Monolithic server HPCC/VM
Sorry, this question fits into the Migration area but it is not related to Hadoop.
We are looking to convert our Oracle system to HPCC. Our business handles dozens of clients locally but we have remote deployments for several clients too. Some of the remote clients already run Oracle, but some purchased servers specifically to run our Oracle-based solutions.
I'm trying to formulate a plan for the existing remote clients, hopefully one that involves re-using their existing hardware to run HPCC.
I realize HPCC is designed to be clustered but has anyone performed a head-to-head comparison between a DBMS and HPCC, where HPCC is running on the same single-server hardware as the DBMS? Something like 16-24 cores and 64+GB RAM.
If not, does the VM have any built-in restrictions that would prevent me from testing this myself?
Thanks
We are looking to convert our Oracle system to HPCC. Our business handles dozens of clients locally but we have remote deployments for several clients too. Some of the remote clients already run Oracle, but some purchased servers specifically to run our Oracle-based solutions.
I'm trying to formulate a plan for the existing remote clients, hopefully one that involves re-using their existing hardware to run HPCC.
I realize HPCC is designed to be clustered but has anyone performed a head-to-head comparison between a DBMS and HPCC, where HPCC is running on the same single-server hardware as the DBMS? Something like 16-24 cores and 64+GB RAM.
If not, does the VM have any built-in restrictions that would prevent me from testing this myself?
Thanks
- aintnomyth
- Posts: 86
- Joined: Wed Jul 13, 2011 7:40 pm
Hello, thanks for the post. We have moved it to this new category based on your input. Thank you!
- admin
- Site Admin
- Posts: 209
- Joined: Thu Jan 27, 2011 10:58 am
I don't know of anyone -quite- having done that. We did do some extensive testing a couple of years back with a 3 node HPCC against a 3 node Oracle cluster; we were 9.something x faster (I remember AE complaining that we hadn't hit 10x)
The 'default' Linux install is single machine - I have it running on 6 CPU & 4CPU machines. It defaults to a single node thor and roxie - that CAN have advantages (no skew etc) - although we have sometimes found that running multiple thor 'nodes' on a single machine (with multiple CPUs) can provide some lift too.
One thing you don't mention is your disks. Depending upon what you are doing Roxie can do quite a few disk hits - if you only have one or two disks then you will need to be a little careful in how you lay out your roxie keys - not a problem - but you need to plan.
The only other thing I can think of is that (obviously) only having one machine rather limits your redundancy - you will need some other way to keep your data backed up - and you will need SLAs that work within the limits of the machine.
The 'default' Linux install is single machine - I have it running on 6 CPU & 4CPU machines. It defaults to a single node thor and roxie - that CAN have advantages (no skew etc) - although we have sometimes found that running multiple thor 'nodes' on a single machine (with multiple CPUs) can provide some lift too.
One thing you don't mention is your disks. Depending upon what you are doing Roxie can do quite a few disk hits - if you only have one or two disks then you will need to be a little careful in how you lay out your roxie keys - not a problem - but you need to plan.
The only other thing I can think of is that (obviously) only having one machine rather limits your redundancy - you will need some other way to keep your data backed up - and you will need SLAs that work within the limits of the machine.
- dabayliss
- Community Advisory Board Member
- Posts: 109
- Joined: Fri Apr 29, 2011 1:35 pm
3 posts
• Page 1 of 1
Who is online
Users browsing this forum: No registered users and 1 guest