Transcript of BriefingsDirect podcast recorded at the Oracle OpenWorld Conference in San Francisco the week of Sept. 22, 2008.
Listen to the podcast. Download the podcast. Find it on iTunes/iPod. Learn more. Sponsor: Hewlett-Packard.
Dana Gardner: Hi, this is Dana Gardner, principal analyst at Interarbor Solutions, and you're listening to a special BriefingsDirect Podcast recorded at the Oracle OpenWorld Conference in San Francisco. We are here the week of Sept. 22, 2008. This HP Live! Podcast is sponsored by Hewlett-Packard, and distributed through the BriefingsDirect Network.
Today we are going to discuss a large and an impactful product announcement at Oracle OpenWorld that took place on Sept. 24. It was the introduction of appliances in a cooperative relationship between HP and Oracle to create some of the most high performing databases and date warehouses in history. We are going to talking about the Oracle Exadata Storage Server and -- when put together in a very impressive configuration -- what becomes the HP Oracle Database Machine.
Here to help us understand how these impressive server configurations and high-speed, extreme-performance databases came together, we are joined by Rich Palmer, the director of technology and strategy for industry standard servers at HP. We are also joined by Willie Hardie, vice president of Oracle database product marketing. Welcome to the show, Willie.
Willie Hardie: Good to be here, Dana.
Gardner: Tell me a little bit about this very momentous announcement. This has been several years in the making, but it’s not just a product announcement. It seems like an architectural shift, and also an alliance and partnership shift in terms of the cooperation between a hardware provider, in this case HP, and Oracle, until now purely a software company.
Hardie: That’s an excellent question. So what we actually announced this week is the Oracle Exadata Storage Server. Now, the Oracle Exadata Storage Server is an intelligent storage device. We’ve basically taken industry standard hardware and storage components from HP, and we’ve combined that with smart intelligence software from Oracle that allows us to offload query processing from the database servers to the storage servers.
So now they can do a lot of the work for us, to allow the stripping off of the rows and columns that we require, and push last data backups through much wider networks.
Gardner: For those of us who are not computer scientists, but are nonetheless interested in the outcomes, architecturally we are putting the intelligence that we usually have in a database server in very close proximity to the data storage itself, connecting that through a very fat pipe in the form of InfiniBand. And, in essence, parallel processing comes to bear, because of the proximity. Is that correct?
Hardie: Absolutely. So what we are able to do for the first time ever is we can use these storage devices to actually do the query processing itself. So the more that the storage server processes and we compute into our configuration, the more of the workload they can take off, which traditionally is done at the database server.
Gardner: Let’s go to Rich Palmer at HP. Tell us a little bit about the history. How did this come about, and what is it that HP has been doing to improve upon the performance of this long-term database lineage?
Rich Palmer: If you look at HP and Oracle as partners in this industry, we have a long-standing history together. We have several reference configurations, more than 50 reference configurations that we do with industry standard hardware and Oracle solutions, which we’ve been delivering for many years now.
Going back all the way to the introduction of Oracle Real Application Clusters (RAC), and even before RAC introductions, the history of the two companies really stems from two leadership positions. HP does more servers on Oracle than any other company. Oracle does more data warehouses than any other company. You bring those two forces together, and you get a very strong formidable entry into this data warehouse appliance market.
Where HP and Oracle really started this discussion stems back a couple of years, and it really became a trend in the market of bringing data and server processing power closer together; that trend has escalated over the last couple of years -- especially as so much data has been growing at exponential rates, every single year. What we found is that, you cannot push so much data over a traditional storage fabric. This new technology allows us to do that.
Gardner: And we are talking about very large data sets, of terabytes and larger, right?
Palmer: Enormous Data sets. Let me give you an example, and I think we are all very familiar with this example. We all use cell phones in today’s industry. Every one of those cell phone calls is a database record somewhere, be it on AT&T’s database or T-Mobile’s database or whomever's database -- they store that data. Now, when they are storing that data, sometimes they are going to want to move it. If you have a narrow pipe to push that data down, and you’re bringing back enormous amounts of data that is erroneous, and you don’t need the other data; all you need is just for what you’re looking for in the query.
So this process allows us to push just the query information across that pipe. Less data over the pipe, a wider pipe, and your performance goes up dramatically.
Gardner: Okay, so let’s unpack this a little bit. We’ve established that the marketplace is demanding better performance, particularly in the use of large data sets, 1 terabyte and larger up to 10 terabytes, and size often. That requires the movement of very large sets of data, and the inhibitor here was the storage’s physical capacity, and ability to deliver the data.
So you’ve re-architected, and we brought together two companies to work together. This brings the question: Why hasn’t the hardware and software duality gotten closer before this? Why now?
