Archive for the ‘pureScale’ Category
Do Oracle check their facts before they issue a press release? Because today there is yet another instance of a blatant mistruth issued by Oracle. This time in an official Oracle press release about an SAP benchmark result. Here is the offending quote:
“Oracle’s superior scalable cluster architecture has full high availability unlike IBM’s that does not scale beyond a single server“
The quote is attributed to Juan Loaiza, Senior Vice President, Systems Technology at Oracle Corporation. Now it is possible that this is not a case of Oracle intentionally trying to mislead the public. It is possible that this is a case of poor fact-checking from Oracle. And if that is the case, then they should have checked yesterday’s SAP benchmark results when an IBM DB2 cluster took top spot in another SAP benchmark.
For the record, IBM DB2 has outstanding scale-out capabilities. IBM DB2 provides both shared-nothing partitioning scale-out capabilities as well as shared-disk clustering scale-out capabilities. Many would argue that IBM DB2 has significantly superior scale-out capabilities when compared with Oracle Database. Especially when it comes to scale-out efficiency.
Note: When this first came to light, I was a little upset. After taking a little time to calm down, I updated some statements in this blog post to tone them down. Thankfully, I probably did this before anyone got a chance to read them
A couple of years ago, IBM introduced the pureScale feature, which provides application cluster transparency (allowing you to create shared-disk database clusters). At the time, IBM had taken their industry-leading clustering architecture from the mainframe, and brought it to Unix environments. IBM subsequently also brought it to Linux environments.
Today, IBM announced its first public industry benchmark result for this cluster technology. IBM achieved a record result for the SAP Transaction Banking (TRBK) Benchmark, processing more than 56 million posting transactions per hour and more than 22 million balanced accounts per hour. The results were achieved using IBM DB2® 9.7 on SUSE Linux® Enterprise Server. The cluster contained five IBM System x 3690 X5 database servers, and used the IBM System Storage® DS8800 disk system. The servers were configured to take over workload in case of a single system failure, thereby supporting high application availability. For more details, see the official certification from SAP.
Philip Howard, Research Director at Bloor Research, recently evaluated the performance, scalability, administration, and cost considerations for the leading integrated systems from IBM and Oracle for OnLine Transaction Processing (OLTP) environments. Here is a summary of his conclusions:
And here is a video with his evaluation. It is packed with practical advice regarding storage capacity, processing capacity, and more.
IBM has prepared a DB2 pureScale Trial and Demo Kit. With this kit, you can run a DB2 pureScale trial environment on 3 physical servers (any generic x86_64 hardware or even a set of VMware images). This kit is very easy to use. You simply download the kit, copy it onto a USB flash drive, and then insert that USB flash drive into a server (or mount it as a device in a Virtual Machine), and then boot from the USB drive in BIOS. It will then run the operating system, DB2 pureScale, demo, and the rest of the kit contents. Here is a screen capture of the kit welcome page.
To get the kit, please visit the DB2 pureScale Trial and Demo Kit page on IBM developerWorks. If you have any questions about the kit, please email firstname.lastname@example.org with subject “IBM DB2 pureScale Acceleration Kit for Trial and Demonstration”.
Last year, IBM released the pureScale feature for DB2, which allows you to create shared-disk clusters for database transaction processing. One of the great things about DB2 pureScale is that it provides near linear scale-out efficiency, which means that when you add a node to the cluster, you almost get a full node worth of additional processing power. In other words, the overhead involved in coordinating cluster activities for DB2 pureScale is relatively small.
Of course, this is in contrast to Oracle RAC, where there are numerous public accounts of poor scale-out performance. Although, to be fair, many of those Oracle RAC horror stories are probably due to people not making their applications cluster-aware, or due to people not having the skills to properly tune those systems. In other words, if you are willing to spend enough time and money, I reckon a lot of those Oracle RAC systems can achieve reasonable scale-out-efficiency. Of course, with DB2 pureScale, you don’t have to worry about any of that. With DB2 pureScale, you simply add or remove nodes as you please. The DB2 pureScale approach does not require that your applications be cluster-aware, and does not require complex tuning. DB2 pureScale actually delivers on the promise of being truly application transparent.
