Integrated Analytics System M4002-001 model overview
The IBM Integrated Analytics System M4002-001 is a small data warehousing appliance available for client-provided rack installation that can be easily incorporated into your existing data center. This purpose-built solution is designed for simplicity and cost efficiency to better store, explore, and manage data for your small, medium, or developing organization.
IAS M4002-001 model is a single Lenovo D2 enclosure that is installed on a customer-provided rack. It contains three SD530 nodes accessible from the front. It attaches directly to customer provided PDUs and network switches. All connections to the appliance are direct links between the node and the customer network.
Linux PPC
) ,
and a separate one for M4002-001 model, which is marked Linux x86
. As of March
2020, the regular IIAS appliances can be upgraded to version 1.0.21.2, while the M4002-001 model is
upgraded to 1.0.19.0. The following release notes apply to M4002-001:
Hardware overview
- 2U Lenovo D2 Chassis
- Three SD530 nodes, each having:
- two M.2 drives - for OS and platform
- four 4 TB NVMe SSD drives - for user data
- 192 GB RAM
- 16 CPU cores (two 8-core 2.1GHz Intel 4110)
- 25G network card (two ports)
- 10Gb EIOM SFP+ with LC Connectors (two ports)
- Network:
- 25Gb ring fabric network via PCI NIC
- 3 or 6 10Gb SFP+ customer connections via EIOM (Ethernet I/O Module)
- 1 1Gb RJ45/ UTP monitoring connection via SMM port
Network
- Ring network among nodes:
Data fabric is a 25GBASE-SR ring of the three nodes. Redundancy is accomplished by running routing on the nodes, so any single failure is tolerated by routing 'in the other direction'.
- User connections:
- 10G user and data connections:
- The only access available is 10GBASE SR (Short Range) SFP+ transceivers with LC connectors.
- Can be single or dual link, or bonded pair.
- Link Aggregation Control Protocol (LACP) active-active only
- 1G hardware platform monitoring
The management network is across the 25G ring, and the customers network. The Baseboard Management Controller (BMC) and System Management Module (SMM) are accessed for management monitoring of the hardware, and those interfaces are customer defined addresses.
- 10G user and data connections:
IBM installers configure the network on customer site. Using apsetup script to configure the network is not supported for 001 models.
Storage
- On M4002-001 appliance there are two types of SSD storage. The M.2 drives are used for operating
system, and NVMe drives are used for database data. There is one GPFS filesystem, using all 12 NVMe
devices assigned as 12 NSDs.
There are 36 TB of data storage available, assuming 4x compression.
- External storage: NFS can be used for file storage, SAN is not supported.
- Tiered storage is not supported.
- Encryption is different than on rack-mounted models. For more information see Storage hardware encryption for M4002-001 model.
- NVMe drives must be enabled and disabled manually by using the ap sd command, as described in ap commands.
For more information on storage, see Appliance storage.
High availability
Hardware architecture is designed in such a way that a single component failure should not disrupt database availability in most cases.
Data protection and redundancy is managed by GPFS. IAS maintains two copies of persistent data in two different nodes and is capable of surviving a node failure as well as a drive loss. Due to the small number of nodes and drives, this arrangement can still achieve adequate resiliency comparable to FlashSystem Arrays in rack-mounted IAS. No dedicated hot spare drive is provided.
Although hardware itself as a system is functional with surviving nodes in the cluster, in M4002-001 appliance, a node (server) failure will cause Db2® Warehouse to fail and requires restart. Therefore, a node failure causes disruption in database operation.
The nodes utilize two M.2 drives in mirror for OS boot. These M.2 devices are not hot swappable. When any M.2 card needs replacement, the node needs to be shut down.
Network: The three nodes are directly attached to each other via non-redundant 25G Ethernet links. Ring Buffer settings ensure that connectivity is still maintained on a single link failure. However, system performance may degrade.
For more information on high availability solutions in IAS, see High availability.
Limitations
- DSX, apdsx command
- RStudio
- Lift
- AFM file-system based replication
- IBM Data Replication for Db2 Continuous Availability
- Tiered storage
- SAN external storage
- Grow on Demand
- JWT SERVICE