Comparing cost and performance for Oracle Database, EnterpriseDB and Tibero Database

Hello and welcome! These are challenging and unpredictable times, more than ever your company needs a database solution that provides value for money. In this slide we emphasize that data access through SQL “SELECT” is the time consuming operation typically 99% of the database engine’s processing time is spent on this operation. Note that SELECTs happen implicitly for most UPDATEs and DELETEs which are also very common operations in your application so how quickly your database processes selects will determine your application’s efficiency the productivity of your employees and therefore the company’s profitability clearly fast select access to data cannot come at any cost so in this video presentation we will be looking at a very interesting cost performance comparison between four key database solution providers these are tibero oracle enterprise edition oracle exadata and enterprise db postgresql throughout this presentation we will use the acronym cpr for cost performance ratio the overall cost you consider should not simply be the initial price of the product and its service fees the real cost needs to take into account the important effect of the database performance if your employee clicks for a report and then waits 30 minutes for it to finish then this is effectively a downtime of 30 minutes for that employee regular downtime for many employees means damaging lost productivity as importantly if you have long running reports or workloads these will consume more cloud computing resources and your monthly bill will be higher you will harm your company by purchasing a database solution with performance disadvantages in this slide we show the environments we used for the tests you can see real world configurations from four different customers and the price they paid for each configuration over five years we chose five years because it is important to not merely look at initial purchase cost but to add in the license fee costs over the longer term the infrastructure and software configuration hosting each database represent the likely real world scenarios you can see the actual costs that would be paid by the customer over a five years operation this includes hardware software and ongoing yearly licenses so what approach did we take to compare database performance and give you a cost price ratio for those different solutions well we firstly grabbed a 128 gigabyte data set from oracle db we migrated this to each of the different database environments in typical configurations of memory and cpu then we executed and timed six sql statements that represent normal day-to-day operations for any database the software tests were identical in each environment and care was taken to ensure there were no competing processes so here’s what you’ve been waiting for the results let me navigate you through the results and their interpretation in this table we are looking at how the same queries perform for each solution each row represents a different sql test and you can find the exact sql statements in a later slide each column heading shows you a client’s solution in terms of database memory and cpu configuration just to clarify faster response times inside the table correspond to a lower number of seconds so client a represents a big german company that owns a regular oracle enterprise edition its timings range from 38 seconds to an execution time so large that it didn’t finish it’s important to compare these timings with those of oracle atp and oracle adw the timings for these two were far better but their total cost is considerably more expensive and for all three clients we are still talking about the same database provider oracle so now let’s compare oracle solutions with other database vendors take a look at the timings for the regular oracle enterprise edition and compare this with the timings for the enterprise db postgresql scaled out solution well edb performed three to ten times worse than this oracle solution whilst its five-year cost of two million dollars is about 15 times more expensive so it is worse than regular oracle on both key measures we can say that the delete operation for edb did succeed when oracles didn’t however it took nearly six hours to delete the data and this isn’t acceptable in real life scenarios oracle ee regular solution has a better cost price performance than the edb solution now we again compare oracle enterprise edition but this time with tibero database oracle ee has a five-year cost three times the cost of tobero however if you compare the timings you can see that the tibero performance was considerably better around 30 times faster tibero database clearly has a much better cpr than oracle enterprise edition so what happens when we move to clients using much more powerful oracle configurations first take a look at the much publicized oracle exadata first the autonomous data warehouse version this is the best oracle solution for olap and its five-year costs amount to an eye-watering 4.4 million dollars its performance figures for read-only selects were good if we compare its timings to those for the edb solution then oracle has at least 30 times better performance and is only double the five year cost so oracle adw has a much better cpr than edb so now let’s compare oracle adw with Tibero what is interesting to observe is that the Tibero client had similar timings to oracle adw and they paid several times less in fact on a totally different budget range Tibero also significantly outperformed oracle adw on the delete operation the cpr for tubero is better than even oracle’s best solution for olap we can now move on to look at exadata oracle autonomous transaction processor this is oracle’s flagship solution for applications that are heavy on oltp here you can see that some of the queries ran twice as fast as tuberos but some for example query 1 ran slower the update from oracle atp ran in just over half the time of tuberose this is only a partial success for oracle however as its rollback time was nearly five minutes whereas tiberos was just 204 seconds a very similar timing to the update itself and most notable of all is that Tibero managed the delete in just over 10 minutes whereas oracle’s delete took an enormous 62.5 minutes… :)… some friendly advice: Don’t press the delete button in your oracle atp application folks! this oracle solution is over 10 times more costly than the Tibero solution with the best oracle timings only twice as fast so Tibero has a better cpr we now compare oracle atp with enterprise db oracle atp performed 20 to over 100 times faster than edb and its cost is only three times as much so oracle atp has a better cpr than enterprise db we will explain in a later slide why a so-called open source EDB (closed-source) which in reality is a commercial database cost two million dollars in this slide we see what performance can be achieved with the open source model and also examine the true cost of open source trends EDB trial represents the free offering from EDB as you can see very little could be achieved in this environment the database engine is simply lacking in performance and could execute only one of the queries and still took over 35 minutes for this when Tibero and oracle took just over a second so a client would conclude quickly that having bought the free version they still have to pay for an upgraded EDB database so for our next test we chose EDB cloud database service (CDS) we use the maximum IOPS allowed for a deployment which is 20.

