In his blog 10 questions you need to ask your SAP HANA hardware vendor John Appleby challenged any of the 6 SAP HANA hardware vendors (Cisco, Dell, Hitachi, HP, IBM and NEC) to answer these questions publicly. Carsten Marcos Nitschke from IBM's Systems & Technology Group took up the mantle and here is the result!
1) What is your budgetary price for 512GB, 2TB, 2.5TB and 8TB appliances?
IBM follows a building block approach for SAP HANA Appliances. The key points for this are:
Only internal storage is used (i.e. there are no expensive network storage units)
ScaleOut - the means of chaining appliances together into a cluster - is achieved via the IBM General Parallel File System (GPFS).
This is the highest level of investment protection for customers since the only thing they will have to do is add SAP HANA nodes of the 512GB size to their landscape. Neither further SAN nor NAS are required, leading to reduced complexity of the SAP HANA landscape.
Customers have the choice of scaling-out with additional 512GB nodes, or, scaling-up existing 512GB nodes to 1TB and then scale-out. For example, 8TB can be built as either 16x 512GB or 8x 1TB. The latter provides the least communication overhead between the nodes and therefore better performance.
IBM is offering ScaleOut configurations for the 256GB, 512GB and 1TB Appliances each for up to 16 nodes.
The really cool outcome of IBM's Building Block approach is that the cost per TB of appliance is slightly better than linear.
2) How big (in Rack Units) and how much power, do the above systems consume during idle, average and peak load?
The IBM SAP HANA Appliances are 2U for the 128GB to 256GB, 4U for the 512GB Model and 8Ufor the 1TB Model - i.e. 8U/TB in all cases.
Table for power Consumption all values are up to:
3) What is your cost and upgrade path from 512GB to 1TB to 2TB?
The IBM M-Size Model of the SAP HANA Appliance (512GB) Model allows the growth to 1TB without any further intervention by simply adding another 512GB node to it. Scaling out will require the use of a 10Gbit Switch in order to connect the nodes. Once the Switch is in place customers can grow up to a 16TB ScaleOut configuration by simply adding nodes. Nothing has to be discarded.
As with an initial installation, the upgrade or the addition of new nodes has to be performed by educated service personnel in order to ensure a minimum downtime of an existing configuration. The growth of existing ScaleOut configurations can even be performed with zero downtime.
With regards to the cost, an upgrade on IBM appliances is linear and can be easily calculated from the initial configuration. The implementation of ScaleOut by educated service personal takes no longer than the initial set-up at the customer.
Failover is something I would suggest to any customer in any case since SAP HANA is for mission critical applications in the customers. IBM Failover is designed to not have any single point of failure. This is due to the Shared Nothing Architecture of the IBM SAP HANA Appliances. The implementation of Failover at the customer consists of at least 1 Standby node of the same size in a ScaleOut configuration. The implementation requirements for Failover on IBM SAP HANA Appliances are similar to the ScaleOut Implementation.
4) What is your upgrade path past 8TB?
The 8TB mark has been surpassed already quite some time ago with the certification of the 16 node 1TB cluster by SAP. This is part of the standard configuration offered by IBM. Nevertheless, there are already customers who need larger SAP HANA ScaleOut configurations than the 16TB, and those customers will simply add more standard SAP HANA nodes of 1TB to it in order to accommodate their needs.
The key takeaways regarding ScaleOut of SAP HANA with IBM are:
ScaleOut certified for:
Up to 16 node 256GB SSD config with HA = 4TB
Up to 16 node 512GB config with HA = 8TB
Up to 16 node 1TB config = 16TB
IBMs Building Block Approach makes the solution scalable without any known limitation. For those customers who will need a ScaleOut configuration beyond the 16TB offered today, IBM offers a joint certification at the customer site working closely with SAP.
You might have seen the Keynote pitch from SAP Founder Hasso Plattner at SAPPHIRENOW in Orlando. Hasso showed a 100TB SAP HANA ScaleOut Cluster with High Availability powered by IBM. This ScaleOut Cluster has been configured, delivered, racked and installed within 4 weeks.
5) How long do you agree to provide compatible upgrade hardware?
IBM is committed to providing an upgrade path for existing ScaleOut configurations upon the availability of new models comprising new options and even with new processors.
Node upgrades (scale-up) will be provided as long as the required components are available, e.g. if a processor will go end of life, IBM will no longer be able to provide upgrades to such systems accordingly.
6) What is your Disaster Recovery capability?
Disaster Tolerance has not yet been certified by SAP with any hardware vendor, whilst there are validations for demo environments. IBM will certainly support Disaster Tolerance capabilities provided on an SAP HANA application layer. In addition, IBM is developing (in cooperation with SAP) a platform based Disaster Tolerance solution that will make use of storage replication between data centers. IBM and SAP are working on a solution that will minimise the performance penalty associated with such a solution.
As of today, the export and recovery of backups to a remote site can be used to mitigate a disaster.
7) What is your SAP HANA Support offering?
IBM offers all the appliances with a 3 year 24*7 support and 4-hour response time for all hardware related matters. The Operating System used in IBM appliances is the SUSE Linux Enterprise Server with priority support for SAP applications. The support of SAP covers the database. In order to assure highest quality and customer satisfaction, IBM or educated partners are performing the initial setup of the SAP HANA appliance(s) at the customer site and realise a go-live check to ensure complete functioning of the IBM Solution for SAP HANA.
8) What is your typical lead time for a 512GB node?
The committed lead-time for IBM System x orders is 6 weeks from contract signature to delivery of the appliance to the customer. However, as the solutions are prebuilt at the manufacturing site, IBM has a track record of providing most of the systems to customers within 2 weeks.
9) What is your try-and-buy policy and typical availability?
This might vary from country to country. The try-and-buy approach provides the highest value to the customer and underlines a firm commitment on all sides to the SAP HANA Project, which in any case will be highly strategic. Upon closure of the contract, try-and-buy systems are delivered with the same lead-time as regularly ordered IBM Solutions for SAP HANA
10) What is your SAP HANA Roadmap and commitment?
IBM Solutions for SAP HANA usually come with a 3-year support & service contract. They can also be ordered with up to 5 years of support and service. IBM plans to provide updated models as well as new models upon the availability of new options, especially processors, and new system building blocks. IBM is working with SAP to get such systems validated as soon as possible. What these new models will look like will be worked out in close cooperation with SAP as their SAP HANA appliance evolves over time.
Carsten holds a degree in Business & International Management from the Hochschule in Bremen Germany. He started his career at IXOS Software International (OpenText) as Sales Manager for Spain and Portugal, before quickly moving onto role of MD for the same region. After 3 years Carsten moved to Ariba where he worked as Business Development Manager with Telefonica and Banco Santander. Following this he moved to Internet Security Systems, an IBM Company, until 2010. Since 2010 he has focused on Business Analytics in IBM Spain, Portugal, Greece and Israel within the Systems & Technology Brand of IBM.
Disclaimer: The opinions expressed here are the views of the writer and do not necessarily reflect the views and opinions of IBM or Bluefin Solutions Ltd.