Palmer: In this market, it’s constantly evolving to a state where you have to bring software tools to the table, and you have to bring high-performance hardware to the table. The evolution of both of those have hit at the perfect time in the last year.
Oracle has been developing the software code for several years now, and HP has been working on the hardware side of this equation to bring together the two forces at this time. We are using industry standard technology, so it’s not something that we are the only hardware guys out there with InfiniBand, and InfiniBand is an evolving technology. But the performance of InfiniBand is at a point now where we can actually leverage it using Oracle software to offload the storage processing from the database server. Those are the two key components -- it’s not just the hardware, and it’s not just the software. You have to marry these two things together.
So why hasn’t it been done in the past? Well, it has to some degree, there are others who had tried to do this, but they haven’t done both. They haven’t been able to achieve both facets, and that’s really why this is the right product at the right time.
Gardner: Okay, Willie, let’s get into the actual product itself. Explain to me what the Oracle Exadata Storage Server actually is? What are we talking about?
Hardie: You see that the Oracle Exadata Storage Server is basically comprised of an industry standard HP DL180 Storage Server. So inside this storage server we have 12 3.5-inch disks to be 12 SATA drives. We have two Intel quad-core processors. We have 8 gigabytes of memory; we have two InfiniBand network connections, and dual power supplies.
So in this storage server we have a lot of storage capacity, we have a lot of processing power, and we have a lot of network bandwidth. Then the real secret sauce here is this intelligence software from Oracle that’s installed into each and every one of those devices. It’s this intelligent software that enables us to offload this query processing, which makes the Oracle Exadata Storage Server really unique.
Gardner: Okay, let's dumb this down a little bit in simplistic terms. Instead of large data sets moving from storage to the database and back, what happens differently now?
Hardie: What happens differently now is, because we are offloading the query processing at the storage server, the storage server can strip out the columns that we don’t need, strips out the rows we don’t need, returns a subset of data back up through this wide InfiniBand network. That’s what makes the difference. We are treating a much smaller data set that we pass up through this network, and the database server can just finish off that query processing much faster than it ever could previously.
Palmer: One of the other values that we achieve here is certainly in the data passing back and forth, or less data over a wider pipe. So you’re going to get exponentially better performance. Now at the storage servers you’ve taken the processing power of doing the query right at the disks, and in every one of these storage servers you have eight cores, these are Intel quad core processors, two of them in each servers, and so you have eight cores on the input/output (I/O) path directly to the disk.
So there is no external I/O going to your disks. Traditionally you’ve had to go outside of the server, go to the disk that is across the fabric -- and everyone else is sharing that fabric.
So you have many people sharing a fabric, versus now you have a dedicated fabric inside of the server. So it’s a copper-to-copper connection inside the server. Those disks are right on top of the processor. That is really the essence of it -- you can pull the data off of this rapidly because it’s all so much faster. As Willie indicated, you can strip out all the unnecessary data and pass a much smaller data set over a much wider pipe, back to your database servers. There are so many levels of performance improvement here.
Gardner: And to your point on the secret sauce -- you are also taking advantage of all those cores via multiple threads, and the software has been a deeply tuned to take advantage of those multiple threads in a concurrent fashion.
Hardie: Oh, absolutely, and Rich touched on that as well.
Palmer: When we add more Exadata Storage Servers into our configuration we can take advantage, not just that additional storage capacity, but we can now take advantage of that additional processing capability -- to own that storage layer, which is a big, big difference.
Gardner: And at the announcement here, Oracle Chairman and CEO Larry Ellison described use cases where improvement typically was 10x to up to 72x over what has been the industry benchmark.
Hardie: Absolutely, when you actually cut away the technology and look at this from a business perspective, what it means for me as a business user -- it means that when you’re accessing those data warehouses that Rich was talking about earlier -- like a call data record -- data warehouse have billions of rules additionally. What this means, when you’re accessing those, your queries are going to run much, much faster than they ever did previously. Not only will they run faster, you can have much more queries and more long-running queries concurrently. That’s what is going to be making the big difference.
So when we hear of customers talking about getting 20x performance, improving 30x performance in one particular instance; in one particular query, 72x performance -- that is extreme performance improvements, in anybody’s measurement.
Gardner: Okay, so we have this engine, this Oracle Exadata Storage Server. We also a new announcement, the HP Oracle Database Machine. Tell me how one relates to the other.
Palmer: The HP Oracle Database Machine is a single rack that contains everything you need to run a large data warehouse. It contains eight ProLiant servers running Oracle Database 11g and RAC. It has four InfiniBand network switches and it has 14 of these Oracle Exadata Storage Servers that we talked about earlier. So in a single unit you have everything you need, ready to load up your data and start running your business queries right away.
Gardner: Tell us a little bit, Rich, about this 42-slot rack configuration and why it’s right for the market now?