Okay, that’s enough background information… on to the actual subject of the blog post. A few months ago, IBM made DB2 pureScale available on x86 servers. That was a big first step towards making DB2 pureScale clusters more affordable. And last week, IBM took another big step in that direction by announcing that DB2 pureScale is now included in DB2 Workgroup Server Edition. DB2 Workgroup Server Edition is our mid-range offering with a list price of $15k per socket.
This means that, for instance, you could purchase a couple of these $15k DB2 Workgroup Server Edition licenses for use with two x86-based servers, each of which has one socket with 4 CPU cores. You could then create a 2-node cluster, where each node has a hefty 4 CPU cores. You would enjoy the great performance of DB2, as well as the continuous availability offered by DB2 pureScale. And all this for a very good price (that is subject to standard software discounts).
DB2 pureScale and DB2 Workgroup Server Edition offer a great alternative to running Oracle RAC for small clusters. Not only is the IBM combination attractive from a pricing standpoint, it is also attractive from a performance and subsequent ease of scaling standpoint.
IBM has just announced the availability of DB2 pureScale on System x at a Smarter Systems Tour event in Beijing.
DB2 pureScale is the Oracle RAC-killer that IBM announced last year. It brings IBM’s industry-leading clustering architecture from the mainframe to distributed systems. DB2 pureScale benefits include centralized lock management, which eliminates the need for the locality of reference that is sometimes hardcoded into applications that use Oracle RAC. Also, when nodes go offline, either for planned or unplanned reasons, centralized lock management improves the rate at which data becomes available again to the application. DB2 pureScale uses remote direct memory access for inter-node communication, improving system efficiency. When you combine all of this with superior scale-out efficiency, DB2 pureScale makes for a very exciting product.
When it was first announced, DB2 pureScale was made available on IBM Power Systems (high-end Unix servers). Many people I talked with, both among analysts and clients, immediately asked when it would be available on x86-based hardware. Well, now it is available on servers ranging from the IBM System x3650 M3 to the IBM System x3850 X5.
For more information, see the DB2 pureScale Web page.
Today, at an event in California, IBM is disclosing details about a new integrated system for transactional workloads called the IBM pureScale Application System. This new integrated system will comprise of DB2 pureScale and WebSphere Application Server on IBM Power 770 servers that run the AIX operating system. When you receive the system, the components are already integrated, already configured, and already tuned for handling transactional workloads.
IBM pureScale Application System is the latest in IBM’s line of workload-optimized systems. IBM’s strategy is to provide integrated systems that are optimized for particular workloads. For instance, the IBM Smart Analytics System that was announced last year is optimized for analytical workloads. Whereas IBM pureScale Application System is optimized for transactional workloads. This optimization extends from hardware configuration through software configuration and storage configuration. By the way, today, IBM also announced new models of the IBM Smart Analytics System for System x (x86) and System z (mainframe).
We are all well aware of the benefits of such integrated systems. These typically include lower initial purchase costs and single-number support. Because the systems are pre-integrated, they take less time and effort to deploy, which means that you enjoy faster time-to-value. I could go on-and-on, but I’m sure you know all of this already.
So, what’s so special about IBM pureScale Application System? Of course, the individual components of this system all have their own merits. From the raw performance, server utilization, and resiliency of the IBM Power 770, to the leading performance, cluster scale-out efficiency, and continuous availability of DB2 pureScale, to the performance, transaction integrity, and flexibility of WebSphere Application Server. These are all very compelling aspects to the system. However, in my opinion, the real value of this system lies in:
- Workload Optimization
Unlike Oracle Exadata, which is designed to handle both transactional and analytical workloads within the same environment, IBM pureScale Application System is optimized for transactional workloads only. Transactional and analytical workloads are fundamentally different in nature. Catering for both types of workload within the same system means that you cannot truly optimize for either of those workloads, which in turn means that you will have a less efficient and more costly IT environment. Database software is very expensive to purchase and maintain. In my opinion, to prudently and efficiently run your IT environment, you cannot afford to run less efficient systems. You need systems that are optimized for the work being performed.
