estimate number of keys on rocks db

[avilevi]

Hi,

I am trying to estimate number of keys at a given minute. 

I created a graph based on avg_over_time with 1hr and 5m interval. looking at the graph you can see that it has high spikes which doesn't make sense (IMO) how can the average have those spikes ? after all since I do not delete keys I would expect to go up or remain the same. 

any ideas what can explain such behaviour ?

attached are graphs 5m and 1 h intervals 

[Yun Tang]

Hi Avi

Unfortunately, we cannot see the attached images. By the way, did you ever use window in this job?

Best

Yun Tang

---

From: Avi Levi <[hidden email]>
Sent: Sunday, March 10, 2019 19:41
To: user
Subject: estimate number of keys on rocks db

 

Hi,

I am trying to estimate number of keys at a given minute. 

I created a graph based on avg_over_time with 1hr and 5m interval. looking at the graph you can see that it has high spikes which doesn't make sense (IMO) how can the average have those spikes ? after all since I do not delete keys I would expect to go up or remain the same. 

any ideas what can explain such behaviour ?

attached are graphs 5m and 1 h intervals 

[avilevi]

Thanks Yun,

Attached. Please let me know if it is ok.

I made several trials including aggregation functions but couldn't figure out why the line is not going straight up and why having those picks . 

On Sun, Mar 10, 2019 at 4:49 PM Yun Tang <[hidden email]> wrote:

Hi Avi

Unfortunately, we cannot see the attached images. By the way, did you ever use window in this job?

Best

Yun Tang

---

From: Avi Levi <[hidden email]>
Sent: Sunday, March 10, 2019 19:41
To: user
Subject: estimate number of keys on rocks db

 

Hi,

I am trying to estimate number of keys at a given minute. 

I created a graph based on avg_over_time with 1hr and 5m interval. looking at the graph you can see that it has high spikes which doesn't make sense (IMO) how can the average have those spikes ? after all since I do not delete keys I would expect to go up or remain the same. 

any ideas what can explain such behaviour ?

attached are graphs 5m and 1 h intervals