Increased overall cost after migrating to H/R datastore?

[Waleed Abdulla]

I'm considering upgrading from M/S to HR datastore, so I compared the latency behavior of both as reported by App Engine system status page. Based on yesterday's graphs (below).

              M/S        H/R

Get Average:  10ms       30ms

Get Spikes:   130ms      50ms

Put Average:  10ms       30ms

Put Spikes:   60ms       70ms

If I'm not mistaken, I should expect to pay significantly more if I upgrade because the average latency is 3 times more, which increases my instance hours. I will also see my pages render noticeably slower. Is that what those of you who migrated experienced? 

Waleed

[Kenneth]

Make the change, you won't regret it.  Unless perhaps you're doing 10-15 sequential writes to the datastore, which you should never be doing, you won't notice any difference.

[Jeff Schnitzer]

In *theory*, if you use python27 and threadsafe, increased latency won't affect instance hours because instances are limited by cpu consumption, not blocked calls.  I would be curious to know the quantitative extent to which this is true.

Jeff

--
You received this message because you are subscribed to the Google Groups "Google App Engine" group.
To post to this group, send email to google-a...@googlegroups.com.
To unsubscribe from this group, send email to google-appengi...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en.

[Brandon Wirtz]

Do you need your Puts to be Synchronous? I use Defer to do all my puts, so they essentially happen in 0ms, but they may not actually happen right away. As a result the latency doesn’t change my pricing.

But we keep an old MS and an Old HR app around.   The Average on the HR is typically 2.5x the MS, but the spikes are FAR less.  We see MS Gets and Puts rise to 500+ ms, but  HR I have never seen go over 120ms.

 

The 2.7 with Multi-threading will likely help you as well.

 

--