000 and you should provision a 09:09 - database of at least 660 gigabytes to select that maximum iops edb cloud database service is quite new and we cannot find any of our customers using it so we became a customer ourselves to perform the test and we will describe in other videos how the tests were performed so back to the EDB Cloud test results and you can see that the smallest infrastructure produced results as bad as the EDB free trial ones once we loaded the data we made a backup and restored it into the biggest possible machine in EDB CDS. Here we did get the sql queries to actually complete represented by customer X in the table but the query timings were the same as the EDB non-cloud in the previous slide and so equally poor so what happens when we scale out by adding five replicas to the more expensive EDB Cloud infrastructures represented by customer Y, you would hope that by increasing the memory cpu and cost five times the performance would significantly improve but NO the timings were the same this is in fact easily explained EDB’s architecture can load balance read-only workloads but NOT read write operations plain common sense should tell you that if you are considering a move from Oracle Cloud to EDB Cloud you need to think twice because in the cloud the performance differences are huge for some scenarios so you save on the software part but then you spend much more on the virtual hardware side on the cloud trying to get the performance to anything close to what you had with Oracle to get any kind of basic performance with EDB you will have to upgrade to their best product at a five-year cost of two million dollars but as our results show the performance will still lag far far behind that of Oracle and Tibero you will also notice a client in column T we wanted to ensure a large infrastructure wasn’t skewing results in favor of tibero so we reduced its server’s memory and cpus considerably you can see that the performance was almost as good as the high specification infra this shows how powerful Tibero’s database engine is maybe you’ve only considered Oracle and EDB and so far you haven’t had time to test Tibero well you should be aware that we are an Automation business and if you do just a few clicks on our Dimensigon website We can provide you with a Tibero install on the cloud also we have access to DELL EMC infrastructure where these tests were executed so let’s interpret the results and see what you need to consider for the “Long Run” what would you do if you had to invest in a database solution right now: would you spend on Oracle products knowing that the performance is similar with Tibero which is a fraction of the cost? or would you invest into a “fake open source” solution like EDB? and then still expend a great deal of money for when you need to scale up your infrastructure well we have seen through the slides that the performance was extremely bad for EDB despite the time we invested in database tuning… Yes, our database experts tried to get better timings but unfortunately even with this tuning. EDB was far behind the competition. You need to take into consideration the queries inside your database can be unexpected and you might not have the chance to optimize all operations unless you are paying a high cost for an extremely good DBA team so it is clear that the total cost of ownership should also be lower for Oracle and Tibero than EDB because their databases (Oracle/Tibero) are running more smoothly when they encounter day-to-day unexpected operations put simply day-to-day costs are lower for Oracle and Tibero than for EDB(EnterpriseDB) because you require less tuning and database team expenditure the leading runners are Oracle and Tibero with EDB barely off the starting blocks from the results we discussed Tibero provides good performance at a cost many times lower than Oracle’s major OLAP and OLTP solutions by the way our other tests show that Tibero is 10 times faster than Aurora(AWS) and 100 times faster for mixed workloads of reporting and transactions Tibero is a clear winner in the cost price performance race In conclusion we have shown you the results and demonstrated that Tibero has the best Cost-Performance relationship We took every care to do the test fairly and accurately but we want you to feel totally confident about the comparisons so we will soon publish the test procedures but if you want them right now just write a comment in this video or contact us through our website (www.dimensigon.com) if you want to have a LIVE demo (Hands-on + Interactive) we still have access to some of the systems although some are owned by customers and we cannot perform the test at the exact time we want.

14:14 - If you can’t wait for the in detail calculation slides and video and how the performance was gathered. You can contact us as well. => www.dimensigon.com We will forward to you the calculation slides directly Feel free to leave your comment in this video and please follow up for updates (LinkedIn / YouTube / Twitter) we will publish more interesting videos about database innovation, database elasticity in the Cloud or Automation. We hope you enjoyed this presentation and we send you the very best wishes! Cheers! .