Hi Guys
( July 2014) My knowlege on the matter has matured over the span on year. Here is link for article on Shared nothing (Teradata ) and Shared disk (Oracle RAC) architure difference article. You can read the below article, Its basic level on different types of architure for more matured reading check latest article.
( July 2014) My knowlege on the matter has matured over the span on year. Here is link for article on Shared nothing (Teradata ) and Shared disk (Oracle RAC) architure difference article. You can read the below article, Its basic level on different types of architure for more matured reading check latest article.
Now so many people have migrated to teradata they say its great for terabytes of information.So i am plannign to write some basic points on teradata.
If you want to compare Oracle and Teradata compare Oracle Exadata with Teradata. Do not compare Oracle RAC with teradata.
The basis difference between Oracle RAC and Teradata is architecture they use. There are two kinds of architecture Shared Nothing and Shared everything. Oracle has shared everthing and Teradata has shared nothing. So whatever the strength of their architecture that is their strength.
Netezza has share nothing architecture like teradata.
Share nothing is like MPP and Shared everything is a SMP. Your dual core processor is a SMP it has 2 processors inside one single intel chip and both processors share same RAM and same hard disk. So disadvantage of this system would be your hard disk speed. Though your processors can do a lot you are limited by hard disk access speed and RAM space. So even in case of parallel processing its not so fast. Remember even if you are only inserting data into database it will check for consitency that is to check that data has not changed so for inserting as well there is overhead for reading.SMP also has some overhead of making sure what other processors are doing.
Share nothing is a MPP. Each processor has its dedicated RAM and dedicated hard disk. It has some overhead with coordination among processors. Apart from that major drawback is data shipping or distribution of data across disk of each processor.
MPP works well for datawarehousing or OLTP the overhead of distributing data across disk is not worth it if you only retrieving single records. MPP is like a train. Good for Throughput. Its not a sports car where speed is fast for retrieval. But if you are retrieving huge data like train pulling lot of boggies you will see train is faster for that load.
But this does not mean Teradata is better that Oracle for datawarehousing. Its not that simple. Check ASK tom for what he says on this. (its not simple as to direct comparison)
SMP appears to use a single disk with help of RAID. Almost all servers use RAID configuration to store your data. Read for basics on RAID
One major difference in Shared Everything is its a SMP (Symmetric multiprocessor ) system. That is all processors use same main memory (RAM) . So its like your intel i5 on your laptop. Shares the same RAM with all its 4 processors inside.
But in a MPP system like teradata. Each processor has its own dedicated main memory (RAM)
Difference between Shared nothing and Shared Everything
http://cognossimplified.blogspot.in/2013/08/basics-about-servers.html
Oracle Exadata uses Infiband to connect between server and disks. It does not use a fibre channel.Fibre channel has a speed of upto 16 Gigabits per second but Infiband has (best one) 300Gigabits per seconds .Your hard disk has rate of 500 giba bits per seconds. So its almost like directly accessing from drive there is no network delay for data to come to server from hard disk. Here hard disk are not part of node(server--only RAM and processor) its share everything architecture
Oracle Exadata is closest to shared nothing . But is not our typical shared nothing with each cpu having its own memory (RAM) .
Very Few database are MPP .Teradata , HP Vertiga , Kognito , Netezza
How hardware VT works---BIOS has VT enable for virtulization of processor
In computing, x86 virtualization is the facility that allows multiple operating systems to simultaneously share x86 processor resources in a safe and efficient manner, a facility generically known as hardware virtualization
In protected mode the Kernel runs at a higher privilege such as ring 0, and applications at a lower privilege such as ring 3. Similarly, a host OS must control the processor while the guest OSs are prevented from direct access to the hardware. One approach used in x86 software-based virtualization is called ring deprivileging, which involves running the guest OS at a ring higher than 0.[1]
How hardware VT works
default username password is root root
To say you need to migrate to Teradata from your usual Oracle RAC. Have you experienced any of the below( I have found anyone seeing a issue and then migrating)
1) Synching of Buffer cache between different RAC nodes causing bottlenecks ?