Palmer: Well, so if you look at the market in data warehousing, the appliance type of delivery is a much simpler deployment of hardware and software configurations. That is emerging as a high-growth area in data warehousing. So with this market trend that’s going on between HP and Oracle, we’ve been able to come together and put everything in customers’ needs in one box. We put it at the customer’s site, and that’s on a global basis.
If you look at HP, one of the strengths that HP brings to this relationship is our ability to distribute and deliver globally. We build all of these database servers or database machines in regions around the globe. They are not just built here in the United States; they are built in United States, they are built in Singapore, they are built in Scotland, and then they are delivered to those regions on a worldwide basis.
So this ability of HP to build the product from the ground up to an exact specification, deliver to the customer, install at to customer's site, and then have Oracle come in and tune the software to make sure it's optimally configured -- that is a no-lose environment. We have the ability here to deliver an appliance-like stack of hardware, put the right software set on that hardware, and target a customer's need for simplicity, high performance, and data reliability -- all in one box.
Gardner: Okay, we've described the marketplace need, the size of data pushing the envelope. Now we are re-architecting to adjust to that. We've described the subset, which is the Exadata Server, and then the configuration, which is the racked Machine. Now, what kind of organizations are going to be interested in having the forklift upgrade to this, bring it right in, drop it in, pre-configured, optimized, and what are they going to do with it? Is this for business intelligence (BI), is this for simply managing scale? What are the speeds that this now provides going to do for companies to improve, or to change, how they do business?
Hardie: The organizations that are going to be interested in Oracle Exadata Storage Server and the HP Oracle Database Machine are those primarily interested in large data warehouses. And by large data warehouses we're talking into the (terabytes and petabytes) and beyond. Now if you look at the organizations that are typically dependent on very large data warehouses, it's organizations that Rich mentioned earlier, the telcos could be an obvious one, call data records, retail organization, very much dependent on analyzing point of sales (POS) transactions. You look at other organizations like trading systems, massive amount of transactions flow through these systems on a daily basis.
Gardner: Especially these days.
Hardie: Absolutely. It is really important to understand what's going on with these transactions, and to make informed business decisions. The beauty of this is you have completely scalable infrastructure from a storage point of view. But more importantly, you've got completely scalable infrastructure from a query performance point of view. As you store more call data records into these systems, more POS transactions, more stock transactions into these systems, you're not going to deteriorate your query performance at all. The more hardware, the more storage servers you put into these systems, the better your performance is going to be.
Gardner: Now that I have this capability to bang on this thing, so to speak, in more ways without the degrading performance, in what ways do you expect these companies to actually "bang" on this? Is this going to provide new and higher level of business intelligence querying? Is this going to provide higher-order analytics? Are there going to more business applications that can derive near real-time data and analytics from this? All of the above? What's the qualitative payback?
Hardie: There is definitely an element of "all of the above." Let me give you some of the examples of some of the queries that customers have actually been experiencing using the Oracle Exadata Storage Server. This probably fits into the context pretty well. You have organizations out there, retail organizations, telcos, for example. You know, some of the queries they are running are literally running for over half an hour. In some cases it is hours.
Moving to this new architecture is bringing down these execution times. One particular example, a query that was running for over 30 minutes is now running in under 30 seconds. It's that scale of improvement. Now when you can set your terminal, your laptop, or your mobile device and then kick off a query and get an answer within seconds -- then you're going to do more of these. If you know that when you kick off a query it is going to take 30 seconds to return it, you're going to pick more times when you choose to kick that off. You don't have to worry timing that anymore. You can just ask queries when you like, and expect to get a quick answer.
Palmer: Willie, I think you are absolutely right. The ability to capture business information has accelerated so much because of this technology. There are customers that cannot access data records beyond a certain time period simply because of the massive size of those data records, or because of how long a query would take to access a historical group of data. That all goes away now.
Now you have the ability. Historically you might have been able to look at the last week's worth of retail records, or medical records. Now you have the ability to go and look at years and years of data in the same timeframe that you were looking at weeks of data, and query a much bigger dataset, because of this architecture. That's a big business value, because now I can trend my business in a much more effective way. I'm putting more productivity tools in the hands of the user, so that they can actually turn data queries and business intelligence back into a fundamental element of growing their business and being more competitive in their markets.
Gardner: I imagine this will also compel companies to put even more data and information into these warehouses, because they are not going to degrade the performance of these essential queries. They are also going to able to do more types of queries. And, again, we're improving the quality and breadth of the data types, but still getting even better performance. So it's sort of a qualitative improvement on many different dimensions.
Hardie: It's a qualitative improvement, and it's a quantitative. I mean, you're absolutely right. Organizations today are more and more dependent on faster access to better information. It's just as simple as that.
