Question LFF HDD for second-hand IBM server - - - SAS 12 vs WD DC HC520 ?

ali_143

Distinguished
Jun 2, 2016
30
1
18,535
Hello, i want 4 x LFF HDDs (with raid 5) for my server (database usage. I know I can use SSDs, but I need LFF for some reason. I have two options :

Option 1 :
4 x IBM 8 TB SAS 7.2K 12G (Refurbished)

Option 2 :
4 x WD DC HC520 12 TB SAS 12G (New)

I asked somewhere about those WD HDDs, and they said its speed is 256 Mb/s not 12Gbps .... is it true ? Which option is better for both speed and stability ?

My server is HPE Gen9 DL360 LFF
 
Hello, i want 4 x LFF HDDs (with raid 5) for my server (database usage. I know I can use SSDs, but I need LFF for some reason. I have two options :

Option 1 :
4 x IBM 8 TB SAS 7.2K 12G (Refurbished)

Option 2 :
4 x WD DC HC520 12 TB SAS 12G (New)

I asked somewhere about those WD HDDs, and they said its speed is 256 Mb/s not 12Gbps .... is it true ? Which option is better for both speed and stability ?

My server is HPE Gen9 DL360 LFF
12Gb is the SAS interface rate. Just as SATA is 6Gb. Spinning hard drives never saturate the interface, even on SATA.
RAID5 for a database is a bad implementation. Generally, multiple RAID1 volumes is a better choice. And I would only use RAID if this were a production machine and downtime is critical. Otherwise, backups are more important. RAID is an UPTIME enhancer, not a data security enhancer. One badly formed SQL statement and the data is gone. RAID doesn't protect against that. Backups do.
 
  • Like
Reactions: ex_bubblehead
12Gb is the SAS interface rate. Just as SATA is 6Gb. Spinning hard drives never saturate the interface, even on SATA.
RAID5 for a database is a bad implementation. Generally, multiple RAID1 volumes is a better choice. And I would only use RAID if this were a production machine and downtime is critical. Otherwise, backups are more important. RAID is an UPTIME enhancer, not a data security enhancer. One badly formed SQL statement and the data is gone. RAID doesn't protect against that. Backups do.
Why not raid 5 for database ???!!! RAID 5 is the best option for database because by breaking a drive, we will have no losing of data, and we can replace that drive with a new one and everything will be just fine ... making backup from a real time database (over 10000 queries in second) is not possible and the whole system goes down (making backup of TB data is not easy) ... also about speed, we just have 20% less speed than raid 1 ... Which is OK.
 
Why not raid 5 for database ???!!! RAID 5 is the best option for database because by breaking a drive, we will have no losing of data, and we can replace that drive with a new one and everything will be just fine ... making backup from a real time database (over 10000 queries in second) is not possible and the whole system goes down (making backup of TB data is not easy) ... also about speed, we just have 20% less speed than raid 1 ... Which is OK.
The RAID 5, or 1, or whatever...only protects against physical drive fail.
It does not protect the data.

And when rebuilding the array after a dead drive replacement, you have to take the whole thing offline anyway.

Any responsible person that runs their system on a RAID array also has a reliable backup.
 
The RAID 5, or 1, or whatever...only protects against physical drive fail.
It does not protect the data.

And when rebuilding the array after a dead drive replacement, you have to take the whole thing offline anyway.

Any responsible person that runs their system on a RAID array also has a reliable backup.
Well I will have over 80 TB new data in each month .... it's not possible to make backup !!!! Cost me too F much ... (The messages in a messenger app) ... Yes, for basic data like users and ..., it's a good thing to do replication ...
 
In this RAID5 array, if 2 or more drives or the controller were to suddenly fail (yes, it does happen) and your 80TB of data were to become unrecoverable, and you do not have a backup, what would be the consequences, and cost to your operations? There's no such thing as "....it's not possible to make backup !!!!"
 
Why not raid 5 for database ???!!! RAID 5 is the best option for database because by breaking a drive, we will have no losing of data, and we can replace that drive with a new one and everything will be just fine ... making backup from a real time database (over 10000 queries in second) is not possible and the whole system goes down (making backup of TB data is not easy) ... also about speed, we just have 20% less speed than raid 1 ... Which is OK.
RAID5 is bad for a database because of the overhead of calculating parity. It is slow. That is why RAID1 is preferred. There is no parity overhead. RAID1 allows failure of a drive and replacement, without losing data. You would have to benchmark rebuild times.
A 10,000 query / second database SHOULD have the critical data cached in RAM. RAM is a much more important attribute for a high performance database system.
 
Well I will have over 80 TB new data in each month .... it's not possible to make backup !!!! Cost me too F much ... (The messages in a messenger app) ... Yes, for basic data like users and ..., it's a good thing to do replication ...
80TB / Month requires 250+ Mb/sec of new data for the entire month. That amount of traffic should NOT be on a single server, IMO. This is a way bigger than a single server or RAID implementation.
 
Well I will have over 80 TB new data in each month .... it's not possible to make backup !!!! Cost me too F much ... (The messages in a messenger app) ... Yes, for basic data like users and ..., it's a good thing to do replication ...
80TB per month.
All on a single server?

What is this server doing?


I asked somewhere about those WD HDDs, and they said its speed is 256 Mb/s not 12Gbps .... is it true ? Which option is better for both speed and stability ?

My server is HPE Gen9 DL360 LFF
You're asking for hardware recommendations, and we have no idea about the workload, budget, or pretty much anything.
 
I asked somewhere about those WD HDDs, and they said its speed is 256 Mb/s not 12Gbps .... is it true ? Which option is better for both speed and stability ?

The interface speed is 12Gb/s. That's gigaBITS per second. The drive is potentially capable of peak sequential transfers (at the beginning of the disk) of around 256MB/s. That's MegaBYTES per second.

The WD 12TB is likely to be slightly faster and more reliable, on account of being bigger and newer. I'll let others address the other concerns.
 
The interface speed is 12Gb/s. That's gigaBITS per second. The drive is potentially capable of peak sequential transfers (at the beginning of the disk) of around 256MB/s. That's MegaBYTES per second.

The WD 12TB is likely to be slightly faster and more reliable, on account of being bigger and newer. I'll let others address the other concerns.
12 gigabits per second is equal to 1500 megabytes per second !!!!!!! How you calculate it, and you got 256 megabytes ??
 
12 gigabits per second is equal to 1500 megabytes per second !!!!!!! How you calculate it, and you got 256 megabytes ??

Your calculation is correct. 12Gb/s (1500MB/s) is the speed of the interface (between the drive and the PC). Few drives, especially hard drives, come close to saturating the interface. In practice, that hard drive will probably top out around 256MB/s, and that will only be for sequential reads/writes, near the beginning of the drive. Near the end of the drive, sequentials will probably hit only around 100MB/s. Random I/O will likely only be a few MB/s.
 
  • Like
Reactions: ali_143