Printable View
I intend to replace the old disks to the passive node. When full volume replication will start again (about 8TB), I see no reason, to enable SSD Cache on both nodes - just increase wearout of SSD and reduced performance due to increased ROC load.
Any better idea?
I use SSD Cache on both nodes - SSDs for cache are "not enterprise" Samsung 850 PRO 512GB.
I would agree with this but the only caching is available for the DSS V7 is with a RAID controller for the LSI CacheCade or the Adaptec MaxCache. As a note for all when using these controllers for the caching please use 2 x SSDs and mirror them, I have seen customers only use 1 and if it fails then there can be data loss.
Yes, for each node I use 2 Samsung SSDs 850 Pro 512GB in RAID1.
512GB is also the largest CachaCade capacity :confused:.
For past 527 days (with 100,0% active cluster!) I used 2 Intel S3500 240GB SSDs in RAID1 for each node and SMART results are:
on "mostly active" node:
F1 155TB
F2 36TB
E9 54%
on "mostly passive" node:
F1 70B
F2 19TB
E9 78%
All these numbers are similar to the expected.
So, I hope for a "peaceful sleep" for the next two-three years.
Typical Intel's SSDs have a 5 year life but good that you have them mirrored but having a spare on hand is also a good idea :)
This 5 year warranty is unfortunate limited with Wearout indicator (or Endurance Rating for easier calculation).
When SSD is used in CacheCade, data is written on it when we write into Datastore and also when we read uncached data.
So we should calculate endurance with adding all traffic to & from datastore.
Intel S3500, for example, has Endurance Rating 140 TBW for 240GB but 275 TBW for 480GB model.
That is good example why to use larger SSD.
When two nodes are used in active/passive mode and they are almost identical, it is good to change activity, time to time, because reading (and writing to SSD) process is only on active node. There is good example on my data - compare F1 SMART readout.