2) High waits on accessing disk blocks- This can be due to any other reason too ?
3) Bottleneck caused by interconnect for disk access. ?
If you want to compare Oracle and Teradata compare Oracle Exadata with Teradata. Do not compare Oracle RAC with teradata.
The basis difference between Oracle RAC and Teradata is architecture they use. There are two kinds of architecture Shared Nothing and Shared everything. Oracle has shared everthing and Teradata has shared nothing. So whatever the strength of their architecture that is their strength.
Netezza has share nothing architecture like teradata.
Share nothing is like MPP and Shared everything is a SMP. Your dual core processor is a SMP it has 2 processors inside one single intel chip and both processors share same RAM and same hard disk. So disadvantage of this system would be your hard disk speed. Though your processors can do a lot you are limited by hard disk access speed and RAM space. So even in case of parallel processing its not so fast. Remember even if you are only inserting data into database it will check for consitency that is to check that data has not changed so for inserting as well there is overhead for reading.SMP also has some overhead of making sure what other processors are doing.
Share nothing is a MPP. Each processor has its dedicated RAM and dedicated hard disk. It has some overhead with coordination among processors. Apart from that major drawback is data shipping or distribution of data across disk of each processor.
MPP works well for datawarehousing or OLTP the overhead of distributing data across disk is not worth it if you only retrieving single records. MPP is like a train. Good for Throughput. Its not a sports car where speed is fast for retrieval. But if you are retrieving huge data like train pulling lot of boggies you will see train is faster for that load.
But this does not mean Teradata is better that Oracle for datawarehousing. Its not that simple. Check ASK tom for what he says on this. (its not simple as to direct comparison)
SMP appears to use a single disk with help of RAID. Almost all servers use RAID configuration to store your data. Read for basics on RAID
One major difference in Shared Everything is its a SMP (Symmetric multiprocessor ) system. That is all processors use same main memory (RAM) . So its like your intel i5 on your laptop. Shares the same RAM with all its 4 processors inside.
But in a MPP system like teradata. Each processor has its own dedicated main memory (RAM)
Difference between Shared nothing and Shared Everything
http://cognossimplified.blogspot.in/2013/08/basics-about-servers.html
Oracle Exadata uses Infiband to connect between server and disks. It does not use a fibre channel.Fibre channel has a speed of upto 16 Gigabits per second but Infiband has (best one) 300Gigabits per seconds .Your hard disk has rate of 500 giba bits per seconds. So its almost like directly accessing from drive there is no network delay for data to come to server from hard disk. Here hard disk are not part of node(server--only RAM and processor) its share everything architecture
Oracle Exadata is closest to shared nothing . But is not our typical shared nothing with each cpu having its own memory (RAM) .
Very Few database are MPP .Teradata , HP Vertiga , Kognito , Netezza
How hardware VT works---BIOS has VT enable for virtulization of processor
In computing, x86 virtualization is the facility that allows multiple operating systems to simultaneously share x86 processor resources in a safe and efficient manner, a facility generically known as hardware virtualization
In protected mode the Kernel runs at a higher privilege such as ring 0, and applications at a lower privilege such as ring 3. Similarly, a host OS must control the processor while the guest OSs are prevented from direct access to the hardware. One approach used in x86 software-based virtualization is called ring deprivileging, which involves running the guest OS at a ring higher than 0.[1]
How hardware VT works
default username password is root root
To say you need to migrate to Teradata from your usual Oracle RAC. Have you experienced any of the below( I have found anyone seeing a issue and then migrating)
1) Synching of Buffer cache between different RAC nodes causing bottlenecks ?
2) High waits on accessing disk blocks- This can be due to any other reason too ?
3) Bottleneck caused by interconnect for disk access. ?
No comments:
Post a Comment