This is so well done. It is rare to find an IT presenter on RU-vid who is personable and can present well, has a good command of the English language, and covers so much at a high-level without losing the non-technical audience. I sincerely thank you
Very well presented. You can see that this talk was created with care for the quality and to serve the audience. This is what tech talks should strive to be.
This tutorial is incredible. The quality of the presentation is great. My one and only critique is the way the slide jumps up and down in the video. The information and presentation are A1 though.
This was very well done. In depth presentation and Q&A. Concise explanations across a broad range of important topics. Overall, well thought out presentation.
for the record, purge is done with truncates in oracle ( which is a smallish update to the data dictionary) - provided you know how to partition your table appropriately. but good talk overall.
Patrick McFadin , can you please post links to your 3 hr data model video you mentioned in this video ? Tried to find but there seem to be bunch of them and not sure which are dups..
keyword 'written' if I'm listening to an hour long plus presentation I don't want to be straining to understand, I'm sure such a presentation could have the most awesome content and the speaker could have phD's coming out of the wazoo but verbal communication is a skill unto itself
Nice video, but a couple of questions here 1. The CQL queries show Tables, isn't the equivalent in Cassandra called a Keyspace? 2. I'm a little confused about the sequential read and write and the partitioning using a hash function? How can a sequential write/read happen when the data is partitioned using a hash function of the primary key?
wow. this was an interesting and informative presentation. learned a couple of new things. keep up the good work. I gotta watch another one of these videos now. :)
This is a great talk, however I do not agree about the comments on High Availability. If you are using Oracle RAC or Exadata, where you have multiple databases in the cluster, you don't have a issue taking one node down for maintenance. One of our production nodes was restarted during the day, and no impact to the users or apps. If you are running a single instance, single node Oracle database, yes you have single point of failure. So, choose the right product feature or option for comparison. Cannot use a generic statement, as if Cassandra is the only product that offers High Availability feature that allows you to patch while database is available.
I disagree to this point oracle has its on pain points to on very rollback segments on racks. Its not possible to add more than 3 rack nodes and get better performance on busy DB clusters.
Which is the better way to install Cassandra, whether from a docker container or from a Linux machine (We use clustering or will make multiple instances)
As far 1:02:00 is concerned, Even though we never feel it, most of the Facebook transactions replicated across geography are eventual consistence, never perfectly consistence.
why an Oracle DBA needs to ask for Downtime to patch? Won't RAC databases address this issue? The use case for Cassandra in comparison to RDBMS could have been clearer. How does Cassandra differ from IBM's solution like DB2 Database partition Feature?
+Balachandran Chandrasekaran: DB2 Partition ??? Scale out Architecture DR HA are all plugged in a Hadoop echo system while you will have to manage that for a Shard DB oracle / DB2.
Pretty interesting watching in 2021 with the rise of fast and cheap NVME storage invalidating many of the assumptions that guided some of cassandra's design decisions. In software, the only constant is change
Nice vid! You are still talking about sequential IO, but that doesn't really matter anymore when everybody is using SSD's (in data centers)? 2 people doesn't like either Facebook, Netflix, eBay or Twitter ;)
You are right about SSDs and this is why we recommend them. There is still a bit of overhead on the transport layer when issuing random seeks. If you were to ask for a contiguous set of blocks and compared that with something issuing pure random reads, the random reads will have a much higher 95th percentile. Because of the sequential write IO, you'll find that Cassandra servers are much nicer on SSDs. Another reason we recommend MLC vs SLC drives. You just don't need the extra protection.
'They' say that SLC is better, faster and more reliable. But SLC is also more expensive. However do you think that MLC are better in data-centers, even considering endurance (life span) of the SLC's. Smaller nanometers NAND chips lead to a lower life cycle.
