!fImemMAwUAGwrlOcWH:matrix.org

OpenLMIS-General

81 Members
34 Servers

Load older messages


SenderMessageTime
2 Aug 2020
@lara:fairydust.spacelara joined the room.11:52:57
@lara:fairydust.spacelara 16:54:58
@josef:netzgemeinde.eujosef 17:11:56
3 Aug 2020
@erich:matrix.kiwifarms.neterich 20:12:18
@joey:halogen.cityjoey joined the room.21:11:31
4 Aug 2020
@_slack_openlmis_U195RKK4G:matrix.orgsbrudzinski changed their profile picture.11:33:48
@_slack_openlmis_U195RKK4G:matrix.orgsbrudzinski Charitable non-profit organizations that are non-government, non-academic, non-commercial in nature, and have no religious affiliation qualify for a 75% discount for Atlassian Cloud products (including add-ons, like Zephyr) - have we explored whether VR could apply for one to reduce recurring Jira costs? https://www.atlassian.com/software/views/community-license-request cc Wes 11:33:48
@_slack_openlmis_UBB6KN8JJ:matrix.orgWes That's a great question... joshzamor, I assume that you would know if we have tried for this? 11:34:36
@_slack_openlmis_U063EBB47:matrix.orgbenleibert We benefit significantly from Atlassian's generous non-profit pricing. Unfortunately, it doesn't extent to marketplace products, though, and our monthly costs are therefore still pretty high. 15:10:14
@_slack_openlmis_U195RKK4G:matrix.orgsbrudzinski benleibert they extended the 75% discount to Cloud marketplace app this month 15:11:03
@_slack_openlmis_U195RKK4G:matrix.orgsbrudzinski so if you are already on a community license, the discount will be automatically applied 15:11:21
@_slack_openlmis_U063EBB47:matrix.orgbenleibert Sorry - I had missed that and it's awesome news. Thank you very much, sbrudzinski! 15:12:34
@_slack_openlmis_U063EBB47:matrix.orgbenleibert We're indeed already signed up as a non-profit. 15:12:55
@_slack_openlmis_U195RKK4G:matrix.orgsbrudzinski Awesome! The Zephyr cost shouldn't hurt as much anymore 🙂 15:13:37
@_slack_openlmis_U195RKK4G:matrix.orgsbrudzinski AWS Savings Plan (Compute Savings Plans) can help us greatly reduce AWS infrastructure costs (no need to commit to specific instance family) - https://docs.aws.amazon.com/savingsplans/latest/userguide/what-is-savings-plans.html in exchange for usage commitment for at least the next year. With Reserved Instances we could save even more, but that requires a commitment to a specific instance (size, family). We have just set that up at SolDevelo to reduce costs greatly. I think it would be worth introducing the same in core, and all our implementations - that should help us save several hundred $ each month joshzamor benleibert kpalkowska Aleksandra Sołtys jjakubowski 15:16:07
@_slack_openlmis_U063EBB47:matrix.orgbenleibert Thank you for being so on top of this Sebastian - it's much appreciated. It really is great to know about Atlassian's new pricing! 🙂 We've been trying to use reserved instances and/or savings plans for a while. Our track record is pretty mixed, in that we're still paying for lots of on-demand capacity, but we're trying to avoid it where possible. Other ways in which we have or are starting to par down costs include: • Use of standard postgres rather than RDS. The latter is significantly more expensive to run and a hassle to decommission. It's not uncommon for implementations to want to turn off their dev server between lengthy periods of development. Although AWS allows RDS to be turned off, they automatically turn it back on every 7 days. Runtime cost even in production, though, is the main disadvantage. • Shifting away from NiFi/Superset in some cases. This strategy may be controversial, but the Superset stack has turned out to be very costly both to host and maintain. Angola, for instance, is currently slated to spend around $600/month to host the component... not counting the cost of tickets like MALAWISUP-2072 (which happens to be open at the moment). Mozambique, in particular, is working to replace the PAV Dashboard, which had recently been developed using Superset, within an alternative based on Power BI. They're similarly planning to use Power BI rather than Superset within their instance of OpenLMIS v3. (This topic is worthy of its own thread. The point here is just that Superset is expensive and not quite a one-size-fits-all solution.) 16:48:49
@_slack_openlmis_U063EBB47:matrix.orgbenleibert Hi sbrudzinski. AWS also offers non-profits $2,000/year worth of credit, while Azure gives away $3,500/year. Although the Azure sponsorship may be ending, they're both worth mentioning to any other non-profits you work with. 23:59:12
5 Aug 2020
@_slack_openlmis_B7JF9KDNK:matrix.orgtwitter @OpenLMIS: We are proud to support the #DigitalPrinciples, 9 living guidelines for #digitaldevelopment, and are excited to share the @digiprinciples new podcast series: Pulse on the Principles. Subscribe to hear experts discuss the intersection of tech & development: https://t.co/GGgo2zjKWc 09:22:52
@_slack_openlmis_UBB6KN8JJ:matrix.orgWes
In reply toundefined
Is there anything we need to do to update our account for Zephyr?
10:42:53
@_slack_openlmis_UBB6KN8JJ:matrix.orgWes
In reply toundefined
(edited) ... for Zephyr => ... for Zephyr?
10:43:00
@_slack_openlmis_U03QATSPZ:matrix.orgjoshzamor
In reply to@_slack_openlmis_U195RKK4G:matrix.org
AWS Savings Plan (Compute Savings Plans) can help us greatly reduce AWS infrastructure costs (no need to commit to specific instance family) - https://docs.aws.amazon.com/savingsplans/latest/userguide/what-is-savings-plans.html in exchange for usage commitment for at least the next year. With Reserved Instances we could save even more, but that requires a commitment to a specific instance (size, family). We have just set that up at SolDevelo to reduce costs greatly. I think it would be worth introducing the same in core, and all our implementations - that should help us save several hundred $ each month joshzamor benleibert kpalkowska Aleksandra Sołtys jjakubowski