- Flexible Configuration
The base system for the IBM pureScale Application System consists of two IBM Power 770 servers, each of which has 4 CPU cores and 32GB of RAM. There are two IBM Power 770 servers for redundancy reasons. They are connected by an Infiniband network. Initially the system is configured with four logical partitions that are dedicated to DB2 pureScale. You can optionally add partitions for WebSphere Application Server. The system then installs the software on these partitions, and optimizes the settings for the chosen configuration.
You can scale within by adding CPU cores to each IBM Power 770 enclosure. You can scale up by adding additional enclosures for each IBM Power 770 server. And you can scale out by adding additional IBM Power 770 servers. Adding WebSphere Application Server instances and DB2 pureScale cluster nodes is flexible and easy.
This flexibility to grow as your needs dictate is in stark contrast to the rigid configurations for the Oracle Exadata system.
- Flexible Licensing
One aspect of the IBM pureScale Application System that excites many clients is the flexible licensing. DB2 pureScale has daily pricing. When you combine daily pricing with true application transparency, where you can easily add and remove cluster nodes in minutes and without application changes, it means that clients do not need to over-provision for the worst case scenario. Instead, they can add or remove capacity as needed, and only pay for the additional capacity for the days on which they use it. This can dramatically reduce software license and maintenance costs for retailers who have huge spikes in activity during holiday periods, for organizations with similar “busy periods”, for business who want to run special promotions, and so on.
IBM has not yet announced the availability date for this new system. However, it is expected to be available for purchase in June of this year.
I mentioned earlier this week that IBM has published scale-out numbers for DB2 pureScale. I thought I’d take a few minutes to let you know about those numbers. To measure the efficiency of scaling out a system, IBM successively added DB2 servers to form larger and larger clusters. A workload that is representative of a Web commerce application (90% read and 10% write) was used. Unlike many Oracle RAC benchmarks, the IBM lab tests did not require any cluster awareness (i.e. transactions were completely random, with no need for transaction routing to specific nodes). The diagram below shows the results. Up to 64 nodes in the cluster, the scalability (compared to the 1 member result) is above 95% and at 128 nodes the scalability is at 84%.
By the way, if you want to see how easy it is to add or remove nodes in the cluster, look at the following video. Please remember that the DB2 pureScale approach does not require any re-partitioning of data, either manual or automated, when nodes are added or removed. This is because DB2 pureScale is based on the approach that has been successfully running on the mainframe for decades, which takes advantage of centralized locking to provide the ultimate database scale-out efficiency.
DB2 pureScale created quite a stir at the IBM Information on Demand conference. A number of people wanted to know more about how to use the “on-demand capacity” aspect of DB2 pureScale. I thought you would be interested to hear how some clients plan to use DB2 pureScale. I have anonymized the company names and removed any specific details about environments due to the early nature of the engagements/discussions:
- Recently, the “business” people at a US airline wanted to implement a big promotion. They built the promotion, only for the IT department to inform them that the systems would not handle the increased transactional workload, and so they cancelled the big promotion. The problem wasn’t that they could not add capacity to their systems. They could. They were using a non-IBM database system by the way. The problem was that adding database nodes to their non-IBM database cluster is a non-trivial project. And then, when you consider all the factors, it was questionable whether it made sense to remove those database nodes after the promotion. Well, this airline are now quite excited about DB2 pureScale’s ability to easily add and remove capacity. This is made possible because DB2 pureScale does not require that your applications are cluster-aware and does not impose best practices of partitioning your data across the nodes in the cluster. DB2 pureScale is truely transparent to applications. When this airline needs additional capacity, they simply add one or more logical partitions (LPARs) of DB2 pureScale to handle the additional capacity, and then remove them when they are no longer needed. They pay for the additional DB2 capacity only for the duration of the promotion.
