of their DS8000 family enterprise-level block storage device. Well, I for one have been a little curious to see what IBM was going to do for naming since they started with the DS8000 and had progressed to the DS8900F. I should have guessed that the numbers were hexadecimal all along. That’s right – the first two models to be released are the follow-ons to the DS8910 and DS8950 and are named the DS8A10 and the DS8A50. IBM has recently been announcing two DS8000 models followed the subsequent year with two more so I’m assuming there will be two additional family members coming out possibly next year.
Playing at the very high end of the block storage spectrum, these systems can do it all (fixed block, IBM i, and CKD storage) with exceptional reliability and performance for the most demanding clients. As such, IBM’s tagline for the marketing of this system is “Always Available, Always Performing, Always Innovating, Always Ready” and today I’m going to focus on performance and innovation.
Contributing mightily to the performance improvements is the new “plumbing” in the box as the PCIe Gen 3 bus has been replaced with PCIe Gen 4 throughout. The I/O bays, Flash Enclosures, and zHyperLink have all been redesigned to take advantage of this upgrade, which doubles the available bandwidth.
The third-generation flash enclosure design swaps out SAS connectivity for NVMe, and couples that with the PCIe enhancements to double both the available throughput and IOPs. There are four options for what IBM calls “Industry Standard NVMe Flash” drives as well as, and this is where it gets very interesting, three sizes of IBM’s own custom-designed FlashCore Modules (FCM 4) [say it with me: “FINALLLLYYYY!”].
IBM FlashCore Modules bring with them some very interesting features (side effects?) and potential enhancements, such as hardware-based compression and internal thin-provisioning. And nothing has been announced, but could we potentially be looking at some ransomware detection capabilities like those that IBM released for their FlashSystem family which use the very same FCMs?
Here's more good news (some might say “the best” news, depending on whether you were the person tasked with managing this) for those of us that have implemented Safeguarded Copy (SGC) and have had to deal with the dreaded “virtual addressing limit.” These systems will, at a minimum, double the virtual address limit from ~3.4 PB to around 8 PB for CKD volumes with small extents. Of course, everyone knows that you should use small extents for effective SGC performance. Well, that is true…unless you’re using FlashCore Modules…
Since pools built with FCMs are thin provisioned on the hardware itself, only written tracks consume space regardless of the extent size. Therefore, you can now achieve excellent performance and efficiency even when using large extents for use cases such as SGC. And when employing said large extents, the virtual addressing limit of the largest system then becomes 29 PB! (limit? What limit?!). Don’t worry Fixed Block users, your limits are even higher.
The DS8A10 and DS8A50 are further examples of IBM’s continued commitment to the Mainframe as IBM persists in applying funding to enhance its z-Synergy features. Case in point – a completely redesigned zHyperLink which IBM is claiming provides enhanced performance metrics such as a whopping 2.5x increase in link write throughput.
In my opinion, this is no mere “turn the crank” incremental release. Among the numerous enhancements, I’m confident that my clients will be particularly excited (wait – do IT professionals get excited?) about the integration of IBM’s custom-designed FlashCore Modules. The immediate benefits, such as improving performance and obliterating the virtual addressing limit, as well as potential future capabilities (can’t directly comment on that) will make life easier for my clients (and therefore for me!) and I am certainly “Always Ready” for that.
#Featured-area-2#IBMChampion