List:Cluster« Previous MessageNext Message »
From:Devananda Date:November 5 2004 5:20pm
Subject:Re: Followup: New important section in manual
View as plain text  
Tomas Ulin wrote:

  > So here are the questions:
  >   0. Are you using/considering using MySQL Cluster for production,
reasearch, fun?       And if still considering, in what timeframe?

Testing and considering for use in production. Will use when disk-based
NDB tables are implemented and we feel the cluster provides sufficient
stability and crash recovery.

  >  1. What application or type of application are you using/intending
  >     to use MySQL Cluster for?  E.g.
  >   * Telecom Operations
  >   * Real-time subscriber information/profiles/authentication (Catalog
  >     Lookups, login/pwd, HLR databases)
  >   * Real-time billing information (usage records produced when a user
  >     uses a service, e.g. VOIP calls, CDRs)
  >   * Dynamic state data for online systems (session management)
  >   * Web applications (e-Commerce site, online shopping carts)
  >   * Real-Time Data Access for Finance (real-time financial information
  >     e.g. stock info,trading info, position info)
  >   * Real-Time Systems Monitoring (log aggregation, looking for
  >     patterns/particular values on a very dynamic datastore)
  >   * or other. What if so?

Real time info on users' names, permissions, account info, geographic
location. Essentially, every pageview that occurs on queries
our Personal database, and we are looking at the possibility of storing
that data in mysql cluster.

  >  2. What are the most important characteristics of your application 
(high availability, response time, scalability)?

Data safety, scalability, ability to handle high volume traffic. 
(response time)

  >  3. If you are considering MySQL Cluster and this is an existing
  >    * What database are you currently  using?    * What critical
database features are you using?
  >    * What features are missing in the current database?

Currently using Oracle 8i for this database, mysql (InnoDB) for most of
our other databases. We do not need anything too fancy, just high 
performance and stablity.

Oracle lacks usable data replication.

  >  4. Why are you using/considering MySQL Cluster (HA, performance,
low-cost, platform support, packaged solution)?

Replication, accessiblity (open source), cost.

  >  5. Are there any specific performance metrics you needed/are trying
to meet?

Transaction volume: 10,000+ queries per second
Read/write ratio: 90/10

  >  6. Do you have any specific HA metric expectations or requirements?

no (un)scheduled downtime

  >  7. What is the size of your database?

Number of tables: 1
Rows: 100,000,000 (large rows, >500b DataMem + >100b IndexMem per record)
We have many more tables, but are looking at what is best for our
primary table.

  >  8. What platforms do you need to support for?  (hardware, CPU,
operating system, other software)?

Linux on Intel

  >  9. What other databases are you considering?
  > 10.  What issues have been encountered while you've been using MySQL
Cluster so far?/
  >    * What version are you using (version, own build, built yourself)?
  >    * Installation issues:
  >    * Configuration issues:
  >    * Setup issues:
  >    * Bugs encountered:
  >    * Documentation issues:
  > 11. Anything keeping you from continuing development and/or deploying
on MySQL
  >      Cluster? (usability, stability, performance, features)

In order to put our primary table into MySQL Cluster, we would require
~60GB of storage space. See an analysis of this on the email list from
late August, under the title "Re: Data persistency". In those emails, it
was estimated at 45GB by Mikael; since then our user base has grown ~15%.
At this time, with the Cluster's requirement to store *all* data in RAM,
it would be unfeasible to store this table in the cluster while leaving
room for growth. Thus, we are waiting on the "Disk-based NDB tables"
feature to be added.

  > 12. Are there any critical features that MySQL Cluster does not
support that's
  >       keeping you from solving your problem? (Foreign Keys, Joins, 
Disk-based data,    Replication, BLOBs, etc)

Disk-based data! (Replication!!)

  > Thank you very much for your input,
  > Tomas Ulin

Thanks to all the developers (and list members!) for the great work and 
awesome support :)

Devananda van der Veen
System Administration Programmer
NeoPets, Inc.

Followup: New important section in manualTomas Ulin3 Nov
  • Re: Followup: New important section in manualNick Sullivan4 Nov
  • Re: Followup: New important section in manualRussell E Glaue4 Nov
  • Re: Followup: New important section in manualTomas Ulin4 Nov
  • Re: Followup: New important section in manualDevananda5 Nov