Choosing an AWS instance for cassandra

Datastax recommends instance m3 for production clusters. And for the right reasons, he also mentions that EBS is not recommended. However, m3 instances don't have that much instance storage. m3.large - 32 GB, m3.xlarge - 80 GB

You cannot scale without EBS. Is the recommendation to maintain these storage levels in memory - m3.xlarge with 80gb for moderate production (with replications) or is there more to it?

Should we just consider m1 instances as they seem to be more economical to store at least?

Thank.

+3


source to share


2 answers


In my experience, you're better off working with spinning m1 / m2 drives than EBS or other shared NAS. Yes, the SSDs in the m3 instances are good, but they are tiny. If it doesn't make sense financially to scale - or to achieve the i2 level - your best bet on AWS is probably the old m1 / m2 series.



+4


source


we started with m1.xlarge just for the same reasons you explain, now we switched to m2.4xlarge and they work amazing (extra drum for heap size for big jobs in Apache sparks works like a charm)



SSDs are great, but m3 doesn't have large volumes, you should definitely be looking at m2 instead of i2 for cost reasons.

+1


source







All Articles