This post is more than 5 years old
2 Posts
0
1417
July 18th, 2016 14:00
Considering SC4020 all flash
Hi,
we are considering a SC4020 all flash array. We got a quote with 6x 1.6TB WI and 12x 3.8TB RI TLC. They told us, that the tier 1 should be sized to hold about 12 days of writing data.
What I understand in the white paper "Flash-optimized Data Progression" is, that beginning with SC OS 6.4 the data will be moved to SSD RI tier after every replay taken. So if we take a replay every 4 hours, we do not need such a large tier 1 with WI SSD. Is this correct?
And a second question: We got also a quote about a single tier SC4020 with RI SSD only (24x 1.9TB), which fits better in our budget. Is there a reasonable maximum size of daily data to be written?
Thanks!
No Events found!
BVienneau
115 Posts
1
July 20th, 2016 08:00
That's pretty old information. They're saying that because the original data progression works on a 12 days cycle for "accessible" data. Inaccessible (data that got overwritten but still needs to be around) data would get pushed immediately on next "original" data progression.
Also, you are right in that ODDP (flash optimized data progress) is MUCH more quick to move data between tiers especially in Flash. In your WI & RI proposed configuration, you want to size the WI tier for the write changes between the expected Replay schedule OR roughtly 24 hours. You can get the 24hr number from DPACK and divide that by the Replay schedule if you are taking them more than every 24hrs. Your thinking is correct here!
Also, I would personally go with the single tier option unless you have extremely high write data that necessitates the WI drives. A single tier RI tier is good for "most" customers, but your DPACK data should dictate that.
MKo24
2 Posts
0
July 20th, 2016 10:00
Thank's a lot!
DPACK shows about 400G daily volume. If I calculate 800G (growth) x 2 (raid 10), I will need 1.6TB if I do one replay per day. But because we will replicate to our Equallogic, we will take every 4 hours a replay.
I will talk to the sales man to calculate with 6x 400G - it will be more than enough :-)
Because the 1.9TB RI MLC will IMHO have about 8MB block size, I don't like the idea writing directly to the SSD (my gut feeling), because our 0.4TB will be written in Raid 10DM (= 1.2TB) and copied to R6 (0.5TB). The sum is about 1.7TB per day without the write amplification, if stripe size is smaller than the SSD block size...
rswislocki
52 Posts
0
August 23rd, 2016 07:00
Fell free to run additional replays on tier1 volume when Tier1 is near full. After replay all frozen data move by data progression mechanism to lower tier.
If you did not create reply and tier1 reaches 100% data progression on demand starts movin data to lower tier.
(Above features need Data Progression License)