3 ms·

with cloud being the talk of the industry these days. how pervasive are SRE roles. I would wager a good system admin | SRE can save a company thousands if they

by dzonga 6y ago

with cloud being the talk of the industry these days. how pervasive are SRE roles. I would wager a good system admin | SRE can save a company thousands if they provisioned their own hardware i.e colo instead of cloud instances etc

chucky_z 6y ago
For LinkedIn that number can be 7+ figures saved per year by a single employee.
xionon 6y ago
You still need SREs for using cloud services at large scale, and they can manage more cloud infrastructure per person than if it was colocated
_jal 6y ago
Yep. We run almost all our systems on our own hardware. (We do use cloud VMs for random things.) It is much cheaper. Running our primary DB cluster in AWS for a month costs more than the purchase price of the hardware the cluster runs on.
unethical_ban 6y ago
Do you think SRE talent isn't needed for cloud? There is an incredible number of metrics around every AWS service, and the decoupling of things via queues, topics, APIs etc. mean there are a lot of moving parts.
arcturus17 6y ago
They could also cost the company ludicrous amounts if they built a Frankenstein at the software layer which made a mess of operations, security or performance. It’s unlikely that a “good” sysadmin / SRE, or even a team of them, could efficiently replicate the versatility and reliability offered by cloud services.
- jonfromsf 6y ago
  Thats true. But you don't need that versatility or reliability a lot of the time. For a bog-standard webapp that talks to a database, gets decent traffic, and is growing at a reasonable pace (say doubling every year), dedicated hosting or colo will save you a LOT of money.
paxys 6y ago
Considering LinkedIn is owned by Microsoft they don't exactly need their own hosting.
- ojbyrne 6y ago
  Eventually they won't: https://engineering.linkedin.com/blog/2019/building-next-infra https://engineering.linkedin.com/blog/2019/building-next-inf...
NineStarPoint 6y ago
Well, given how expensive cloud services are compared to provisioning hardware, a good SRE can save a ton of money in just utilizing cloud services more effectively. The decision for where the hardware will exist may be advised by SREs, but is generally not exactly in their wheelhouse. SREs tend more towards lowering the cost and maximizing the utilization of whatever methods the company as a whole decides to use.
0xbadcafebee 6y ago
It's the opposite. A good SRE can build a system faster and cheaper in the cloud, and only self-host the things that make sense. You need a half dozen engineers working full time to replicate what the cloud gives you, and still buy the hardware ahead of time. That's 6 months of straight project development with 6 people and capital for a rack full of gear.. that's like $1.5M USD outlay for one year. For close to feature parity for core services I mean, nobody would normally build all that just to put a crappy web app online. OR hire one SRE to set up a cloud stack in two months and only pay for what you use, $200K for one year or $100K for a short contact.

by dzonga 6y ago

chucky_z 6y ago

For LinkedIn that number can be 7+ figures saved per year by a single employee.

xionon 6y ago

You still need SREs for using cloud services at large scale, and they can manage more cloud infrastructure per person than if it was colocated

_jal 6y ago

Yep. We run almost all our systems on our own hardware. (We do use cloud VMs for random things.) It is much cheaper. Running our primary DB cluster in AWS for a month costs more than the purchase price of the hardware the cluster runs on.

unethical_ban 6y ago

Do you think SRE talent isn't needed for cloud? There is an incredible number of metrics around every AWS service, and the decoupling of things via queues, topics, APIs etc. mean there are a lot of moving parts.

arcturus17 6y ago

They could also cost the company ludicrous amounts if they built a Frankenstein at the software layer which made a mess of operations, security or performance. It’s unlikely that a “good” sysadmin / SRE, or even a team of them, could efficiently replicate the versatility and reliability offered by cloud services.

jonfromsf 6y ago
Thats true. But you don't need that versatility or reliability a lot of the time. For a bog-standard webapp that talks to a database, gets decent traffic, and is growing at a reasonable pace (say doubling every year), dedicated hosting or colo will save you a LOT of money.

paxys 6y ago

Considering LinkedIn is owned by Microsoft they don't exactly need their own hosting.

ojbyrne 6y ago
Eventually they won't: https://engineering.linkedin.com/blog/2019/building-next-infra https://engineering.linkedin.com/blog/2019/building-next-inf...

NineStarPoint 6y ago

Well, given how expensive cloud services are compared to provisioning hardware, a good SRE can save a ton of money in just utilizing cloud services more effectively. The decision for where the hardware will exist may be advised by SREs, but is generally not exactly in their wheelhouse. SREs tend more towards lowering the cost and maximizing the utilization of whatever methods the company as a whole decides to use.

0xbadcafebee 6y ago

It's the opposite. A good SRE can build a system faster and cheaper in the cloud, and only self-host the things that make sense. You need a half dozen engineers working full time to replicate what the cloud gives you, and still buy the hardware ahead of time. That's 6 months of straight project development with 6 people and capital for a rack full of gear.. that's like $1.5M USD outlay for one year. For close to feature parity for core services I mean, nobody would normally build all that just to put a crappy web app online. OR hire one SRE to set up a cloud stack in two months and only pay for what you use, $200K for one year or $100K for a short contact.