- We are currently engaged with several large retailers. These retailers typically engage in capacity planning projects every six months or so. As part of this exercise, they forecast their needs for the subsequent two or three years. These retailers determine their peak workloads, add a cushion, and then provision the necessary hardware, software, and storage. The thing is, given the nature of the retail business, much of the capacity they provision is unused for most of the year. These retailers buy a lot of hardware, and license a lot of software for that hardware, just to handle peak workloads during “busy periods”. They have to pay for all of this capacity even when they do not use it during the “normal periods”. Because DB2 pureScale has daily-based pricing and because it is so easy to add and remove capacity, many retailers can now provision the software on-demand and only pay software license fees for the capacity they actually use. (Note that this lowers software costs, not hardware costs.) These companies are forecasting that using DB2 pureScale to add and remove capacity on-demand will free up significant amounts of IT budget.
- A large insurance company is talking to IBM about being able to handle large volumes of transactions at short notice. For instance, the insurance company needs to be able to process a high number of transactions after a particularly damaging hurricane or tornado. However, they cannot accurately predict the severity or timing of these natural events. As such, their approach has been to provision for the worst case scenario. But doing this has resulted in a large amount of their IT budget being tied up in servers and software that is not being used most of the time. Now DB2 pureScale allows them to recapture a significant amount of the IT budget that is spent on database licensing and maintenance and invest it in supporting the business in new and innovative ways that help them get ahead of the competition.
- It seems obvious now, but I didn’t realize that telecommunications companies encounter large spikes in transactional workload during holidays. All those calls to family and friends generate a lot of transactions on the back end. It should not surprise you that DB2 pureScale’s ability to add or remove capacity on-demand is generating a lot of interest from telecoms providers. Of course, the continuous availability enhancements in DB2 pureScale are also very important for telecoms providers. Again, the primary benefit for these companies is the cost savings involved in not paying for extra capacity when you are not actively using that capacity.
Today, IBM announced DB2 pureScale. I thought I’d take a few moments to tell you about DB2 pureScale. The first thing you should be aware of is that DB2 pureScale is the culmination of a multiple-year collaboration between the IBM DB2 and IBM Power Systems teams. Some people may initially think that this announcement is a reaction to recent publicity stunts from Oracle Corporation. But when you dig beneath the surface, you will discover that this is a tremendous piece of technology.
As you know, DB2 on System z is the industry gold standard for scalability, availability, and reliability. Nothing comes close to DB2 for z/OS. However, there are organizations who are not taking advantage of the unique capabilities of IBM System z. Often, such organizations have turned to database clusters on distributed systems, like Oracle RAC, to achieve their scalability and availability goals. If your organization attempted to use Oracle RAC, you are only too aware of the issues involved in deploying and updating RAC systems.
DB2 pureScale directly addresses these issues. For some time now, clients have been telling IBM that they need an easier solution for adding capacity to transactional systems. They want to get away from having to specially tune their database systems for use in clusters and from having to make their applications cluster-aware. DB2 pureScale directly addresses these issues by providing application transparency. With DB2 pureScale, you can easily add or remove nodes. You don’t need to re-tune the database systems or update your application code.
Not only is DB2 pureScale much easier to work with, but it also offers higher levels of availability and it offers greater levels of efficiency when adding capacity to your system. Higher levels of availability are made possible by adopting a design with 15 years of proven leadership on System z. These innovations basically ensure that more data is available to your applications for greater amounts of time, when compared with solutions from other vendors. The key to the increased levels of efficiency are the unique architecture and the use of Remote Direct Memory Access (RDMA) calls, which are much more efficient that IP-based calls.
Here is a link to a short video interview where I talk about DB2 pureScale:
If you want to get more information about DB2 pureScale, make sure to watch the upcoming DB2 Chat with the Labs Webcast on the 21st of October. This Webcast will feature experts from the IBM labs who will dive into the technical details of DB2 pureScale. They will explain the architecture and how the RDMA calls are so much more efficient.