Why do large banks, financial institutions, and e-commerce business prefer our flagship IBM DS8000 series storage for their mission-critical workloads?
You guessed it right! Storage solutions from IBM not only provide high performance, enhanced security/cyber resilience & rich RAS feature set, they are essential to all major organizations for their service.
DS8000 provides the world’s most resilient storage environment. Whether customers are implementing highly available (HA) systems or disaster recovery (DR) strategies, the DS8000 provides lower recovery point objectives. This results in significantly lower recovery time objectives. With safeguarded copy, the DS8000 further extends its lead into business resiliency and the best HA/DR capability. The DS8000 even provides customers with the ability to recover from cyber-attacks (malware, ransomware & other internal threats) in hours instead of weeks, months or even years.
Now, let’s have a look at the building blocks of DS8K storage.
Power of DS8K:
IBM DS8K is the only storage system built on IBM POWER processor with AIX as the OS and is the most reliable server and processor platform in the industry today (after mainframes and z/OS). Moreover, a mature and stable microcode establishes the DS8K as Flagship storage in the industry.
IBM DS8K uses 3-layer architecture as mentioned below:
Host Adapters manage external FC path protocols for distributed host I/O and FICON path for z/OS mainframe I/O. It also helps you replicate data to other DS8K systems.
Raid Adapters / Device Adapters manage the internal storage devices. It is used for managing paths to disk enclosures, RAID protection, drive sparing & rebuilds when drive failure occurs.
Redundant Controller Nodes power of DS8K storage comes from our own IBM Power based active-active controller architecture which uses 4-way SMP and positioned between host and device adapters layer. Power controllers leverage AIX based kernel to manage the Power9 processors and power memory as a data caching area.
IO Enclosure Pairs:
Both host adapters and device adapters physically present in I/O Enclosure. By default, I/O Enclosure will be available in pairs for redundancy.
In DS8K, we call the disk enclosures as “HPFE” high performance flash enclosure and HPFE can be added only in pairs. HPFE Gen2 designed to deliver high performance and extremely low response times through PCIe connectivity to controller nodes via I/O enclosure pair.
HPFE supports 800 GB, 1.6 TB, 1.9 TB, 3.2 TB, 3.84 TB, 7.68 TB and 15.36 TB flash drives with 5, 6 and 10 RAID options.
Apart from IO Enclosure pair, HPFE pair and controller pair, we have management enclosure which has dual HMCs which is the point for management & monitoring activities. Management enclosure provides the communication to client’s local area network.
1U KVM unit is used for managing the storage in the data center.
DS8900F Offering Models:
As you are aware, storage industry is moving towards an all-flash model. IBM DS8900F has been built as an all-flash storage to align to the market, just how our clients need it!
DS8900F is categorized in two offering models as below:
- Flexibility class
- Agility class
Within flexibility class, we have two sub-categories namely, rack-less and racked system. The Rack-less model can be integrated into customer’s existing standard 19” rack (or) we can integrate it with the existing mainframe (or) LinuxOne rack which has 16U reserve space for mounting the DS8K storage.
For detailed specifications, please refer to the data sheet:
Highlights of DS8900F:
- DS8900F storage can be used in both mainframe as well as open/distributed host environments
- With “zHyperLink”, clients can achieve ~18 microseconds latency in mainframe environments
- Enterprise resiliency with better than seven 9’s availability
- 2 to 4 seconds of RPO for 2-site replication and it is far better than other players
- Capability to configure 3- 4 site replications
- Enhanced cyber resiliency with safeguarded copy
- End-to-end data encryption (in-flight, at-rest and in the cloud)
The unique cache architecture helps to increase the effective cache utilization and improves performance and latency. Cache is segmented in 4K and it improves the overall performance of OLTP workloads.
Unique replication methodology can give an RPO of as low as 2-4 seconds for a 2-site asynchronous replication and with the capability to handle errant network links would make more effective utilization of the links.
3-to-4 site replication will fulfil zero RPO requirements.
IBM remote mirroring feature provides,
It offers a synchronous copy option that constantly updates a secondary copy at the target site. Usually, we use metro-mirror replication between DC site and NDR site
It offers long-distance asynchronous replication and is used for DC to DR (or) NDR to DR replication
Metro/Global Mirror function combines the capabilities of both Metro Mirror and Global Mirror. It is configured using a minimum of three storage units, one each at DC, NDR and DR.
Like this mighty warrior had a vulnerable point, every enterprise today in journey to Hybrid Multi-Cloud, Multi-Site Replication has a similar pain point which is Cyber Attacks.
DS8900F SafeGuarded Copy can:
- Protect against data corruption (or) loss caused by malicious events
- Point-in-time with more frequency
- Prevents against accidental / deliberate deletion actions
- Non-accessible to host infra and safeguarded backups will be in hidden mode
- Enable any previous recovery point to be made available on a set of recovery volumes while the production environment continues to run
Other than the mentioned functions, we have zHyperLink which can help us in delivering as low as ~18 microseconds latency when connected to “Z” hosts like IBM z14 and higher
Logical Storage Construct:
So far, we covered building blocks and key capabilities of DS8K.
Let’s jump in to the logical layers for volume provisioning
Disk: Disk or Flash card is the physical disk installed in HPFE pairs
Array Site: Every eight drive is grouped as an array site
Array: Array is formed when we apply RAID on Array site
Rank: Rank will be created on an array by defining the storage type either FB (or) CKD
FB refers fixed block and the storage can be presented to open / distributed host environments.
CKD refers to Count Key Data which can be mapped to mainframe environment.
When defining a rank, we can also specify the extent size:
- Extents for FB can be small 16 MiB FB extents (or) 1 GiB FB extents
- Extents for CKD can be 21 Cylinders (or) 1113 Cylinders
When we have smaller 16MiB extents, movement of extents between tiers during “Easy-Tier” promotion will be faster and more granular based-on data access patterns
Extent with similar storage type (FB or CKD) pooled together can act as an “extent pool”.
A volume is a fixed amount of storage on a storage device either fat (or) thin-provisioned / ESE (extent space efficient).
Host is a server / initiator that interacts with storage array (should be zoned properly)
The logical subsystem (LSS) that enables one (or) more host I/O interfaces to access a set of devices.
DS8900F storage can be managed through,
- DSGUI (or)
- DSCLI (or)
- DS8K Restful API
In addition to the above, the copy services manager will be used for managing Flash copies and Replication (metro-mirror, global-mirror, metro-global mirror, safeguarded copies etc.).
DSGUI/DSCLI & Copy-services-manager (CSM) can be installed & made available in HMC itself.
CSM-on-HMC can manage up-to four DS8K storage arrays. If you have more than four DS8K, then we can install a dedicated CSM server considering the CPU, memory resource requirements.
There are so many other factors/metrics keeping us differentiated from any other storage provider in the market. To learn more about, please land it to IBM sites.
Interactive Product Demo: 3D (ibm.com)
To summarize, IBM’s DS8K series storage is highly available and highly secured for your mission-critical data & workloads. It has been around for over 15 years and is evolving to meet changing client needs.