Often when this is brought up, the issue is that many of our grants don't have a 1 year or + term, and we're very hesitant to make the commitment. It's a good idea though.
17:23:41
@_slack_openlmis_U03QATSPZ:matrix.orgjoshzamor
In reply to@_slack_openlmis_U063EBB47:matrix.org
Thank you for being so on top of this Sebastian - it's much appreciated. It really is great to know about Atlassian's new pricing! 🙂 We've been trying to use reserved instances and/or savings plans for a while. Our track record is pretty mixed, in that we're still paying for lots of on-demand capacity, but we're trying to avoid it where possible. Other ways in which we have or are starting to par down costs include: • Use of standard postgres rather than RDS. The latter is significantly more expensive to run and a hassle to decommission. It's not uncommon for implementations to want to turn off their dev server between lengthy periods of development. Although AWS allows RDS to be turned off, they automatically turn it back on every 7 days. Runtime cost even in production, though, is the main disadvantage. • Shifting away from NiFi/Superset in some cases. This strategy may be controversial, but the Superset stack has turned out to be very costly both to host and maintain. Angola, for instance, is currently slated to spend around $600/month to host the component... not counting the cost of tickets like MALAWISUP-2072 (which happens to be open at the moment). Mozambique, in particular, is working to replace the PAV Dashboard, which had recently been developed using Superset, within an alternative based on Power BI. They're similarly planning to use Power BI rather than Superset within their instance of OpenLMIS v3. (This topic is worthy of its own thread. The point here is just that Superset is expensive and not quite a one-size-fits-all solution.)
benleibert: How did Angola get to $600/mo?
17:26:09
@_slack_openlmis_U063EBB47:matrix.orgbenleibert
In reply to@_slack_openlmis_U03QATSPZ:matrix.org
benleibert: How did Angola get to $600/mo?
Hi joshzamor. I’m sorry about that – I meant to mention Malawi alongside MALAWISUP-2072. Including sales tax, their Superset related costs are likely closer to $650/month or, as I recall, about 1/3 of their overall AWS expense. This is just an approximation based on their Superset related EC2/RDS costs. For comparison, hosting the Superset based PAV Dashboard costs around $420/month including tax. Angola (by which I actually mean “Angola”) has Superset related costs of around $130/month, or about 17% of their total AWS expense. Because their system sees negligible load at the moment, though, their costs are of limited use for forecasting. Malawi’s instance is far more representative. Hopefully that helps. : ) I didn’t set out to do a careful analysis, but think the above numbers (and, this time, country names!) should be close enough for this thread’s general discussion.
19:10:54
@_slack_openlmis_U03QATSPZ:matrix.orgjoshzamor
In reply to@_slack_openlmis_U063EBB47:matrix.org
Hi joshzamor. I’m sorry about that – I meant to mention Malawi alongside MALAWISUP-2072. Including sales tax, their Superset related costs are likely closer to $650/month or, as I recall, about 1/3 of their overall AWS expense. This is just an approximation based on their Superset related EC2/RDS costs. For comparison, hosting the Superset based PAV Dashboard costs around $420/month including tax. Angola (by which I actually mean “Angola”) has Superset related costs of around $130/month, or about 17% of their total AWS expense. Because their system sees negligible load at the moment, though, their costs are of limited use for forecasting. Malawi’s instance is far more representative. Hopefully that helps. : ) I didn’t set out to do a careful analysis, but think the above numbers (and, this time, country names!) should be close enough for this thread’s general discussion.
I think there's more here than I need to understand, as in I'm a little lost about the Angola thing, but that's alright. My question really is that in design we wanted to target about $300 / mo for the entire stack. It sounds like Malawi is about double that, right? My question is really a what question: what doubled the costs? Do they have high load and so need bigger instances (vertical scaling), did they need to more features and so have more services than the standard reporting stack?
23:08:43
6 Aug 2020
@_slack_openlmis_U063EBB47:matrix.orgbenleibert
In reply to@_slack_openlmis_U03QATSPZ:matrix.org
I think there's more here than I need to understand, as in I'm a little lost about the Angola thing, but that's alright. My question really is that in design we wanted to target about $300 / mo for the entire stack. It sounds like Malawi is about double that, right? My question is really a what question: what doubled the costs? Do they have high load and so need bigger instances (vertical scaling), did they need to more features and so have more services than the standard reporting stack?
Hi joshzamor. To the best of my knowledge, our implementations haven't added features to the reporting stack which would impact performance. Instead, their costs are driven by usage and, in turn, the degree to which they have to vertically scale.
00:35:49
7 Aug 2020
@paul:converser.eupaul 00:15:07
@nicolas:cybre.spacenicolas 09:41:17
8 Aug 2020
@joe:hope.netjoe 21:52:07
10 Aug 2020
@herbert:synod.imherbert joined the room.21:10:34
11 Aug 2020
@_slack_openlmis_B7JF9KDNK:matrix.orgtwitter @OpenLMIS: We created a new, simplified configuration of OpenLMIS that can be quickly deployed for the integrated management of COVID-related supplies @VillageReach @CHAI_health https://t.co/jnucOuw2sF 07:51:46

There are no newer messages yet.


Back to Room List