Gardner: We've talked about the types of organizations that we'll use this now in its current configuration. I expected this re-architecting of the database and the storage will also move down market a bit. What possible other use-case scenarios do you envision for leveraging this technology beyond the high-end of the market, into other areas of the market?
Palmer: If you look at some of the growing and emerging markets today, just think of cloud computing and all of the massive amounts of data that we're storing in other locations on the Internet, or through a paid service, and the massive amounts of storage that's being deployed for those types of applications. That's not going to slow down at all. This allows us through the Database Machine to go in and drop in a configured environment for that workload, specifically dedicated to a workload.
You can now scale this product by connecting multiple racks together, you can now scale just the storage component, if the processing side of the database environment is sufficient. You can now just scale the storage nodes, so it is a scalable grid architecture that can grow on the fly. So cloud computing is a very good example where we really don't know what the upper limit of that storage is going to be. So deploy a configuration, say, on a HP Oracle Database Machine and then grow it as your needs grow. This is one application where we know this is going to succeed.
Gardner: Willie, we're also aware that organizations will just want the Oracle Exadata Storage Server. They might have their own environments, their own preference for configuring what's available to them, and what would become available to them in the future.
Hardie: Any organization that wants to run their data warehouse on the Oracle Exadata Storage Server -- all they have to do is buy the Oracle Exadata Storage Server. It's just as simple as that. Oracle and HP of long given customers a choice of configurable options. So if customer feels that something like HP Oracle Database Machine is not the right fit for their organization, if it does not fit the standard needs for their organization, then they have the option of buying the individual components, the Oracle Exadata Storage Server, the InfiniBand connectors, connecting to the database servers, they have that option.
Gardner: Looking at this again through how to get started, where do organizations go? Now that this is available immediately, both of these configurations, is the sales happening through both HP and Oracle?
Palmer: It's a cooperative effort, but Oracle is leading the sales process. So the Oracle sales representatives on a global basis are leading this process, and HP is certainly as their partner going to join with them and make sure that the customer receives the best from both companies.
Gardner: HP is going to service the hardware, but the support comes through Oracle, is that correct?
Hardie: Oracle is the first point of contact if you want to buy an Oracle Exadata Storage Server, Oracle is your first point of contact. So talk to your local Oracle sales representatives. If you do decide to buy one, and you want to resolve a support issue, you call Oracle, and Oracle will bring in HP as and when required to resolve any issues.
Gardner: To sum up a little bit, for those folks who perhaps are a few steps removed from the IT department, who are doing queries, or using business applications, what's the big take away for them? What about this announcement is going to change their world?
Hardie: For these types of users you just mentioned, a little bit or a couple of steps removed from the IT department ... To be quite honest, they don't really care what their systems run on. What they are interested in is getting fast answers to their business queries. It's just simple as that. So when these business users know that they can get instantaneous response times, they can get real extreme performance of their date warehouse, or of their business intelligence applications -- that's what's going to make a big difference for them.
Gardner: Rich, at HP, let me flip the question to you. For those people inside the IT department, who want to come in Monday morning without big headaches, what is this new configuration and architectural approach mean for them?
Palmer: Simplicity, higher performance, the ability to increase their service level agreements (SLAs) with their customers in the warehousing world. This is a solution built on industry standard hardware, with Oracle software that is just well accepted in the industry as an enterprise software leader. The IT departments are very comfortable with both of those facts. They're very comfortable with HP; they're very comfortable with Oracle. Putting the two together is a natural event for any IT manager.
Gardner: We've been talking about a large and impactful announcement here at Oracle OpenWorld, the introduction of the Oracle Exadata Storage Server -- the first hardware product from Oracle. Isn't that right?
Gardner: We've also looked at the configuration of those Exadata servers into the HP Oracle Database Machine, which is in effect a data warehouse appliance. Joining us to help explain this, we have been happy to have Rich Palmer, director of technology and strategy in the industry standard servers group at HP. And also Willie Hardie, vice president of Oracle database product marketing. Thanks to you both.
Hardie: Thank you, Dana.
Palmer: Thank you very much, Dana.
Gardner: Our conversation comes to you today through a sponsored HP Live! Podcast from the Oracle OpenWorld Conference in San Francisco. Look for other podcasts from this HP Live! event series at hp.com, as well as via the BriefingsDirect Network. I'd like to thank our producers on today's show, Fred Bals and Kate Whalen.
I am Dana Gardner, principal analyst at Interarbor Solutions. Thanks for listening, and come back next time for more in-depth podcasts on enterprise IT topics and strategies. Bye for now.
Listen to the podcast. Download the podcast. Find it on iTunes/iPod. Learn more. Sponsor: Hewlett-Packard.
Transcript of BriefingsDirect podcast recorded at the Oracle OpenWorld Conference in San Francisco. Copyright Interarbor Solutions, LLC, 2005-2008. All rights reserved.