You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@trafodion.apache.org by Nieyuanyuan <ni...@huawei.com> on 2015/09/14 08:22:17 UTC

How's Trafodion/NonStop SQL's Performance compared to Other Databases like MySQL/PostgreSQL/Oracle

Hi, Hans, Narendra,

Looks like we claimed that EsgynDB is targeting for OLTP workloads in Hadoop ecosystem which makes it special, compared to other open source solutions like Hawq, Impala, etc.

My question is, do we have any official or non-official performance testing reports based on Trafodion or even its predecessor - NonStop SQL/NeoView? Especially the comparison between it and other famous OLTP databases, like MySQL/PostgreSQL/Oracle.

Thanks.

Re: How's Trafodion/NonStop SQL's Performance compared to Other Databases like MySQL/PostgreSQL/Oracle

Posted by Rohit Jain <ro...@esgyn.com>.
Nieyuanyuan,


We do not run competitive benchmarks using artificial benchmarks, such as TPC or YCSB. Most databases are optimized to run well on artificial benchmarks. Most vendors use these artificial benchmarks internally to assess performance, as do we.  We run them internally to assess how our product is improving for certain workloads — OLTP, operational, query / reporting — from one release to another.  It makes no sense for anyone to use these benchmarks to compare products when they are making a decision which database to go forward with.  These numbers are rigged, databases are fine-tuned for them, and so they have very little relevance to how that database will perform for your specific workloads.  The strength of an optimizer and execution engine for any database is in how it can handle varied workloads that are thrown at it.  While benchmarks like the TPC-DS attempt to do this, they fail in being representative of customer specific workloads. Which is why your own assessment is relevant.  Customers need to spend the time and effort to understand what workloads they will deploy, and create a representative POC that tests products on their response times, scalability, concurrency, availability, data integrity, features, … all requirements relevant to them to meet their service level objectives.  

The other aspect of a database engine is its architecture.  This often reveals the strength, flexibility, agility, extensibility, ability to handle mixed workloads, of a database product.  Unfortunately many customers do not take the time to do that or don’t understand enough about database architectures to make such an assessment.  We challenge any other database out there in how our database engine is architected for performance, scalability, availability, concurrency, data integrity, and so on.  You have to build your house on a solid foundation and a strong architecture.  That is what we have done in what you see in Trafodion.

So it is up to you to spend the effort and time to make such as assessment for what is important to what you are trying to accomplish. If you do it right, you will conclude that Trafodion is one of the best database engines in the relational proprietary and open source worlds of Hadoop.  Of course, every database engine has to be tuned.  We just replaced our storage engine with HBase.  So we are going through that tuning process and have come a long ways in the past year.  But that is where you have to make an assessment of the architecture for what the potential of a technology is, with enough evidence of what it can do today.  

Rohit








On 9/14/15, 1:22 AM, "Nieyuanyuan" <ni...@huawei.com> wrote:

>Hi, Hans, Narendra,
>
>Looks like we claimed that EsgynDB is targeting for OLTP workloads in Hadoop ecosystem which makes it special, compared to other open source solutions like Hawq, Impala, etc.
>
>My question is, do we have any official or non-official performance testing reports based on Trafodion or even its predecessor - NonStop SQL/NeoView? Especially the comparison between it and other famous OLTP databases, like MySQL/PostgreSQL/Oracle.
>
>Thanks.