Does Oracle Have a Winner in Exadata V2?
Could Oracle's Exadata V2 enable a new -- and altogether more ambitious -- approach to enterprise performance management?
- By Stephen Swoyer
- July 21, 2010
Veteran industry watcher Merv Adrian recently took an in-depth look at Oracle Corp.'s Exadata version 2.0 -- aka, the Oracle Sun Database Machine.
Adrian liked what he saw. He says the revamped offering's combination of features -- cutting edge storage (SSD and Flash Cache), high-performance I/O (which exploits solid state storage to reduce latencies and increase I/O operations), and a quasi-analytic database architecture (Oracle's "Hybrid Columnar Compression") -- achieves a kind of gestalt effect in which the Exadata V2 whole is considerably greater than the sum of its parts.
"The real 'secret sauce' of Oracle Exadata V2 is the way in which these technologies complement each other to deliver additional performance and scalability," writes Adrian, a principal with IT Market Strategy.
"Traditionally, server, storage, and database technologies from different vendors achieved excellent optimization within their own spheres, but that optimization often worked at cross-purposes with other parts of the data-management architecture," he continues. "[D]atabase and storage vendors have traditionally optimized data storage for different user needs, so that a [SAN] … solution aimed at remote-PC document and video data could undercut the optimization of a data-warehouse data store."
Exadata V2 serves up a more synergistic arrangement, Adrian notes. This could help explain Oracle's extremely optimistic performance claims.
"In querying, the [Exadata] database engine works in concert with the storage solution, with server and storage being allocated query tasks in a version of load balancing across CPU, SSD, and disk-array processing," he explains. "Storage allocation and compression are integrated between the storage manager and the database engine. The CPU can operate on less data, as retrieval is assisted by pushdown processing in the storage layer."
Adrian also lauds Exadata's workload management capability. Data warehousing stalwart Teradata Corp. likes to trumpet its own workload management expertise, which it says permits its DW platform accommodate a mix of different workloads -- and prioritize certain workloads based on the roles, responsibilities, or needs of users. For a while now, rival Hewlett-Packard Co. (HP) has been making similar claims about its Neoview DW platform. More recently, vendors such as analytic database specialists Kognitio, Netezza Inc., and Vertica Inc. have trumpeted their own mixed workload bona fides. Although none of these players (with the exception of HP) claims to offer Teradata-like workload management, all claim to achieve good enough service (in some cases, they claim, by virtue of the superior speed of their analytic database platforms) for most users.
Oracle, too, likes to claim that it can consider the goods -- i.e., adequate query response times and availability levels that are appropriate to user classes or roles -- mixed query-wise. If nothing else, Adrian writes, Exadata V2 seems to have the makings of a mixed workload performer.
"[W]hile complex-query data-analysis solutions use hybrid-columnar technology up and down the integrated stack to deliver deeper analysis on smaller, compressed data sets that have been rapidly pre-filtered by smart storage, Oracle Exadata's OLTP write-heavy solutions can take advantage of the new SSDs to process transactions faster, with fewer buffer flushes, and less forced swapping to disk," he points out.
Adrian offers an assessment of Exadata V2 as part of a new wave of category-inaugurating devices. "[T]he potential of Oracle Exadata lies in its ability to do what is already being done, but faster and cheaper, or in the assurance [for existing or prospective customers] that they may now be able to keep ahead of ever-growing demands for more data," he comments. "But an equally significant -- and perhaps more exciting -- use of Oracle Exadata is the promise of using its raw power to do something that hasn't been done before. One example of this might be Enterprise Performance Management."
EPM, according to Adrian, has a much more ambitious scope than vanilla business or corporate performance management.
"Twenty years ago, it might take a month to do a yearly plan for a large-scale enterprise, and a month to record the quarterly 'actuals' and come out with a quarterly set of balance sheet, income statement, and cash analysis," he explains. "Today, consolidation of weekly results from a 150-line-of-business global enterprise, followed by comparison with plan, replanning, and rebudgeting, can be done in less than a working day."
The takeaway? New and hitherto unimaginable responsiveness, according to Adrian. "This speedup means that enterprise-wide plans are not cast in concrete for a full year; they can be modified, drastically if necessary, on a monthly or even weekly basis, when the organization changes strategy in midstream," he continues. "But the real new EPM application potential is in using deeper what-if analysis during the consolidation process to drive more granular plan/budget modifications." In other words, Adrian concludes, an offering like Exadata makes it possible for "EPM to move surprisingly close to real-time reaction to events."