The Information Management world is abuzz with talk of Oracle’s TimesTen-based “engineered system” – Exalytics! It is a high-performance, in-memory appliance that delivers fast, actionable interactive and insightful analytics by integrating optimized hardware and software components to deliver a complete analytical solution – combining modeling, planning and reporting all on a single box!
So, what does that mean? Let’s get a little under the hood and check out Exalytics’ foundational building blocks.
The Exalytics Eco-system
- Runs the BI layer on a multi-core 1 TB server
- An in-memory cache (Oracle TimesTen 11.2.2.2) is used to accelerate the BI part of the stack
Key Hardware Components:
- Sun Fire X4470 M2 Server
- 1 TB RAM, 40 Cores (four Intel Xeon© E7-4800 series processors), 3.6 TB HDD
Key Networking Components:
- InfiniBand: 2 quad-data rate (QDR) 40 GB/s InfiniBand ports are available with each machine expressly for Oracle Exadata connectivity. When connected to Exadata, Exalytics becomes an integral part of the Exadata private InfiniBand network and has high-speed, low latency access to the database servers. When multiple Exalytics machines are clustered together, the InfiniBand fabric also serves as the high-speed cluster interconnect.
- 10 GB Ethernet: Two 10 GB/s Ethernet ports are available for connecting to enterprise data sources and for client access.
- 1 GB Ethernet: Four 1 GB/s Ethernet ports are available for client access.
- Dedicated Integrated Lights Out Management (ILOM): Ethernet port for remote management.
Key Software Elements:
- Optimized OBIEE 11.1.1.6
- Oracle Essbase 11.1.2 with enhancements
- Oracle TimesTen for Exalytics (supports columnar compression)
- Runs on 64-bit Oracle Linux
- OBIEE and Essbase are licensed as Oracle BI Foundation and their Exalytics-specific functionality and features can only be used in conjunction with the Exalytics Hardware.
A typical Exalytics Prototype Environment
Exalytics Architecture: Developer tools are used for defining and maintaining aggregate definitions, client tools for OBIEE reporting, TimesTen and Oracle DB; the Exalytics layer contains OBIEE, TimesTen and Essbase and the DB tier contains FMW DB repository, usage tracking and summary stats.
In addition to an Exalytics machine and an external DB, the following components go on a Developer Workstation: Oracle BI Admin Tool, Summary Advisor Wizard, TimesTen Client, SQL Developer, Oracle DB 11g Client.
Exalytics includes two in-memory analytics engines that provide the analytics capability – Oracle TimesTen In-Memory Database for Exalytics and Oracle Essbase with in-memory optimizations for Exalytics. These two data management engines are leveraged in the following four techniques to provide high performance in-memory analytics for achieving the aforesaid query optimizations:
- In-memory Data Replication
- In-memory Adaptive Data Mart
- In-memory Intelligent Result Cache
- In-memory Cubes
Some of these are Oracle’s genuine “secret sauce” (new algorithms and functionality) and others are simply descriptions of standard existing BI/DW strategies extended to take advantage of the in-memory concepts.
My next blog will make comparisons between SAP HANA and Exalytics.
I look forward to your next blog comparing HANA with Exalytics. I encourage a hearty, fact-based debate and hope that you will portray HANA accurately, unlike Oracle Corporation blogs which have made exceedingly false claims about how HANA works or doesn’t. If you have any questions about how HANA works, please feel free to reach out to me, I’ll be glad to pass you documentation or information to ensure technical and functional accuracy.
Thanks so much for your interest in my blog and this debate. My next blog may disturb you a bit, in that it is going to present Oracle’s position on Exalytics vis a vis SAP HANA. However, the followup blog which I will publish soon thereafter will make a more critical comparison between the two, based on other inputs.
So, please stay tuned and let’s have fun with this!
Thanks, Neetu. It doesn’t disturb me that you posted Oracle’s claims, though it does disturb me that Oracle would make such egregious claims – every one of which are false. (I watched the webcast where Thomas Kurian made these claims) Not only are they false, they are easy to check! The documentation for HANA is online and anyone can see that it supports indexes and you can see the SQL for how to create them. Limited & Non-standard SQL? It’s ANSI standard. Perhaps I should send Thomas a link? 🙂
In addition to that, SAP has performed scale-out testing across 16-node clusters with 100TB raw data, and even most recently across a 51-node cluster with 625TB raw data, and published these results. So to say HANA doesn’t support scale-out is crazy.
Of course, they didn’t limit their false claims to HANA – they claimed that TimesTen supports columnar storage, which it does not. (it supports columnar compression, but that is hardly the same thing, and they claimed that separately)
I look forward to your analysis of Oracle’s claims!