Veeam Cisco Repository specs

[bcampbell]

We are getting ready to purchase our Veeam Repositories from our Vender. We went with a Cisco s3620. The vender has quoted us using 2 1TB NVME Drives in each server. I can seem to find any requirements for these drives. Is anyone using these in their 3620 Repository and what for? We have 2 ssds for the OS included as well. I just started at the company and this is an ongoing project. I am just overlooking the quote and making sure we don't pay for something we don't need.

[Gostev]

These are to put in RAID1 and use for OS, instant recovery write cache and guest file system catalog. You can go with regular SSDs, but I do recommend at least 512GB (however, 1TB is optimal because it gives you more headroom for long-running instant VM recoveries).

[bcampbell]

Apparently cisco doesn't make the 500g nvme anymore so it has to be 1TB.

But we are quoted both the NVME drives and the 480GB Boot SSDs. Are the ssds for the UCS OS?

I guess i'm just confused why the quoted both.

[tsightler]

Some people like to have the NVMe as dedicated cache since they have much higher throughput and lower latency vs the boot SSDs. I've seen people use the NVMe for all kinds of things, instant recovery write cache or file index, as Gostev mentioned, but I've also seem some people share it out with Windows NFS server, or via iSCSI, and use it as a super fast datastore to store templates or even to press into service as an emergency datastore if there is a disaster with the primary storage.

[Gostev]

I guess i'm just confused why the quoted both.

Me too... looks like an overkill to have a separate drive just for OS. Go with either 2 SSD, or 2 NVMe.

[bcampbell]

Some people like to have the NVMe as dedicated cache since they have much higher throughput and lower latency vs the boot SSDs. I've seen people use the NVMe for all kinds of things, instant recovery write cache or file index, as Gostev mentioned, but I've also seem some people share it out with Windows NFS server, or via iSCSI, and use it as a super fast datastore to store templates or even to press into service as an emergency datastore if there is a disaster with the primary storage.

Thanks Those are some very good ideas. I never really thought of that.